Neuroinflammation, Glutamate Regulation and Memory

Neuroinflammation, Glutamate Regulation and Memory.

DISSERTATION

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

in the Graduate School of The Ohio State University

Holly M. Brothers, M.A.

Graduate Program in Psychology

The Ohio State University

2013

Dissertation Committee:

Gary G. Berntson, John P. Bruno, Phillip G. Popovich, and Gary L. Wenk

Copyrighted by

Holly Marie Brothers

2013

Abstract

Neuroinflammation and excessive glutamatergic signaling have deleterious effects in the brain, are mutually promoting, and play a role in the onset and progression of neurodegenerative diseases. It was my goal to better understand the relationship between neuroinflammation, glutamate dysregulation and clinical symptoms, as well as identify potential therapeutic targets. To do this, I studied molecular, cellular and behavioral outcomes across various time points in young rats with experimentally-induced chronic neuroinflammation and older rats with age-associated neuroinflammation. I focused on

Alzheimer’s (AD) and Parkinson’s (PD), and tested only treatments that had therapeutic potential in a clinical setting. These studies led to a number of important discoveries, summarized below.

I first investigated the effects of reducing pre-synaptic glutamate release by caffeine treatment, and discovered that caffeine attenuated experimentally-induced but not age- associated neuroinflammation and was not sufficient to improve cognitive performance

(Chapter 2). During these investigations, I discovered that young rats with experimentally- induced neuroinflammation reared on their hind legs less, a behavioral change associated with PD. Therefore, I next expanded the scope of my studies to include a thorough investigation of brain systems that degenerate in PD, the basal ganglia and brainstem, using our model of chronic neuroinflammation (Chapter 3). I found that neuroinflammation drives changes that indicate a decline in neurotransmitter production and function, and that all these changes recovered over time, despite the continued presence of a ii neuroinflammatory stimulus. I paralleled these investigations in the hippocampus, a region that degenerates in AD, while gathering preliminary data on delayed treatment with memantine, a drug which reduces post-synaptic glutamate function that is used for AD

(Chapter 4). Like the basal ganglia and brainstem, I found compensatory recovery in hippocampal function over time, despite continued neuroinflammation, as revealed by recovery of cognitive performance in a hippocampal-sensitive behavioral task. These results were particularly interesting because delayed treatment with memantine, a drug which prevents AD-like cognitive decline under neuroinflammatory conditions, worsened performance in rats that would have otherwise recovered. This indicated that delayed reduction of post-synaptic glutamate interferes with a naturally occurring compensatory response. I speculated that this compensation was likely in the glutamatergic system and, if understood, could be augmented pharmacologically and become a therapeutic drug target for AD patients. Therefore, I investigated indicators of pre-, post- and extra-synaptic glutamatergic function, and discovered that recovery from AD-like memory decline correlated with a protein that clears excessive glutamate from the synapse, GLT1 (Chapter

5). In response to this discovery, I investigated whether two drugs that increase glutamate clearance would prevent AD-like cognitive decline in aged rats or young rats with experimentally-induced neuroinflammation (Chapter 6), and found that the drugs had modest, but beneficial effects on neuroinflammation and cognition.

This document is a compilation of the projects outlined above (Figure 1). Chapter

1 is a thorough review of background material, Chapters 2-6 describe each project in a journal article format, Chapter 6 contains the final series of studies described in my

iii candidacy proposal, and Chapter 7 briefly summarizes the knowledge gained and highlights my contribution to the field.

Figure 1. Thesis overview Chapters 2-6 examine microglia activation in a model of chronic neuroinflammation (Chapters 2-6) and in natural aging (Chapters 2 and 6). Chapter 2 investigates modulation of pre-synaptic glutamate release by caffeine, Chapter 3 examines neuroinflammation in the midbrain and brainstem, Chapter 4 explores attenuation of post- synaptic NMDAR activity by delayed memantine treatment, Chapter 5 demonstrates compensatory changes in extra-synaptic glutamatergic regulation in response to chronic neuroinflammation and Chapter 6 investigates augmentation of glutamate clearance by Ceftriaxone and Riluzole.

Dedication

To my great-grandma who inspired me to begin this work, to my grandpa who inspires me to do this work and to my daughter Adelina who inspires me to continue this work. Adelina, I hope you will follow your dreams and do what makes you happy. Grandma and Grandpa, thanks for always giving me so much love. Mom, thanks for believing that my work will make a difference. Gram and Dad, thanks for the moxie. Scott, thanks for being one of my best friends. John Bidwell, your curiosity sparks mine. Thank you to the Rosses for making me part of your family, Javy and Adelina for being my family, the Sotos for joining our family and thank you Isa, Roxanne and Sarah for making our lab like a family.

Acknowledgments

Dr. Gary Wenk thank you for you incredible mentorship, I appreciate both your confidence and your criticisms greatly. Thanks Dr. Yannick Marchalant, Dr. Francesca Cerbai, Dr. Patricia Fernanda Schuck, Dr. Gustavo Ferriera, Dr. Åsa Konradsson-Geuken, Dr. Isabelle Bardou, Sarah Hopp, Roxanne Kaercher, Sarah Turner, Mollie Mitchem, Clelland Gash and David Bortz for teaching me and helping me to complete my experiments. Thanks Dr. Katrina Paumier and Dr. Daniel Ankeny for demonstrating stereological estimation, Dr. Kristina Kigerl for helping me test LPS on microglia culture, Dr. Angela Wynne Corona and Ashley Fenn for helping to test microglia using flow cytometry and Dr. Glenn Lin for sharing control tissue for GLT1 staining. Thanks to Dr. John Bruno, Dr. Jonathan Godbout and Dr. Phillip Popovich for sharing their equipment, students and post-docs.

Vita

2001...... McKinley Sr. High School, Honors

2005...... Bachelor of Science, Honors in Arts and

Sciences, Distinction in Psychology, Minor

in Neuroscience, Magna Cum Laude,

The Ohio State University

2005...... Phi Beta Kappa

2006 ...... University Fellowship,

The Ohio State University

2010 ...... Master of Arts, Department of Psychology,

The Ohio State University

2010 ...... PhD Candidate, Department of Psychology,

The Ohio State University

2011 ...... University Presidential Fellowship,

The Ohio State University

Publications

Cerbai F, Lana D, Nosia D, Petkova-Kirovab P, Zecchi S, Brothers HM, Wenk GL and

Giovannini MG (2012). The neuron-astrocyte-microglia triad in normal brain

vii

ageing and in a model of neuroinflammation in the rat hippocampus. PLOS1,

7(9): e45250.

Bardou I, DiPatrizio N, Brothers HM, Kaercher R, Baranger K, Mitchem M, Hopp SC,

Wenk GL, Marchalant Y (2012). Pharmacological manipulation of cannabinoid

neurotransmission reduces neuroinflammation associated with normal aging.

Health, 1(4): 1-6.

Norman GJ, Morris JS, Karelina K, Weil ZM, Zhang A, Al-Abed Y, Brothers HM,

Wenk GL, Pavlov VA, Tracey KJ, DeVries AC (2011). Cardiopulmonary Arrest

and Resuscitation Disrupts Cholinergic Anti-inflammatory Processes: A Role for

Cholinergic α7 Nicotinic Receptors. J Neurosci, 31(9): 3446-52.

Brothers HM, Marchalant Y, Wenk GL (2010). Caffeine attenuates lipopolysaccharide-

induced neuroinflammation. Neuroscience Letters, 480(2): 97-100.

Marchalant Y, Brothers HM, Wenk GL (2009). Cannabinoid agonist WIN-55,212-2

partially restores neurogenesis in the aged rat brain. Molecular Psychiatry,

14(12): 1068-1071 and cover image.

Marchalant Y, Brothers HM, Norman GJ, Karelina K, DeVries AC, Wenk GL (2009).

Cannabinoids attenuate the effects of aging upon neuroinflammation and

neurogenesis. Neurobiology of disease, 34(2): 300-7.

Marchalant Y, Brothers HM, Wenk GL (2008). Inflammation and aging: can

endocannabinoids help? Biomedicine and Pharmacotherapy, 64(4): 212-7.

Knox D, Brothers HM, Norman G, Berntson GG (2008). Nucleus basalis

magnocellularis and substantia inominata corticopetal cholinergic lesions

viii

attenuate freezing induced by predator odor. Behavioral Neurosci, 122(3): 601-

10.

Marchalant Y, Cerbai F, Brothers HM, Wenk GL (2008). Cannabinoid receptor

stimulation is anti-inflammatory and improves memory in old rats. Neurobiology

of Aging, 29(12): 1894-901.

Marchalant Y, Brothers HM, Wenk GL (2008). Neuroinflammation in young and aged

rats: influence of endocannabinoids and caffeine. Abstract. Journal of

Neuroimmunology, 197(2): 168.

Fields of Study

Major Field: Psychology

Table of Contents

Abstract ...... ii Dedication ...... v Acknowledgments...... vi Vita ...... vii Table of Contents ...... x List of Tables ...... xvii List of Figures ...... xviii Chapter 1: General background: Neuroinflammation, glutamate dysregulation and functional impairment ...... 1

1.1 Introduction ...... 1 1.2 Neuroinflammation ...... 1 1.2.1 Microglia activation states ...... 2 1.2.1a Resting...... 3 1.2.1b Classical activation (M1) ...... 4 1.2.1c Alternative activation (M2a) and acquired deactivation (M2c) ...... 6 1.2.1d Microglia activation: a spectrum of its own ...... 7 1.2.2 Microglia and aging; priming or senescence?...... 11 1.2.3 Chronic neuroinflammation ...... 13

1.3 Glutamate regulation ...... 15 1.3.1 Synaptic regulation: Pre-synaptic glutamate release, post-synaptic excitation and LTP ...... 16 1.3.2 Astrocytes: extra-synaptic glutamate clearance and release ...... 17 1.3.2a Glutamate clearance ...... 17 1.3.2b Glutamate-cysteine exchange (xCT) ...... 19 1.3.2c Gliotransmission: Astrocytic vesicular glutamate release ...... 19

1.4 Neuroinflammation and glutamate dysregulation interact, leading to loss of function and cell death ...... 20 1.4.1 Mechanisms of impairment and cell death ...... 20 1.4.1a Elevated glutamate ...... 24 1.4.1b Oxidative stress ...... 26 1.4.1c TNFα activation of programmed cell death ...... 27 1.4.2 Neuroinflammation and glutamate dysregulation are mutually promoting ..... 31

1.5 Neuroinflammation and glutamate dysregulation in Alzheimer’s and Parkinson’s diseases ...... 34 1.5.1 Features of AD and PD ...... 35 1.5.1a Alzheimer’s disease ...... 35 1.5.1b Parkinson’s disease ...... 37 1.5.1c Locus coeruleus and raphe nuclei in AD and PD ...... 40 1.5.2 Neuroinflammation in AD and PD ...... 41 1.5.2a Factors that initiate neuroinflammation in AD and PD ...... 41 1.5.2b Neuroinflammation observed in AD and PD ...... 43 1.5.2c Epidemiology of neuroinflammation in AD and PD ...... 46 Epidemiological evidence of of protection against AD by anti-inflammatory use ...... 47 Epidemiological evidence of protection against PD by anti-inflammatory use 48 1.5.2d Neuroinflammation as a pharmacological target ...... 49 1.5.3 Glutamate dysregulation in AD ...... 52 1.5.3a Evidence of dysregulated glutamate in AD ...... 52 1.5.3b The glutamatergic system as a pharmacological target ...... 57

1.6 Chronic neuroinflammation model ...... 59 1.6.1 Mechanism by which lipopolysaccharide induces inflammation ...... 60 1.6.2 Chronic i.c.v. LPS reproduces components of AD and PD ...... 63

Chapter 2: Caffeine modulates microglia activation in the hippocampus of LPS-infused young rats ...... 67

2.1 Brief Rationale: ...... 67 xi

2.2 Introduction ...... 67 2.2.1 Caffeine is protective against AD and PD ...... 67 2.2.2 Caffeine mechanisms of action and protection ...... 70 2.2.2a Caffeine reduces levels of extracellular glutamate ...... 71 2.2.2b Caffeine modulates immune activation ...... 73 2.2.3 Hypothesis and approach ...... 74

2.3 Methods ...... 74 2.3.1 Subjects, surgical procedures and drug delivery ...... 74 2.3.2 Behavior ...... 76 2.3.3 Histological procedures ...... 76 2.3.3a Single peroxidase staining...... 76 2.3.3b Double-immunofluorescent staining ...... 77 2.3.4 Biochemical procedures ...... 78 2.3.4a Western Blots ...... 78 2.3.4b Binding ...... 78 2.3.5 Statistics ...... 79

2.4 Results ...... 79 2.4.1 Behavior ...... 79 2.4.1a Spatial learning and memory: Morris water maze ...... 79 2.4.1b Open field...... 82 2.4.2 Caffeine reduces LPS-induced microglia activation in the DG ...... 84

2.4.3 LPS infusion increased A1R binding ...... 91 2.4.4 LPS elevates p-Erk and caffeine elevates p-38 ...... 93

2.5 Discussion ...... 94 2.5.1 Caffeine reduces LPS-induced microglia activation ...... 94 2.5.2 A1 receptor expression and binding ...... 96 2.5.3 LPS-induces increased protein levels of p-Erk and caffeine increases p-p38 . 97 2.5.4 Conclusions ...... 98

Chapter 3: Chronic neuroinflammation drives time-dependent changes that recover in the midbrain and brainstem...... 99 xii

3.1 Brief Rationale: ...... 99 3.2 Introduction ...... 99 3.3 Methods ...... 101 3.3.1 Subjects, surgical procedures and LPS administration ...... 101 3.3.2 Behavioral testing ...... 102 3.3.2a Open field ...... 103 3.3.2b Hanging Task ...... 103 3.3.3 Histological procedures ...... 104 3.3.4 Biochemical procedures ...... 107 3.3.4a Western blot analysis ...... 108 3.3.4b DAT Binding ...... 109 3.3.4c ELISA ...... 110 3.3.4d Bradford protein analysis ...... 110 3.3.5 Statistics ...... 110

3.4 Results ...... 110 3.4.1 Behavior ...... 110 3.4.1a Open field analysis of gross motor and anxiety-related behaviors ...... 110 3.4.1b Hanging task evaluation of balance and forelimb strength ...... 111 3.4.2 SN and projections to the striatum ...... 113 3.4.2a Microglia activation and transient TH loss in the substantia nigra ...... 113 3.4.2b Integrity of SNpc projections to the striatum ...... 117 3.4.3 Brainstem: LC and raphe nuclei...... 119 3.4.4 Cerebellum microglia activation ...... 124 3.4.5 Data Summary ...... 125

3.5 Discussion ...... 126 3.5.1 Time-course of microglial activation ...... 126 3.5.2 Compensatory recovery of aminergic systems during chronic neuroinflammation ...... 128 3.5.3 Motor impairment due to neuroinflammation ...... 130 3.5.4 Conclusions ...... 131

xiii

Chapter 4: Delayed memantine does not reverse LPS-induced microglia activation or memory impairment; A pilot investigation...... 132

4.1 Brief rationale: ...... 132 4.2 Introduction ...... 132 4.3 Methods ...... 134 4.3.1 Subjects, surgical procedures and drug administration ...... 134 4.3.2 Spatial learning and memory: Morris water maze ...... 135 4.3.3 Immunohistochemistry ...... 136 4.3.4 Statistics ...... 137

4.4 Results ...... 137 4.4.1 Morris water maze ...... 138 4.4.2 Microglia activation ...... 141

4.5 Discussion ...... 142

Chapter 5: Glutamate transport may promote time-dependent recovery from inflammation-induced hippocampal-dependent spatial memory deficit ...... 145

5.1 Brief rationale ...... 145 5.2 Introduction ...... 145 5.3 Methods ...... 147 5.3.1 Subjects, surgical procedures and LPS administration ...... 147 5.3.2 Spatial learning and memory: Morris water maze ...... 148 5.3.3 Histological procedures ...... 148 5.3.4 Western blot analysis ...... 149 5.3.5 Nitric oxide (NO) release from BV-2 microglia ...... 149 5.3.6 Statistics ...... 150

5.4 Results ...... 150 5.4.1 Behavior: Spatial learning memory in the Morris water maze ...... 150 5.4.2 Hippocampal histology: MHCII, SNAP25 and GLT1...... 153 5.4.3 Molecular and chemical analyses by Western blot ...... 160 5.4.4 Nitric oxide release from LPS-stimulated BV-2 microglia cell culture ...... 160

xiv

5.5 Discussion ...... 160 5.5.1 Microglia ‘activation’ and hippocampal vulnerability...... 161 5.5.2 Increased glutamate uptake may reestablish cognition ...... 165 5.5.3 Conclusions ...... 168

Chapter 6: Manipulation of glutamate transporters by Ceftriaxone and Riluzole in in aging and a model of neuroinflammation ...... 170

6.1 Brief rationale ...... 170 6.2 Introduction ...... 170 6.3 Methods ...... 172 6.3.1 Experimental groups ...... 172 6.3.2 Chronic neuroinflammation via continuous i.c.v. LPS ...... 173 6.3.3 Drug treatment to increase glutamate clearance: Ceftriaxone and Riluzole .. 173 6.3.4 Spatial learning and memory: Morris water maze ...... 174 6.3.5 Biochemical analysis ...... 174 6.3.5a rtPCR mRNA analysis ...... 175 6.3.5b Protein analysis: Bradford...... 176 6.3.5c Protein analysis: BioPlex ...... 176 6.3.5d Protein analysis: Flow Cytometry ...... 177 6.3.6 Statistics ...... 178

6.4 Results ...... 179 6.4.1 Spatial memory ...... 179 6.4.2 Microglia activation phenotype and relationship with spatial memory ...... 185 6.4.2a Inflammation-related mRNA expression ...... 185 6.4.2b Cytokine protein expression ...... 188 6.4.2c LPS-induced TLR4, MHCII and GLT1 but not CX3CR1 on microglia .. 193 6.4.3 GLT1 and xCT expression and relationship with spatial memory ...... 195

6.5 Discussion ...... 198 6.5.1 Neuroinflammation and age are associated with similar cognitive impairments but qualitatively different inflammatory profiles ...... 198 6.5.2 IL-1α and IL-1ß correlate with LPS-induced cognitive impairment ...... 200 xv

6.5.3 Elevated extracellular glutamate may drive cognitive impairment in aged rats ...... 203 6.5.4 Glutamate regulation as a target for inflammation- and age-associated cognitive impairment ...... 205

Chapter 7: General Conclusions ...... 208 Appendix A: List of Abbreviations...... 249 Appendix B: Microglia activation states and immune factors ...... 254 Appendix C: Epidemiological evidence and clinical trials of NSAIDs and other anti- inflammatories in AD and PD...... 265 Appendix D: Summary of outcomes from i.c.v. LPS infusion ...... 269 Appendix E: Experiments using Ceftriaxone with evidence of increased glutamate transport and/or cognitive effects...... 275 Appendix F: Experiments using Riluzole with evidence of increased glutamate transport and/or cognitive effects...... 277

xvi

List of Tables

Table 1. Hyper- and Hypoglutamatergic states across aging and in AD ...... 56

Table 2. Caffeine treated experimental groups ...... 76

Table 3. Animal groups...... 102

Table 4. Schedule of infusion and behavioral testing ...... 102

Table 5. Results summary ...... 126

Table 6. Microglia activation state markers examined ...... 175

Table 7. PCR Primers ...... 176

Table 8. Correlation between TLR4 mRNA expression and spatial memory ...... 187

Table 9. IL-1α/ß mRNA and protein expression correlate with poor spatial memory .. 190

Table 10. Correlation between GLT1 and xCT mRNA and spatial memory ...... 196

xvii

List of Figures

Figure 1. Thesis overview ...... iv

Figure 2. Microglia morphology changes with activation ...... 3

Figure 3. Spectrum of microglia activation states ...... 9

Figure 4. Glutamate regulation between neurons and astrocytes ...... 16

Figure 5. Interaction between neuroinflammation and disease pathology ...... 22

Figure 6. Neuroinflammation and glutamate dysregulation: Mechanisms of cell death . 29

Figure 7. Hippocampal connectivity ...... 36

Figure 8. Effect of SN loss in PD on the basal ganglia motor pathway ...... 38

Figure 9. Protein aggregates in AD and PD related dementias ...... 40

Figure 10. Plaques and Tangles ...... 44

Figure 11. Toll-like receptor 4 signaling pathway ...... 62

Figure 12. Potential targets for caffeine in the basal ganglia in Parkinson's ...... 72

Figure 13. Water maze latency, swim speed and thigmotaxis ...... 81

Figure 14. Water maze probe trial ...... 82

Figure 15. LPS reduces rearing ...... 83

Figure 16. Distribution of hippocampal MHCII+ microglia in rats treated with caffeine water ...... 85

Figure 17. Quantification of hippocampal MHCII+ microglia in rats treated with caffeine water ...... 86

xviii

Figure 18. Distribution of hippocampal MHCII+ microglia of Aged and LPS-infused rats treated with caffeine (40 mg/kg/day i.p.) ...... 88

Figure 19. Quantification of hippocampal MHCII+ microglia in Aged and LPS-infused rats treated with caffeine (40 mg/kg/day i.p.) ...... 89

Figure 20. Distribution of MHCII+ microglia in the hippocampus of LPS-infused rats treated with caffeine (0.5-40 mg/kg/day i.p.) ...... 90

Figure 21. Quantification of MHCII+ microglia in the hippocampus of LPS-infused rats treated with caffeine (0.5-40 mg/kg/day i.p.) ...... 90

+ Figure 22. Adenosine receptor A1 and MHCII microglia in the CA3 ...... 92

Figure 23. Adenosine receptor A1 protein expression and functional binding ...... 93

Figure 24. Protein levels of the phosphorylated MAPKs Erk and p38 ...... 94

Figure 25. Regions of interest ...... 105

Figure 26. Regions of interest evaluated by histological analysis ...... 107

Figure 27. Open field rearing and hanging task results ...... 112

Figure 28. Open field zone preference results ...... 113

Figure 29. Distribution of TH+ neurons and MHCII+ microglia in the SN ...... 115

Figure 30. Quantification of TH+ neurons and MHCII+ microglia in the SNpc ...... 116

Figure 31. pTH distribution in SN after 2 weeks infusion ...... 117

Figure 32. Striatum TH, SNAP25 and DAT binding do not change ...... 118

Figure 33. Distribution of MHCII+ microglia and neurons in the LC and raphe nuclei 121

Figure 34. LC TH and brainstem MHCII expression ...... 122

Figure 35. Hippocampal DBH ...... 123

Figure 36. Cerebellar and brainstem MHCII+ microglia and TH/TrypH+ cells ...... 125

xix

Figure 37. Memantine prevents LPS-induced spatial memory deficit and microglia activation (Rosi et al. 2006)...... 137

Figure 38. Morris water maze ...... 140

Figure 39. Microglia activation...... 142

Figure 40. Morris water maze latency, swim speed and thigmotaxis ...... 152

Figure 41. Morris water maze probe trial results ...... 153

Figure 42. Regional distribution of activated microglia after LPS infusion ...... 154

Figure 43. Distribution of MHCII+ microglia in the hippocampus ...... 155

Figure 44. Quantification of hippocampal MHCII+ microglia ...... 157

Figure 45. Distribution and quantification of hippocampal SNAP25...... 158

Figure 46. Distribution and quantification of hippocampal GLT1 ...... 159

Figure 47. LPS maintains potency after 8 weeks incubation in osmotic minipump ...... 160

Figure 48. Latency to find the hidden platform ...... 180

Figure 49. Distance ...... 181

Figure 50. Velocity ...... 182

Figure 51. Thigmotaxis ...... 184

Figure 52. Probe ...... 185

Figure 53. Hippocampal inflammatory marker mRNA expression ...... 186

Figure 54. Increased TLR4 mRNA correlates with impaired spatial memory ...... 187

Figure 55. Hippocampal cytokine protein expression ...... 189

Figure 56. IL-1ß mRNA and protein expression increase similarly after LPS ...... 190

Figure 57. Increased IL-1ß correlates with impaired spatial memory in LPS ...... 191

Figure 58. Serum cytokine protein expression ...... 192

Figure 59. LPS induced MHCII and TLR4 and GLT1 but not CX3CR1 on microglia . 194

Figure 60. Hippocampal GLT1 and xCT mRNA expression ...... 195

Figure 61. Reduced GLT1 and increased xCT correlate with impaired spatial memory

...... 197

Figure 62. Proposed relationship between GLT1, xCT and cognitive impairment ...... 205

Figure 63. Summary of drugs tested and their generalized effects on glutamate ...... 212

xxi

Chapter 1: General background: Neuroinflammation, glutamate dysregulation and functional impairment

1.1 Introduction Neuroinflammation is an immune response in the central nervous system (CNS) that involves the activation of microglia, resident immune cells of the brain. The principle role of the immune system is designed to protect, but an aberrant or prolonged immune response can potentially disrupt tissue and function. Glutamate dysregulation induced by neuroinflammation may be one way in which neuroinflammation negatively affects the brain. Neuroinflammation and changes in glutamatergic regulation are characteristic of the aging brain and many neurodegenerative diseases, including Alzheimer’s disease (AD) and Parkinson’s disease (PD). The experiments described in this thesis aim to better understand the interaction between neuroinflammation, glutamate dysregulation, and behavioral impairment as well as investigate pharmacotherapies that attenuate the functional consequences of neuroinflammation. In order to provide the background upon which these experiments were designed, this chapter will describe the neuroinflammatory process, the negative consequences of excessive glutamate upon cognitive function, the ways in which neuroinflammation and glutamate dysregulation interact, as well as evidence that neuroinflammation and glutamate dysregulation contribute to pathology in neurodegenerative disease. Finally, I introduce the model used to experimentally produce neuroinflammation in the following chapters.

1.2 Neuroinflammation The CNS was once considered ‘immune privileged’ because of protection by the blood-brain barrier (BBB) and due to the lack of humoral immune response involving leukocytes and antibodies, like that seen in the periphery. Key features of inflammation such as tumor, rubor, dolor and calor are not observed in the brain. The concept of brain

1 immune privilege was revised in the 1990’s, and it is now generally acknowledged that some brain areas (the meninges, ventricles and circumventricular organs) are not immune privileged) and that brain parenchyma is capable of an innate immune response, albeit characteristically different from the peripheral innate immune response (Matyszak 1998). The CNS has an active and dynamic immune response in which resident macrophage-like cells called microglia scan, mobilize, phagocytose, present antigen and release a diverse array of factors. The following section will provide background on the neuroinflammatory response, focusing on the role of microglia.

1.2.1 Microglia activation states Microglia are the primary immune cells of the CNS, analogous to macrophages of the periphery. Microglia comprise 5-20% of all the glial cells in the nervous system and are as abundant as neurons (Kettenmann & Ransom 2005). William Ford Robinson likely stained microglia first in 1900, calling the cells mesoglia and noting that they “seem to have a phagocyte action in certain pathological conditions” (Kettenmann & Ransom 2005). Two decades later in 1921, Pio del Rio-Hortega stained these cells, coined the name ‘microglia’, described the transition from resting state to contracted amoeboid morphology and also proposed phagocytic activity (Kettenmann & Ransom 2005) (Figure 2). Almost 100 years later, the macrophage-like properties of microglia are well acknowledged, and the shared lineage between microglia and macrophages has been detailed. In a recent study, Ginhoux and colleagues (2010) used a gene mapping technique to label macrophages in embryonic mice and revealed that nearly the entire population of adult microglia arose from a single population of yolk sac macrophages; confirming the shared origins of microglia and brain macrophages. Reflective of their myeloid lineage, microglia release cytokines that communicate with other microglia and amplify or dampen the immune response, express dendritic cell-like behavior (antigen presentation) and also macrophage- like behavior (phagocytosis).

Figure 2. Microglia morphology changes with activation Microglia microphotographed by Pio del Rio-Hortega as they progress in morphology from resting to ramified to amoeboid (A-F) (from Kettenmann & Ransom 2005). Microglia activity can be categorized into approximately four generally recognized successive activation states that are often compared to macrophage activation states (i.e. M1, M2a and M2b): resting, classical activation (defense/attack, M1), alternative activation (restoration/repair, M2a) and acquired deactivation (M2c) (Figure 3 and Appendix B, reviewed by: Town et al. 2005; Colton 2009; Lucin & Wyss-Coray 2009; Lynch 2009; Ransohoff & Perry 2009; Cameron & Landreth 2010; Colton & Wilcock 2010; Varnum & Ikezu 2012). Microglia in these states differ in morphology and activity (i.e. motility, phagocytic activity, protein expression, factors released). Activation may be considered a continuum in which classically activated microglia exercise an innate immune response (i.e. phagocytosis) and alternatively activated microglia express properties of the adaptive immune response (i.e. antigen presentation) (Town et al. 2005). There is considerable overlap in the properties and activities of microglia between states (Figure 3), although the distinction between active and resting is the most apparent.

1.2.1a Resting Resting microglia are never truly at rest as once believed; this concept was unequivocally replaced when Nimmerjahn et al. (2005) demonstrated using in vivo two- photon imaging that microglia have long thin processes which continually protrude and withdrawal (within minutes) and scan the whole of the parenchyma (within hours). Microglia support neuronal function and play an active role in synaptic plasticity (Graeber & Streit 2010; Graeber 2010). Microglia are maintained in a quiescent, non-inflammatory state by signals including CD200 and fractalkine (CX3CL) which are constitutively

3 expressed on the surface of neurons (Lyons et al. 2007; F. F. Cox et al. 2011; Koning et al. 2009; Hoarau et al. 2010). While the rest of this section will describe the phenotype of microglia in various states of activation (Figure 3), it should be acknowledged that there is considerable variation in phenotype within an activation state, including the resting state, and variation across brain regions (Lawson 1990; Mittelbronn et al. 2001; Phillips et al. 1999; Sheffield & Berman 1998). For example, De Haas et al. (2008) evaluated 11 common microglia surface proteins across CNS regions of healthy adult mice. They found that hippocampal microglia expressed the most F4/80 and the least CD45, CD80, CXCR3 and CCR9 compared to the cortex, striatum, cerebellum and spinal cord. Human gray matter microglia express less MHCII and CD45 than white matter microglia (Melief et al. 2012). This type of variation in microglia phenotype may underlie regional vulnerability in disease and suggests regional differences in response to an activating stimulus.

1.2.1b Classical activation (M1) Classical activation (M1) is the expression of a pro-inflammatory innate immune response to injury or invasion that is structured to protect the host. Microglia recognize injury and invasion through surface pattern recognition receptors (PRRs). PRRs are alerted to injury by components of dead or dying cells known as damage-associated molecular pattern molecules (DAMPs). Common DAMPs include DNA, purine metabolites (i.e. ATP and adenosine), heat-shock proteins (HSP), high-mobility box group 1 (HMGB1) and S100 cell proteins. Microglia also respond to a decrease in neurotransmitter release and changes in extracellular ion concentration or pH. Microglial PRRs react to host invasion by recognition of pathogen-associated molecular patterns (PAMPs), motifs that are highly conserved across invading organisms. For example, microglia toll-like receptors (TLRs), a subcategory of PRRs, recognize the gram-negative bacterial cell wall component lipopolysaccharide (LPS). Similarly, scavenger receptors (SRs) like the macrophage receptor (MARCO), macrophage scavenger receptor 1 (MSR1) and CD36 are PRRs that populate the microglial surface and recognize low-density lipoproteins from pathogens. In the same study in which Nimmerjahn and colleagues (2005) demonstrate active surveillance of the microenvironment by ‘resting’ microglia, they provide a time-lapse video of microglia activation in response to infusion of LPS in which microglia processes quickly wrap around the tip of the injection pipette. Importantly, components of AD and 4

PD pathology, such as protein aggregates as well as signals from degenerating and dead neurons, activate microglia PRRs and perpetuate an innate immune response (Barger & Basile 2001; Mandrekar-Colucci & Landreth 2010; Sondag et al. 2009; Gao et al. 2011; Lee et al. 2010). After microglia stimulation, transcription factors are activated, including signal transducer and activators of transcription 1 and 4 (STAT1, STAT4) and nuclear factor-κB (NFκB) (Colton & Wilcock 2010b); indicating the activation of genes for a pro- inflammatory response. Pro-inflammatory cytokines are communicative factors produced and released by M1 microglia to propagate the immune response between cells. In particular, M1 microglia express the pro-inflammatory cytokines TNFα, IL-1ß, IL-6, IL-12 and IFNγ. Some of these pro-inflammatory cytokines, like TNFα and IL-1ß, are also sufficient to independently initiate an M1 state (Chen et al. 2008), highlighting the self-propagating nature of the innate immune response. M1 microglia also release a number of other pro- inflammatory cytokines in the interleukin family as well as chemokines (cytokines that encourage microglia migration). M1 microglia contract their processes into a ramified morphology or retract their processes even further to a rounded, amoeboid form (Figure 2) that allows microglia to mobilize and follow gradients of pro-inflammatory cytokines to areas of pathogen invasion or injury. In addition to cytokines, M1 microglia upregulate oxidative stress elements in a respiratory burst which produces reactive oxygen and nitrogen species (ROS/RNS) like nitric oxide (NO), oxygen ions (i.e. superoxide), peroxides and oxygen-containing free radicals. ROS/RNS are released as a cytotoxic ‘respiratory burst’ against pathogens and infected cells, and are also highly concentrated within phagosomes where they assist with for phagocytosis. ROS/RNS kill infected cells and invading pathogens through damage to proteins, lipids and DNA and also interfere with homeostatic redox signaling. Oxidative stress is discussed later in this chapter as a mechanism of inflammation-driven neurodegeneration. In addition to the release of cytotoxic ROS/RNS, M1 microglia express proteins that suggest the M1 state is associated with glutamate dysregulation and potentially with excitotoxicity. These include the glutamate transporter 1 (GLT1; excitatory amino acid transporter 2, EAAT2; solute carrier family 1 member 2, SLC1A2) and the

5 cysteine/glutamate transporter system xCT (SLC7A11). The relationship between neuroinflammation and glutamate dysregulation is the core content of this thesis, and will be further elaborated later in this chapter. An M1 activation state can be deduced better from the production of pro- inflammatory cytokines, ROS/RNS than the expression of surface proteins, although some surface proteins are indicative of an activated phenotype. Activated microglia increase expression of surface proteins, such as TLRs, to recognize pathogens, and major histocompatibility complex II (MHCII) to present antigen to T lymphocytes and initiate a humoral immune response (Tooyama et al. 1990). Though MHCII and its co-factors B7- 1 and B7-2 are upregulated in M1 microglia, they are also highly expressed by M2a/alternatively activated microglia (Wei & Jonakait 1999); suggesting that MHCII may remain on the microglial surface through the transition between activation states. Morphology and surface markers are useful to predict the activation state of microglia, however, it cannot be determined from these indicators exactly how microglia are interacting with their microenvironment and what products they are releasing.

1.2.1c Alternative activation (M2a) and acquired deactivation (M2c) The aggressive innate immune response to injury/invasion is, under non-chronic and non-disease conditions, punctuated by a transition to an anti-inflammatory, alternative activation state (M2a) focused on resolution of the inflammatory response, phagocytosis of pathogens, removal of cellular debris, and tissue repair. This is followed by acquired deactivation (M2c), characterized by immunosuppression and clearance of apoptotic cells. Finally, microglia may transition back to a ‘resting’ state, although they are likely to maintain cellular memory and may be primed for subsequent exposure or remain immuno- suppressed by other M2c microglia in the immediate environment. M2a is characterized by anti-inflammatory cytokine release, and an M2a response can be directly driven by application of either IL-4 or IL-13 (Wei & Jonakait 1999; Szczepanik 2001; O’Keefe et al. 1999); much like an M1 state both produces and can be initiated by the pro-inflammatory cytokines IL-1ß, IL-6, TNFα and IFNγ. M2a microglia upregulate transcription factor STAT6 which enacts the anti-apoptotic effects of IL-4 (Kelly-Welch et al. 2003). Although M2a is an anti-inflammatory state, microglia remain phagocytic and in some cases are more phagocytic than M1 microglia (Durafourt et al. 6

2012). There is an increase in the both c-type lectin receptors (mannose receptor, DC- SIGN, Fizz) and the chitinase-like lectins (Chi3L1, Chi3L2, YM1) that recognize lipids and carbohydrates of pathogens and neuronal debris and facilitate phagocytosis (Mosser & Edwards 2008; Colton & Wilcock 2010). As opposed to the toxic release of TNFα and ROS/RNS by M1 microglia, phagocytosis is an internal and controlled process that serves the microenvironment by removal of pathogens and cellular debris. Deactivation of microglia to M2c is both characterized by and can be driven by elevations in the anti-inflammatory cytokines IL-10 and TGFß. The deactivated M2c state is further characterized by elevated glucocorticoids and nerve growth factor (NGF). M2c microglia are characterized by cystolic SMAD that carries signals from TGFß to the nucleus as well as suppressors of cytokine signaling (SOCS) 1 and 3 and translocation of STAT3. In addition to the reduced production of pro-inflammatory mediators and the increase production of anti-inflammatory mediators, oxidative stress is diminished. Inducible nitric oxide synthase (iNOS) is absent and there is an increase of arginase 1 (Arg1) to compete with iNOS for the substrate arginine; therefore NO is not produced (Rojo et al. 2010). Microglia transition from M2c into a deactivated state that is similar to a resting state, but may leave microglia ‘primed’ for future events.

1.2.1d Microglia activation: a spectrum of its own Microglia activation states are often compared to macrophage activation states, which are based upon the Th1 and Th2 immune responses of T cells (Gordon 2003; Mosser & Edwards 2008). There are considerable similarities between microglia and macrophage activation states, but there are differences between these cell types in expression and activity as well (reviewed by Ransohoff & Perry 2009). Durafourt and colleagues (2012) tested the activity of human microglia and macrophages biased toward an M1 phenotype with LPS and IFNγ or towards an M2a phenotype with IL-4 and IL-13 in vitro. The authors uncovered overexpression (≥4-fold) exclusive to either microglia or macrophages of 41 genes and/or proteins within a shared activation state. Similarly, Melief and colleagues (2012) biased human microglia and macrophages toward an M1 state with LPS and IFNγ, toward an M2a phenotype with IL-4, or toward an M2c activation state with the glucocorticoid dexamethasone. Microglia did not increase TNFα release in response to TNFα and IFNγ as robustly as macrophages did (100-200-fold), unless primary microglia 7 cultures were allowed time to upregulate CD14 (a co-receptor to TLR4); demonstrating that the M1 response in microglia is blunted compared to macrophages unless they are ‘primed’. This study also found that MR was upregulated in IL-4-exposed macrophages, but not microglia, and that M2c-biased microglia overexpressed CD163, MR and CCL18 but M2c macrophages only upregulated expression of CD163. These data caution against using criterion that define macrophage activation states to categorize microglia activation states. Furthermore, these data illustrate that multiple factors on multiple levels (i.e. morphology, gene expression, surface markers, proteins released, and phagocytic behavior) must be used in order to approximate (not define) microglia activity. Understanding microglia activation states may be useful as a predictor of function and guide to therapeutic targets. Yet, the data available suggest that these states are not distinct conditions, nor are they points on a continuum from resting to activated (Town et al. 2005), but are better thought of as a particular points in a spectrum. A similar idea has been suggested by Mosser and Edwards (2008) for macrophages wherein macrophages express characteristics within a non-linear spectrum between “regulatory”, “classically activated” and “wound-healing”. This approach better accommodates the large array of microglia states that may co-exist within a diseased brain as well as the diversity of microglia activation states that may be produced from interactions between disease, age, drug exposure and a plethora of environmental and biological factors. The following chapters describe experiments in which MHCII was used to define microglia activation. We believed this was adequate, because we knew that LPS directly activated the innate immune response and a classical M1 activation state in microglia. However, an experiment in which microglia maintained MHCII expression when LPS- induced behavioral impairment improved (Chapter 4) led to doubt that MHCII was, at that time point, labeling microglia in a classical activation state. In the project proposed for my candidacy exam (Chapter 6) I examined various markers of microglia activation in order to address these concerns and better predict the activation state of microglia in our samples. This includes markers of an M1 state (IL-1ß, IL-6, IL-12, IL-1a, IFNγ, TNFα, GM-CSF, TLR4, xCT, GLT1), M2 states (IL-4, IL-5, IL-10, IL-13, PPARγ) and resting (CX3CR, CX3CL).

Figure 3. Spectrum of microglia activation states Factors that induce or are elevated in a particular activation state are shown and relationships between them are visualized. Colors are used to represent the continuous and overlapping nature of activation across a spectrum of activation: resting microglia (green), M1 microglia (red), M2a (blue) and M2c (teal). Morphology was traced from images of activated microglia (Streit & Graeber 1996). Phagocytic activity is depicted for M1 and M2a phenotypes. Elements in this schematic are generalized and not exhaustive. This schematic is designed to be used with the accompanying table in Appendix B: Microglia activation states and immune factors, in which more detail is given. Abbreviations: pathogen activated molecular pattern (PAMP), danger activated molecular pattern (DAMP), lipopolysaccharide (LPS), amyloid ß (Aß), heat-shock protein (HSP), high mobility box group 1 (HMGB1), pattern recognition receptor (PRR), toll-like receptor (TLR), scavenger receptor (SR), macrophage receptor with collagenous structure (MARCO), macrophage scavenger receptor 1 (MSR1), cluster of differentiation (CD), tumor necrosis factor α (TNFα), interleukin- (IL-), interferon γ (IFNγ), TNFα receptor (TNFαR), TNFαR death domain (TRADD), glutamate transporter 1/excitatory amino acid transporter 2 (GLT1), glutamate-cystine antiporter (xCT), reactive oxygen - and nitrogen species (ROS/RNS), nitric oxide (NO), superoxide O2 , antigen presenting cell (APC), major histocompatibility complex II (MHCII), mannose receptor (MR), dendritic cell-specific intercellular adhesion molecule-3-grabbing non-integrin (DC- SIGN), found in inflammatory zone (FIZZ), chitinase-3-like protein 1 (Chi3L1), chitinase-3-like protein 2 (Chi3L2), Chi3L3 (YM1), cannabinoid receptor 2 (CB2), transforming growth factor ß (TGFß), nerve growth factor (NGF), fractalkine (CX3CL1), fractalkine receptor (CX3CR), suppressor of cytokine signalling (SOCS), nuclear factor κ-light-chain-enhancer of activated B cells (NFκB), signal transducers and activators of transcription (STAT), peroxisome proliferator-activated receptor γ (PPARγ).

Figure 3.

1.2.2 Microglia and aging; priming or senescence? There are two competing perspectives on the activation state of aged microglia that are crucial approach we take in treating neuroinflammation in disease: 1) that microglia become primed and more reactive with age (Godbout & Johnson 2009; Wynne et al. 2010b; Henry et al. 2009; Wynne et al. 2009; Barrientos et al. 2010; Frank et al. 2010) and 2) microglia become senescent and less effective with age (Miller & Streit 2007; Streit et al. 2004; Streit et al. 2009; Streit 2006; Flanary et al. 2007). These concepts stand in stark contrast. The first suggests that immunotherapy should be used to suppress microglia activation in the diseased brain, and the latter proposes that we aim to augment microglia reactivity and promote their immune function. This is an important issue to address, because the single strongest correlate with Alzheimer’s onset and progression is advancing age, and one of the strongest known correlates with cognitive impairment is neuroinflammation (Cagnin et al. 2001; Cagnin et al. 2006; Edison et al. 2008; Okello et al. 2009; Parachikova et al. 2007). Microglia increase expression of MHCII with age in humans (Overmyer et al. 1999). Our lab has shown that the number of MHCII+ microglia in the hippocampus increase in aged rats, yet the number of MHCII+ microglia is not enhanced by continuous i.c.v. LPS as robustly as it is in young rats (Bardou et al., 2013). Similarly, microglia increase expression of MHCII and MHCII-related proteins in aged mice, but this is not exacerbated by i.p. injection of LPS as it is in young mice (Godbout et al. 2005; Frank et al. 2010). These observations imply that there may be a ceiling effect in which the majority of microglia in aged animals express MHCII. MHCII indicates that the basal activity of aged microglia may be active, but does not distinguish between M1 and M2a microglial phenotypes. However, there is experimental evidence that microglia in aged animals are biased toward an M1 state due to the elevated basal and LPS-induced expression of pro-inflammatory cytokines as well as the reduced presence of and response to anti-inflammatory cytokines. The Maier/Watkins group identified an age-related basal increase in gene expression of microglia activation indicators including MHCII (and MHCII-related B7-2 and CIITA), CD11b, Iba1 as well as IFNγ, and demonstrated that the microglia of aged rata have an exaggerated expression of IL-1ß to LPS ex vivo and IL-1ß protein in the hippocampus after peripherally

11 administered LPS (Frank, Barrientos, et al. 2006; Frank, Barrientos, Watkins, et al. 2010; Barrientos et al. 2009; Barrientos et al. 2010). Increased activation and reactivity of aged microglia are further supported by the lack of mRNA for CD200, which encourages quiescence, and reduced IL-10, which biases microglia toward M2c (Frank, Barrientos, et al. 2006). Consistent with these findings, the Godbout lab demonstrated that microglia from aged mice do not move toward an M2 phenotype in response to IL-4 ex vivo (Fenn et al. 2011). Furthermore, aged microglia express lower levels of CXCL1, a protein that promotes microglia senescence, and when challenged with peripheral LPS in vivo, aged microglia express less CX3CR1, less anti-inflammatory TGF and more pro-inflammatory IL-1ß (Wynne et al. 2010a); demonstrating a propensity to bias toward a M1 response and away from a M2 phenotype. Chen and colleagues (2008) demonstrated with laser microdissection of hippocampal sub-regions that IL-1ß and TNFα mRNA were increased in aged mice and that peripheral LPS challenge enhanced these pro-M1 cytokines more than in young animals. This evidence suggests that aged microglia are more likely to have a prolonged period of classical activation in response to innate immune challenge. Interestingly, hippocampal-dependent memory deficits and sickness behavior were exaggerated in aged mice treated with peripheral or i.c.v. LPS (Huang et al. 2008; Frank et al. 2010; Barrientos et al. 2010; Chen et al. 2008). The relationship between cognitive impairment and chronic neuroinflammation, such as that found in aging or disease, will be discussed in detail throughout the remainder of this thesis. Despite the fairly consistent literature on microglia priming with age, there is a competing hypothesis that microglia become senescent and that it is not microglia activation but lack of microglia protection that contributes to damage and cognitive impairment in the aging and diseased brain (Streit 2006; Streit & Xue 2010). This theory is primarily supported by work from Dr. Wolfgang Streit’s lab. They have demonstrated that telomere length is shortened and telomerase activity reduced in microglia of aged rats and humans, particularly in humans with AD (Flanary et al. 2007). Ramified microglia surround plaques in AD patients (Streit et al. 2009); suggesting by morphology that aged microglia, even in a disease state are not activated. Furthermore, dystrophic microglia are more abundant under conditions of AD pathology, particularly around tau aggregates and they phagocytize Aß less efficiently (Hickman et al. 2008; Flanary et al. 2007; Streit et al.

2009); suggesting that even if aged microglia are ‘primed’, they are less effective at fighting disease pathology. The studies outlined in this thesis investigate the relationship between chronic neuroinflammation, glutamate and cognitive impairment. The debate over whether activated microglia drive pathology or senescent microglia fail to protect against pathology is important to AD. Activated microglia are predictive of and correlate better with disease onset and progression in AD than other disease pathology, like Aß plaques (Cagnin et al. 2006; Edison et al. 2008a). Furthermore, the products of M1 microglia are sufficient to drive cognitive impairment. Therefore, prolonged pro-inflammatory microglia activation state, whether induced by pathogen or age, influences cognitive impairment and neurodegeneration independent of AD pathology, through dysregulation of glutamate before. From this perspective, it would be beneficial to attenuate microglia activation in the aged and diseased brain.

1.2.3 Chronic neuroinflammation Neuroinflammation is often referred to as a ‘double-edged sword’ because of its potential to both protect and harm. In general, immune activity in the brain is beneficial. The immune response protects the brain from pathogens, promotes the healing after injury, conserves resources by quickening the death of cells when damage is irreversible, removes extracellular debris, and promotes cellular repair. However, the mechanisms by which microglia may protect the brain, may cause damage to surrounding tissue when prolonged. Independent of co-existing disease pathology, chronic neuroinflammation may lead to behavioral impairment and neurodegeneration. Classically activated microglia produce and release pro-inflammatory factors in order to attract and induce activation in other microglia, creating a cycle of activation, and also to respond to injury and pathogens. Some of these factors are used in defense, such as ROS/RNS which can induce oxidative stress and kill bacteria. Under normal circumstances, this response is transient and does not cause damage to surrounding tissue. Microglia subsequently enter a phase of resolution and repair, alternative activation, and become deactivated. Harm to neurons can occur, however, if the pro-inflammatory phase is prolonged, and the inflammatory process transitions from being primarily protective to predominantly pathological. 13

Not only is there considerable overlap in microglia phenotype between activation states and variance in phenotype within a state and across brain regions, ages and disorders, as discussed above, but microglia of multiple activation states may also coexist. In fact, the diseased brain is characterized by the presence of microglia across a range of activation states (Colton & Wilcock 2010b). This may occur under circumstances of chronic neuroinflammation and disease, because there is not a distinct period in which a microglia population is activated and later deactivated. In neurodegenerative disease, pathological protein species and signals from dying and dead neurons fuel immune activation over decades. Therefore, neuroinflammation is sustained over time, punctuated by the activation and deactivation of individual microglia. This further suggests that the microenvironment may be influenced by competing interests, both attack and repair. In addition to continual exposure to immune-activating disease pathology, a chronic neuroinflammatory response is also self-sustaining. For example, the pro- inflammatory cytokines IL-1ß and IFNγ released by activated microglia, are all sufficient to induce microglia activation (Lee et al. 1993; Loughlin et al. 1993); therefore, activated microglia recruit nearby microglia to an activated state, and this sustains a population of activated microglia. Furthermore, disease pathology and a pro-inflammatory immune response can interact to sustain a chronic inflammatory state. For example, expression of the PRR receptor for advanced glycation endproducts (RAGE) is increased on microglia from AD brains, indicating that they are 'primed' for activation. When RAGE is exposed to Aß, microglia produce macrophage colony stimulating factor (M-CSF), which recruits microglia and increases their expression of RAGE (Lue et al. 2001). This sets up a cycle by which the immune system may be continually activated by the Aß present in AD. Similarly, the complex cascade of events initiated by chronic microglia activation include increased surface expression of MHCII. MHCII on microglia stimulate T cells which subsequently renew the release of pro-inflammatory factors from microglia (Lynch 2009) that both maintain and recruit activated microglia. Finally, a pro-inflammatory state contributes directly to disease pathology (discussed in the next section), and disease pathology stimulates a pro-inflammatory state. All of these relationships demonstrate the self-perpetuating attributes of neuroinflammation.

1.3 Glutamate regulation Glutamate is the primary excitatory neurotransmitter and is approximately 1000- fold more concentrated in the central nervous system than other neurotransmitter (Butcher & Hamberger 1987). This is, in part, because glutamate is an important metabolite for many cellular products, including the antioxidant glutathione, as well as the neurotransmitter GABA, amino sugars and nucleotides. Glutamate is compartmentalized in pre-synaptic terminals (~100mM), post-synaptic terminals (~10mM) and astrocytes (~2mM), and is highly concentrated in these compartments compared to the extracellular space (<1µM) or CSF (~1µM) (Nedergaard et al. 2002). In a simplified view, glutamate is released from pre-synaptic neurons after an action potential, depolarizes the post- synaptic membrane by interaction with post-synaptic AMPARs and NMDARs, is cleared from the synapse by transporters (GLT1) on astrocytes, is metabolized within astrocytes to glutathione and glutamine, and is recycled in pre-synaptic neurons from glutamine to glutamate and packaged in vesicles for exocytosis. In addition, metabotropic glutamate receptors (mGluRs) populate astrocytes, neurons and microglia and modulate glutamatergic neurotransmission (Niswender & Conn 2010), however, they are not a main focus of this thesis and will not be reviewed here. The action and regulation of glutamate within the tripartite synapse between neurons and astrocytes will be explained in detail and is represented schematically in Figure 4.

Figure 4. Glutamate regulation between neurons and astrocytes Glutamate (Glu) released by presynaptic neurons opens 2-amino-3-(3-hydroxy-5-methyl- isoxazol-4-yl)propanoic acid receptors (AMPARs) and, after sufficient membrane depolarization, N-Methyl-D-aspartic acid receptors (NMDARs). Glutamate is sequestered by glutamate transporter 1 (GLT1) on astrocytes. Glutamine synthetase converts glutamate to glutamine (Gln), and glutamine is transported to neurons for glutamate synthesis, but glutamate is also synthesized in and released by vesicular exocytosis from astrocytes. The cystine/glutamate antiporter (xCT) exchanges glutamate for cystine (Cys), a precursor to the antioxidant glutathione (GSH), primarily on astrocytes, but also on neurons.

1.3.1 Synaptic regulation: Pre-synaptic glutamate release, post-synaptic excitation and LTP Glutamate is synthesized primarily as a produce of the citric acid cycle in astrocytes, however, the concentration of glutamate is relatively low in astrocytes compared to neurons because it is converted to glutamine by glutamine synthetase in astrocytes (Nedergaard et al. 2002). Glutamine is then exported into the extracellular space by a Na+-dependent transporter and acquired by pre-synaptic neurons where is it converted from glutamine to glutamate by glutaminase. Glutamate is packaged by vesicular glutamate transporters into synaptic vesicles that are docked at the synaptic cleft by the soluble N-ethylmaleimide sensitive fusion protein attachment protein receptor (SNARE) complex and released by exocytosis.

Glutamate release from the pre-synaptic terminal stimulates ion channels at the post-synaptic site, kainate and α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors (APMARs) that allow the entry of Na+, and N-methyl-D-aspartic acid receptors NMDARs that allow Ca2+ influx. Entry of these Na+ and Ca2+ ions depolarize the post- synaptic site. If the membrane potential depolarizes past a certain threshold, NMDARs will lose their voltage-dependent Mg2+ blockade, and Glutamate released into the synapse will bind NMDARs and cause the influx of Ca2+. As Na+ and Ca2+ enter the post-synaptic site, they depolarize the post-synaptic terminal, producing an excitatory post-synaptic potential (EPSP). The membrane is restored over time to resting potential (approximately -70 mV) by the Na+/K+ ATPase, which returns Na+ against its concentration gradient to the extracellular space and is very energetically expensive. Determined by spatial summation and temporal summation, i.e. closeness in distance and timing, EPSPs will have a combined effect in which the post-synaptic membrane is depolarized enough to cross the threshold in which voltage-dependent Na+ channels open and an action potential is issued from the neuron. Dependent upon the timing and strength of successive stimuli, the post-synaptic site can become more responsive to the same magnitude of synaptic input, and this process is called long-term potentiation (LTP). Depolarization of the post-synaptic membrane can drive transcription factors that lead to an increase of AMPARs and NMDARs at the synapse, the addition of synapses, increased spine density, and overall arborization of the dendritic tree. This process is the physical realization of the Hebbian theory in which ‘neurons that fire together wire together’. The functional manifestation is long-term enhancement of the post-synaptic response to event-related glutamate release from the pre- synaptic terminal. LTP is conceptualized as the physical basis of memory formation.

1.3.2 Astrocytes: extra-synaptic glutamate clearance and release

1.3.2a Glutamate clearance The termination of glutamate activity in the synapse is primarily accomplished through removal by excitatory amino acid transporters (EAATs). EAATs are high affinity, Na+-dependent transporters that co-transport glutamate with Na+ in exchange for intracellular potassium (Zerangue & Kavanaugh 1996), effectively clearing glutamate from

17 the synapse. However, these transporters can operate in reverse, releasing glutamate into the extracellular space, under conditions of ischemia or metabolic distress (reviewed by Malarkey & Parpura 2008). There are five sub-types of EAAT which are abundant in the brain (Chaudhry et al. 1995; Lehre et al. 1995), located predominantly on astrocytes but also neurons and microglia. EAATs can concentrate glutamate within astrocytes up to 10,000-fold compared to the extracellular space (Milton et al. 1997). The significance of these stores is appreciable, given that astrocytes outnumber neurons (Kettenmann & Ransom 2005). Evoked glutamate release from neurons reaches peak synaptic concentrations of up to 1.1 mM, and glutamate is removed from the synapse within milliseconds in hippocampal slice cultures (Clements et al., 1992). In contrast, glutamate concentration in the extracellular space is maintained at <1µM (Nedergaard et al. 2002). NMDARs are normally not responsive to glutamate concentrations in the µM range due to the presence of the voltage-gated Mg2+ blockade, but under energetic stress, this blockade is removed and NMDARs become sensitive to glutamate concentrations <35µM (Cox et al. 1989). Sustained extracellular glutamate levels of 30 µM (less than 3% of synaptic release release) over a 24 hr. period lead to apoptosis in approximately 40% of cultured neurons (Ankarcrona et al. 1995), illustrating the essential role of glutamate clearance. GLT1 is the principle EAAT found on astrocytes and is responsible for approximately 80-90% of glutamate clearance in hippocampal tissue, followed by the glutamate asparate transporter (GLAST) (Danbolt 2001; Maragakis & Rothstein 2004; Selkirk et al. 2005). Immunohistochemical analysis of GLT1 in human brain tissue demonstrates that GLT1 expression is almost exclusively on astrocytes, particularly perisynaptic astrocytes in regions receiving dense glutamatergic innervation, such as the molecular layer of the DG (Milton et al. 1997; Rothstein et al. 1994; Lehre et al. 1995). Microglia and neurons do not generally express GLT1 (Milton et al. 1997), however, GLT1 is present on the surface of microglia exposed to LPS (Persson et al.2005) and neurons under pathological conditions such as the AD brain (Thai 2002; Sasaki et al. 2009; Pow & Cook 2009). Decreased neuronal GLT1 expression is linked to increased vulnerability to glutamate-induced excitotoxicity (Selkirk et al. 2005), and higher expression of GLT1 is positively correlated with memory performance in rodents (Heo et al. 2011).

1.3.2b Glutamate-cysteine exchange (xCT) - The cystine/glutamate antiporter (xCT; system xc ) is found primarily on the surface of astrocytes, although also found on neurons and microglia, and exchanges extracellular cystine for intracellular glutamate, releasing glutamate into the extracellular space (Moran et al. 2003; reviewed by Lewerenz et al. 2012). Cystine is reduced to cysteine and used in the production of the antioxidant glutathione (GSH). Antioxidant production is an important form of cellular protection, particularly in a pro-inflammatory brain environment, and loss of GSH can lead to cell death (Tan et al. 2001; see Section 1.4.1b). In fact, cultured neurons die without cystine in the medium, high concentrations of glutamate reduce cystine uptake and this reduction is directly proportional to the degree of neuronal toxicity, and neurons are protected from glutamate toxicity in the presence of antioxidants such as GSH (Miyamoto et al. 1989; Murphy et al. 1989). Import of cystine is necessary to protect neurons, yet, xCT transport of glutamate into the extracellular space can amplify pathology, and is necessary for glutamatergic neurotoxicity induced by activated microglia in vitro (Piani & Fontana 1994). Whether xCT is neuroprotective by producing the GSH, or contributes to excitotoxicity by transport of glutamate into the extracellular space may depend upon the accompanying function of EAATs (Lewerenz et al. 2012; Lewerenz et al. 2006; Sheldon & Robinson 2007); therefore, increasing EAAT function and glutamate clearance is a preferable pharmacological target to modulating xCT function.

1.3.2c Gliotransmission: Astrocytic vesicular glutamate release Astrocytes clear glutamate from the synapse, recycle glutamate and also release vesicular glutamate (gliotransmission) in a manner that can augment pre-synaptic release. Communication between neurons and astrocytes is bi-directional. Astrocytes respond to neuronal activity with transient elevations in Ca2+ (Dani et al. 1992; Porter & McCarthy 1996). The machinery to transport glutamate into vesicles (vesicular glutamate transporters 1 and 2) is found in approximately 25-40% of astrocytes as well as the machinery for exocytosis (the SNARE protein cellubrevin), and release of quantal glutamate can be observed within 200 milliseconds of mGluR stimulation (Bezzi et al. 1998; Bezzi et al. 2004).

Astrocytic glutamate release regulates neuronal transmission (Nedergaard 1994; Parpura et al. 1994; Hassinger et al. 1995; Mennerick et al. 1996; Pasti et al. 1997; Volterra & Steinhäuser 2004; Bezzi et al. 2001; Vesce et al. 2007) Ca2+ increases in cultured astrocytes can be pharmacologically induced with bradykinin and glutamate release from astrocytes was sufficient to drive Ca2+ transients in neurons in astrocyte-neuron co-cultures (Parpura et al. 1994; Volterra & Steinhäuser 2004). Interestingly, astrocytes release vesicular glutamate in a TNFα-dependent manner. Using patch-clamp recordings in brain slices of transgenic mice deficient in TNFα, Voltera’s laboratory (Santello et al. 2011) demonstrate that glutamate release from astrocytes is slowed because of a change in vesicle docking and potentiation of pre-synaptic NMDARs is abolished. The implications of this homeostatic role of TNFα in astrocytic glutamate release in the context of neuroinflammation and elevated TNFα is discussed later in this chapter.

1.4 Neuroinflammation and glutamate dysregulation interact, leading to loss of function and cell death

1.4.1 Mechanisms of impairment and cell death Chronic neuroinflammation and glutamate dysregulation can contribute to cognitive impairment and facilitate neurodegeneration. Neuroinflammation may cause cell death through glutamatergic excitotoxicity, oxidative stress and TNFα engagement of the TNFαR death domain (TRADD, see Appendix B). The following schematic (Figure 5) summarizes the intricate interactions between neuroinflammation, glutamatergic dysregulation and processes that may underlie loss of function and neurodegeneration in AD and PD. Factors that disrupt normal function, in addition to accumulated cell, are important targets for disease treatment. For example, approximately 80% loss of SN dopaminergic neurons is the threshold at which motor symptoms become apparent in PD. This suggests that the function of the basal ganglia is protected by compensatory processes during the pre-symptomatic period when there is sub-threshold, yet substantial neurodegeneration. If identified, these compensatory mechanisms may be augmented by pharmacological therapy. Likewise, accumulated neuronal degeneration leads to cognitive impairment in AD. Yet, patients with AD have a degree of variance in cognitive acuity across days, and

20 this variation suggests that there are other factors in addition to cell loss (which is progressive and does not reverse) that contribute to impairment that can be pharmacologically manipulated. One such factor is glutamatergic regulation (Palop et al. 2006; Huang & Mucke 2012; Palop & Mucke 2010). Together, neuroinflammation and excessive glutamate signaling are two factors that can overcome compensatory processes and lead to functional impairment (both on the cellular and behavioral levels) in addition to neuronal degeneration (Wyss-Coray & Mucke 2002; Akiyama et al. 2000).

Figure 5. Interaction between neuroinflammation and disease pathology Schematic represents the intercommunication between microglia, astrocytes and neurons under healthy (blue) and inflammatory (red) conditions. Pathological protein aggregates and degenerating neurons elicit microglia activation. Once activated, microglia present specific surface markers and release pro-inflammatory factors that amplify immune activation and communicate to astrocytes and neurons. Activated microglia become phagocytic against protein aggregates. Under inflammatory conditions, glutamate levels rise, in part due to reduced sequestration by astrocytes and increased release from glia triggered by TNFα and IL-1ß . Elevated glutamate levels lead to mitochondrial dysfunction that is exacerbated by oxidative stress, and may lead to cell death. This figure is not exhaustive, but illustrates most factors discussed in this section. Abbreviations: cluster of differentiation factor 200 receptor (CD200R), fractalkine (CX3CL1), interleukin (IL-), fractalkine receptor (CX3CR1), transforming growth factor ß (TGFß), glial cell line-derived growth factor (GDNF), brain derived neurotrophic factor (BDNF), glial fibrillary acidic protein (GFAP), glutathione (GSH), interferon ƴ (IFNγ), tumor necrosis factor α (TNFα), nitric oxide (NO), reactive oxygen species (ROS), reactive nitrogen species (RNS), major histocompatibility complex II (MHCII), pattern recognition receptors (PRRs), toll-like receptor 4 (TLR4), glutamate transporter 1 (GLT1), receptor for advanced glycation endproducts (RAGE), TNF receptor-associated death domain proteins (TRADD), adenosine triphosphate (ATP), high-mobility box group 1 (HMGB1), amyloid ß (Aß), and neurofibrillary tangles (NFTs).

Figure 5.

1.4.1a Elevated glutamate Glutamate is a tightly regulated excitatory neurotransmitter, but pathological conditions, such as neuroinflammation, lead to loss of homeostatic regulation and an increased availability of extracellular glutamate (Rogawski & Wenk 2003). Mechanisms of extracellular glutamate accumulation include enhanced astrocytic or pre-synaptic vesicular release, release from microglia, reduced clearance or reversal of glutamate transporters, increased xCT activity, swelling-induced astrocytic expulsion, release through ionotropic purinergic receptors and release through hemichannels on astrocytes or microglia (reviewed by Malarkey & Parpura 2008). Neuroinflammation can drive an increase in extracellular glutamate. Elevated extracellular glutamate disrupts the tight regulation of glutamate necessary for memory formation, excessive glutamate can lead to cell death, and accumulated cell loss manifests as gross cognitive impairment. Elevated extracellular glutamate is likely to interact with AMPARs, and enable enough Na+ entry to depolarize the post-synaptic membrane. If the membrane is depolarized, then the Mg+ blockade may be release from NMDARs and the likelihood increases that extracellular glutamate will also activate voltage-gated NMDARs. When post-synaptic NMDARs are activated, Ca2+ enters the post-synaptic membrane. Under normal conditions, extracellular glutamate may regulate synaptic activity, LTP and memory formation (reviewed by Featherstone & Shippy 2008; Volterra & Steinhäuser 2004). However, in a pathological situation, NMDAR activation and Ca2+ entry is no longer tied to event-related release of glutamate after an action potential. In this case, the post-synaptic neuron receives false ‘signals’ that disrupt LTP. In fact, in a model of chronic neuroinflammation produced by continuous i.c.v. LPS, we see loss of LTP and spatial memory performance and increased induction of the Ca2+-activated immediate early gene, Arc, in response to a novel environment (Hauss-Wegrzyniak et al. 1998; Rosi et al. 2005). This indicates that neuroinflammation is associated with increased Ca2+ entry through NMDAR activation by elevated extracellular glutamate, and that this is sufficient to drive disrupt cognitive processes. Elevated extracellular glutamate can promote memory impairment, and may also lead to cell death. Elevated Ca2+ entry through open NMDARs may cause mitochondrial stress, which leads to less energy production for the cell. Energetic failure leads to loss of 24 cellular homeostatic function, including the inability to maintain the resting membrane potential, as this is dependent upon ATP and very energetically expensive. If the membrane remains depolarized, the response to extracellular glutamate is likely to continue, until excessive Ca2+ entry and mitochondrial failure culminate in excitotoxic cell death. In addition to excitotoxicity, glutamate that is excessive enough to ‘spill over’ the synaptic cleft is likely to engage extra-synaptic NMDARs that are rich in the NR2B subunit (unlike synaptic NMDARs that contain more of the NR2A subunit). Extracellular NMDARs comprise an estimated 36% of all NMDARs (Harris & Pettit 2007), and their activation is more likely to lead to cell death (Hardingham & Bading 2010; Xia et al. 2010; Stanika et al. 2009; Liu et al. 2007), and hippocampal neurons develop tolerance to glutamate excitotoxicity by down-regulation of the NR2B subunit (Kambe et al. 2010). A strong connection between NR2B and AD was made by an Alzheimer’s Disease Neuroimaging Initiative (ADNI) study in which 742 subjects were observed over 5 years, and found that a polymorphism in the NR2B gene is the only genetic association of over 500,000 investigated genes that correlated with reduced temporal lobe volume and poor mini-mental state exam (MMSE) score (Stein et al. 2010). Animal models in which glutamatergic activity is enhanced demonstrate impairment in hippocampal-sensitive tasks and neuronal loss. Partial transgenic elimination of GLT1 accelerates cognitive decline in a mouse model of AD (Mookherjee et al. 2011), and GLT1 correlates positively with performance in spatial memory tasks in mice (Heo et al. 2011), suggesting that reduced glutamate clearance promotes elevated extracellular glutamate and memory impairment. Extracellular glutamate release from activated microglia and subsequent neuronal toxicity in vitro are dependent upon cystine, and the presence of xCT (Piani & Fontana 1994; Lewerenz et al. 2006; Lewerenz et al. 2012). Inhibition of glutamate metabolism through glutamate dehydrogenase 1 (GLUD1) in a transgenic mouse model is associated with increased levels of extracellular glutamate, increased frequency and amplitude of EPSPs, enhanced KCl-induced synaptic glutamate release, elevated Arc expression, decreased GLT1 expression, reduced glutamate clearance, phosphorylation of the NMDA-R2B in the hippocampus, and regionally- specific/age-dependent neuronal loss (Camacho et al. 2007; Michaelis et al. 2011; Xiaodong Bao et al. 2009; Xinkun Wang et al. 2010).

1.4.1b Oxidative stress Oxidative stress from factors release by microglia may directly drive neuronal degeneration. Phagocytes, such as microglia, produce ROS/RNS that include H2O2 and peroxynitrite as well as free radicals such as NO, superoxide and hydroxyl radicals. Phagocytes use ROS/RNS to eliminate pathogens, but high levels of ROS/NRS can weaken mitochondria and impair cellular energy production and homeostasis, as well as cause damage to the lipids, proteins and DNA of nearby cells. Oxidative stress eventually leads to apoptosis or necrosis, depending upon the degree of oxidative stress. In addition, the oxidative burst characteristic of oxidative stress consumes available GSH, and the loss of GSH triggers glutamate release from microglia that contributes to excitotoxicity (Barger et al. 2007). In addition, oxidative stress modifies EAATs and decreases clearance activity (Volterra et al. 1994; Trotti et al. 1998). Moreover, elevated extracellular glutamate and oxidative stress interact; both weaken mitochondria, through elevated Ca2+ or disruption of the respiratory cycle, and make mitochondria more vulnerable to failure (Brown & Bal- Price 2003; Brown & Neher 2010). Neuronal toxicity and glutamate release induced by LPS- or cytokine-activated microglia in vitro is attenuated by administration of NO and NOS inhibitors or stabilization of NO with superoxide dismutase (SOD) as well as by removal of cystine or treatment with an NMDAR antagonist (Boje & Arora 1992; Piani & Fontana 1994); demonstrating that microglia activation directly decreases the viability of a neuron, both through glutamate excitotoxicity and oxidative stress. A review of findings in AD patients (Gibson & Shi, 2010) observed that mitochondrial dysfunction and oxidative stress are invariable across the AD disease spectrum; i.e. they are present both in early onset/genetically-driven disease types as well as late onset in which the disease- initiating factors are unknown. The production of ROS/RNS is associated with AD and PD. Levels of iNOS and the cytokines IL-1α and TNFα are elevated in the in the amygdala, hippocampus, EC and insular cortex of AD and Lewy Body dementia (DLB) brains (Katsuse et al., 2003). Polymorphisms of NOS genes are risk factors for PD (Lo et al., 2002, Hancock et al., 2008). The NOS inhibitor, asymmetric dimethylarginine, decreases in the cerebral spinal fluid (CSF) of AD patients and correlates with decreased cognition (Abe et al., 2001).

Abnormal proteins and aggregates found in AD and PD can induce oxidative stress.

Exposure to Aß can induce increased levels of H2O2, iNOS, NOS, NO and other ROS/RNS predominantly from microglia but also from astrocytes and neurons, and these compounds may mediate Aß-induced toxicity in the diseased brain (Behl et al., 1994, Goodwin et al., 1995, Hunot et al., 1996, Cassarino et al., 1997, Wallace et al., 1997, Akama et al., 1998, Weldon et al., 1998, Combs et al., 2001, Quinn et al., 2002, McLellan et al., 2003). Neurons containing neurofibrillary tangles (NFTs) in AD express elevated levels of NOS

(Vodovotz et al., 1996). Both Aß and α-synuclein increase the production of H2O2 that is converted to hydroxyl radicals upon exposure to iron (Tabner et al., 2001, Tabner et al., 2002), which is increased in the SN of the PD brain (Sofic et al., 1988, Sofic et al., 1991).

The production of H2O2 and other ROS/RNS is increased with exposure to α-synuclein (Multhaup et al., 1998, Junn and Mouradian, 2002), and α-synuclein-induced increased intracellular ROS/RNS makes cultured neurons more susceptible to apoptosis upon exposure to dopamine (Junn and Mouradian, 2002). Dopaminergic neurons are particularly vulnerable to the processes of oxidative stress. Activated microglia are replete in the SN of PD (Imamura et al. 2003; McGeer et al. 1988). Microglia from elderly human brains activated by LPS are more toxic dopaminergic neurons than other types of neurons, particularly when dopamine release is experimentally induced (Mastroeni et al. 2009); demonstrating that dopamine potentiates neuronal vulnerability. SN neurons pigmented with neuromelanin have a high iron content which increases with advanced age and is particularly high in PD. Iron converts H2O2 to toxic hydroxyl radicals (Hirsch et al., 1991, Good et al., 1992, Tabner et al., 2001, Zecca et al., 2001, Tabner et al., 2002). Furthermore, oligomerization of α-synuclein, the primary pathological protein species in PD, is enhanced by H2O2 (Kang and Kim, 2003). Consequently, pathological protein fibrils and aggregates can cause neuroinflammation and oxidative stress (see Section 1.5.2a), and oxidative stress can chaperone the formation of protein aggregates, amplifying the cycle of neuroinflammation and negative cellular consequences.

1.4.1c TNFα activation of programmed cell death TNFα plays a regulatory role in glutamatergic signaling under normal conditions (see Section 1.3.2c), but when elevated under inflammatory conditions, TNFα not only 27 facilitates the increase in extracellular glutamate, but is the only known cytokine to activate a cascade to programmed cell death directly. TNFα activation of the TNFα receptor 1 (TNFR1) subtype activated the TNFα receptor death domain (TRADD) and downstream signaling that eventually leads to activation of caspase 8 and initiation of apoptosis (Hsu et al., 1996). This mechanism of action is not a main focus of this thesis, but it should be noted that this pathway of cell death is connected to AD. TNFα and TNFR1 levels increase in AD brains, correlate with the level of AD pathology, TNFR1 is necessary for Aß- induced cell death, and a clinical trial using the TNFα-inhibitor etanercept produced modest cognitive improvements in AD patients (Potter 2010; Tobinick 2009; Li et al. 2004; Cheng et al. 2010; Chandler et al. 2010).

Figure 6. Neuroinflammation and glutamate dysregulation: Mechanisms of cell death This schematic illustrates glutamate (Glu) as product of the citric acid cycle (TCA) in astrocytes and conversion to glutamine (Gln) by glutamine synthase (Gln Syn). Glutamine is transported into the synaptic space and taken up by pre-synaptic neuronal terminals where Gln is metabolized by glutaminase to form glutamate. Glutamate is packaged into vesicles by the vesicular glutamate transporter (VGluT) and docked at the synapse by the soluble NSF-attachment protein receptor (SNARE) complex which includes the docking protein synaptosomal-associated protein 25 (SNAP25). Released glutamate activates 2-amino-3-(3-hydroxy-5-methyl-isoxazol-4-yl)propanoic acid receptors (AMPAR) which allow Na+ entry and, if the membrane is depolarized enough to remove of the voltage-dependent Mg+ blockade, N-Methyl-D-aspartic acid or N- Methyl-D-aspartate receptors (NMDAR) which allow Ca2+ influx. Excitation from Ca2+ entry can potentiate excitation to subsequent glutamate release, a process called long- term potentiation (LTP) and a mechanism for memory processing. Excessive glutamatergic signalling, however, elevates intracellular Ca2+ levels and drives excitotoxic cell death. Extrasynaptic NMDARs are more likely to contain the NR2B NMDAR subunit and are more strongly associated with excitotoxicity. Glutamate is cleared from the synapse by the glutamate transporter (GLT1) on astrocytes, but can also be cleared by GLT1 and other excitatory amino acid transporters (EAAT) on microglia and neurons. Glutamate inside astrocytes can be released by reversal of GLT1, hemichannels (grey), TNFα-dependent vesicular release or in exchange for cystine by the cystine-glutamate anti-porter (xCT). Microglial hemichannels also release glutamate. The xCT is found also on neurons and microglia, and intracellular cystine is used to form the antioxidant glutathione (GSH). Activated microglia monitor neurons, which signal their health by surface expression and release of CD200 and fractalkine (CX3CL). Upon activation, microglia transform from an anti-inflammatory ‘resting’ state to a classically activated (M1)/pro-inflammatory state associated with the release of reactive oxygen and nitrogen species (ROS/RNS) as well as prostaglandins, TNFα and the interleukins (IL-) IL-1ß, IL-6 and IL-12 . ROS/RNS cause oxidative stress which damages mitochondria and destabilizes the resting membrane, leading to more reactivity to glutamate and cell death. TNFα, IL-1ß and prostaglandins are involved normal LTP and vesicular release of glutamate from astrocytes that potentiates pre-synaptic glutamate release, but high levels of TNFα and IL-1 elevate extracellular glutamate by attenuating GLT1 function and increasing speed of xCT. TNFα can also act through its receptor (TNFαR) to activate the TNFαR death domain (TRADD), which initiates an intracellular cascade leading to apoptosis. Neuronal death by excitotoxicity, oxidative stress or apoptosis release internal components and leave neuronal debris, which become damage-associated molecular patters (DAMPs) that are recognized by pattern recognition receptors (PRRs) and activate microglia. If not continually activated, microglia will transition into an alternative activation (M2a) state, associated with release of IL-4 and IL-13, and then to an acquired deactivation (M2c) phenotype associated with release of IL-10 and transforming growth factor ß (TGFß).

Figure 6.

1.4.2 Neuroinflammation and glutamate dysregulation are mutually promoting Neuroinflammation is associated with impairments in hippocampal-sensitive learning and memory tasks, which are primarily dependent upon glutamatergic signaling. Neuroinflammatory factors, such as the pro-inflammatory cytokines TNFα and IL-1ß, do not impair memory directly, but neuroinflammation likely impairs memory through dysregulation of glutamatergic neurotransmission as well as accumulated cellular dysfunction and death over time. Glutamatergic activity is elevated under conditions of chronic neuroinflammation. To illustrate this point, activated microglia release glutamate, TNFα and IL-1ß amplify glutamatergic signaling and increase extrasynaptic glutamate availability and oxidative stress increases xCT export of glutamate. Glutamate, in turn, drives energetic failure and neuronal cell death that stimulates the neuroinflammatory process. Activated microglia release glutamate into the extracellular environment. Glutamate is released from cultured microglia activated by LPS or TNFα (Piani & Fontana 1994). Interestingly, LPS-activation increases microglial expression of the glutamate transporter GLT1 in vitro (Persson et al., 2005), and it is not understood if this functions to further activate microglia or to negatively regulate microglial glutamate release. LPS- stimulated glutamate release from microglia requires the oxidative burst associated with oxidative stress (Barger et al. 2007). Oxidative stress is also associated with increased xCT function, which promotes production of the antioxidant glutathione, but also additional export glutamate to the extracellular space. Glutamate released by TNFα- or LPS-activated microglia in vitro is produced within microglia and released through hemichannels, as glutamate release is blocked in a dose-dependent manner by either a glutaminase inhibitor or a gap junction blocker (Shijie et al. 2009; Takeuchi et al. 2008; Takeuchi et al. 2006a). Astrocytes, like microglia, play an active role in the neuroimmune response (Cerbai et al. 2012) and may be credited with significant regulatory control over extracellular glutamate (see Section 1.3.2). In addition to stimulating microglia activation, and being released by activated microglia, the pro-inflammatory cytokines TNFα and IL-1 regulate homeostatic glutamatergic neurotransmission, and may elevate glutamate under inflammatory conditions. Low, physiological levels of pro-inflammatory TNFα and prostaglandin E2 (PGE2) are necessary for vesicular release of glutamate from astrocytes 31

(Santello et al. 2011; Bezzi et al. 1998; see Section 1.3.2c) which potentiates pre-synaptic glutamate release and facilitates LTP (Parpura et al. 1994; Volterra & Steinhäuser 2004). These combined processes may lead to excessive extracellular glutamate under pro- inflammatory conditions in which TNFα and PGE2 are elevated. For example, when LPS- activated microglia are added to the astrocyte cultures, astrocytic glutamate release is amplified 4-fold (Bezzi et al. 2001). In addition, TNFα upregulates microglial glutamate synthesis and glutamate release through hemichannels (Takeuchi et al. 2006b). In this way, TNFα released from either microglia or astrocytes sustains a pro-inflammatory environment and promotes astrocytic, neuronal and microglial release of glutamate. Like TNFα, low physiological levels of IL-1ß are required for the normal induction of LTP and successful performance in tasks sensitive to hippocampal damage (Avital et al. 2003; Goshen et al. 2007). IL-1ß enhances pre-synaptic glutamate release and the post- synaptic response to glutamate, reduces glutamate clearance and elevates astrocytic glutamate release. IL-1ß triggers neuronal glutamate release in vitro (Liu et al. 2011), dose-dependently increases surface expression of glutamate AMPARs (Lai et al. 2006) and increases Ca2+ entry through hippocampal NMDARs by (Viviani et al. 2003; Wang et al. 1999). In addition to potentiation of neuronal glutamate signaling, IL-1ß increases the velocity of cystine-glutamate exchange by xCT, which increases cystine uptake and elevates extracellular glutamate levels in murine cortical cultures (Fogal et al. 2007). Furthermore, IL-1ß reduces GLT1 on astrocytes and dose-dependently decreases glutamate uptake (Prow & Irani 2008; Hu et al. 2000; Sama et al. 2008; Ye & Sontheimer 1996). The combined actions of TNFα and IL-1ß on glutamate include: potentiation of pre-synaptic release, increased expression of AMPARs and NMDARs, reduced glutamate clearance, enhanced xCT glutamate export, increased vesicular glutamate release from astrocytes, and elevated glutamate release from microglial hemichannels; all of which could lead to a hyperglutamatergic state under neuroinflammatory conditions in which TNFα and IL-1ß are dramatically elevated. Taken together, these observations suggest that the pro-inflammatory cytokines TNFα and IL-1ß disrupt hippocampal-dependent learning and memory through exacerbation of glutamatergic neurotransmission and may promote excitotoxic cell death.

Dysfunction of glutamate transport may be both a consequence of and contributor to the neurodegenerative process. TNFα, caspase-3 and reactive oxygen species (Trotti et al. 1998), all elevated in an inflammatory environment, reduce the function of GLT1. The pro-inflammatory cytokine TNFα suppresses the GLT1 gene (Sitcheran et al. 2005). Caspase-3 is a protease involved in inflammation, apoptosis and necrosis, is the primary caspase that cleaves amyloid precursor protein in AD. Caspase-3 also cleaves GLT1 and leads to a reduction in glutamate uptake (Boston-Howes et al., 2006); likely resulting in increased extracellular glutamate levels. Astrocytes exchange intracellular glutamate obtained by EAATs with extracellular cystine through the cystine/glutamate antiporter, and transform the intracellular cystine to cysteine and then the antioxidant glutathione (GSH). Elevated extracellular glutamate concentrations can block the cystine/glutamate exchanger, leading to reduced production of the antioxidant glutathione (Lewerenz et al., 2006 & 2009). In summary, loss of GLT1 function likely leads to reduced protection for neurons from the anti-oxidant glutathione, an elevation in extracellular glutamate, both of which may increase the probability of excitotoxic neuronal death. Failing neurons and debris from dead neurons stimulate microglia activation and promote an inflammatory environment. Activated microglia release reactive oxygen species that further contribute to oxidative stress. Moreover, reactive oxygen species inhibit glutamate uptake in cultures astrocytes (Trotti et al. 1998). The connection observed between immune activation and increased glutamatergic signaling at the synaptic level as well as decreased clearance are supported by evidence on the functional and behavioral levels. Peripheral or central injection of IL-1ß drives a measurable increase in extracellular glutamate in rats (Chao et al. 1995; Kamikawa et al. 1998; Mascarucci et al. 1998). IL-1ß concentration directly correlates with hyperexcitability of glutamatergic circuitry in multiple sclerosis patients (Rossi et al., 2012). Innate immune activation by LPS drives an increase in evoked cortical excitability that develops into seizures in rats within 30 minutes, as well as deficit in LTP, that are both blocked by treatment with IL-1Ra (Rodgers et al. 2009; Chapman et al. 2010). In contrast to acute, local application of LPS, Dr. Gary Wenk’s lab has used a model of prolonged continuous i.c.v. LPS infusion. Results from this model demonstrate an increase in Arc expression due to post-synaptic Ca2+ influx and disruption of LTP and spatial memory that

33 are alleviated by drug treatments that reduce pre- or post-synaptic glutamatergic signaling as well as a TNFα-synthesis inhibitor (Hauss-Wegrzyniak et al. 1998; Hauss-Wegrzyniak et al. 2002; Brothers et al. 2010a; Rosi et al. 2006; Marchalant et al. 2009; Rosi et al. 2005; Tweedie et al. 2012; Belarbi et al. 2012). Similarly, Arc mRNA is elevated in glutamate- overexpressing transgenic mice (Michaelis et al. 2011). Taken together, these data suggest that the pro-inflammatory environment created by the innate immune response elevates glutamatergic activity, leading to increased cortical excitability, disruption of LTP and working memory, seizure activity and neuronal death. Excessive glutamatergic signaling, in turn, leads to reciprocal neuroimmune activation. For example, microglia release pro-inflammatory cytokines when exposed to extracellular glutamate (Taylor et al. 2005), and stimulation of neuronal NMDARs elevates neuroinflammatory markers including IL-1ß, TNFα, and iNOS (Chang et al., 2008). Likewise, death of neurons in an inflammatory/oxidative/excitotoxic environment stimulates a pro-inflammatory immune response. For example, microglia stimulated by necrotic hippocampal neurons increase expression of glutaminase, an enzyme that produces glutamate, and the elevated activity of this enzyme leads to enhanced extracellular levels of glutamate (Pais et al. 2008). Furthermore, the AD-related protein, Aß, drives neuronal cell death in vitro only in the combined presence of NO and stimulation of TNFαRs and NMDARs; none alone is sufficient to cause cell death (Floden et al. 2005). Taken together, these observations demonstrate that neuroinflammation, oxidative stress and glutamatergic excitation form a cycle of mutual activation, and work in concert to disrupt neurotransmission, system function and behavior, as well as lead to cell death.

1.5 Neuroinflammation and glutamate dysregulation in Alzheimer’s and Parkinson’s diseases AD and PD are distinguished by different principal pathology, regions of degeneration, etiology and clinical features, yet, there is significant overlap in each of these domains as well. I have chosen to investigate neuroinflammation and glutamate regulation as they relate to AD and PD, and therefore will provide background in this section on the characteristics of AD and PD, as well as aging, a shared risk factor for both. Neurons in AD and PD may be particularly vulnerable to the negative consequences of

34 neuroinflammation; the hippocampus to glutamate dysregulation and the nigra to oxidative stress, respectively. More emphasis is placed on AD in this section, because more of the experiments described in this thesis focus on AD. This section will also provide evidence of neuroinflammation and glutamatergic dysregulation in these diseases and address pharmacological trials and drugs currently in use.

1.5.1 Features of AD and PD

1.5.1a Alzheimer’s disease AD is defined by a progressive loss of memory and the development of dementia. The symptoms are related to pathology that occurs in brain regions such as the cortex, basal forebrain, and temporal lobe. Atrophy of temporal lobe regions, including the entorhinal cortex and hippocampus occurs early in AD and contributes to memory impairment in AD. The studies in this thesis focus on the hippocampus as it related to AD because it has a propensity to develop neuroinflammation with age and in models of experimentally induced neuroinflammation (Hauss-Wegrzyniak et al. 1998). The hippocampus is an area regarded as important for the formation of episodic and spatial memory, and LPT (Section 1.3.1) has been studied and defined in this region of the brain. Damage to the hippocampus may cause both retrograde and anterograde amnesia. The hippocampus is comprised of subregions that include Cornu Ammonis areas 1 and 3 (CA1, CA3) and the dentate gyrus (DG), through which a circuit is made and information flow is typically unidirectional (Figure 7). Input from the entorhinal cortex (EC) arrives at the DG region of the hippocampus through the perforant path. The granule cells of the DG project through mossy fibers forward to the pyramidal cells of the CA3 region. From CA3, information projects to CA1 and the subiculum, and finally out to the EC, completing a trisynaptic circuit within the hippocampus.

Figure 7. Hippocampal connectivity Inputs arrive from the entorhinal cortex (EC) layer II and follow the perforant pathway (PP) to the dentate gyrus (DG). The DG sends information through the mossy fibers (MF) to the pyramidal cells of the CA3 region which projects through its Schaffer collaterals (SC) to the CA1 region. From the CA1 information travels to the subiculum (Sb) and out of the hippocampus to layer III of the EC.

Pathology in AD includes two types of protein aggregates, plaques which are formed in the interstitial space by aggregates of inappropriately cleaved amyloid ß (Aß) and neurofibrillary tangles (NFTs) which are formed from intracellular fibrils of the hyperphosphorylated tau protein. Plaques and tangles were first described by the psychiatrist Alois Alzheimer in 1906, after examining the tissue of his famous patient, Auguste D. Plaques and tangles develop in the same temporal lobe regions that show the most atrophy in AD. Aß is formed by amyloid precursor protein that is cleaved by secretases to form multiple isotopes, primarily Aβ40. However, a more fibrillogenic form

Aβ42, can be cleaved, and it is this form that aggregates in the extracellular space and develops into plaques. Some genes that have been identified with early onset AD skew amyloid precursor protein cleavage to this pathogenic form. It is unknown if Aß plaques contribute directly to neuronal death. Therefore, it is still debated whether plaques are a cause or a symptom of AD. Unlike Aß plaques, NFTs form within the cell body of the neuron. NFTs are formed by aggregates of hyperphosphorylated tau, a protein that is associated with microtubules. When hyperphosphorylated, tau forms paired helical 36 filaments that aggregate and may directly contribute to neuronal dysfunction or death. Plaques and NFTs do not correlate strongly with the onset or severity of symptoms, although NFTs do correlate with cognitive deficit better than do plaques (Lue 1996). In fact, plaques and NFTs are present in approximately 30% of normal aged adults with no symptoms of dementia (Fukumoto et al. 1996; Fukumoto et al. 1996; Zhan et al. 1995). Neuroinflammation is stimulated by the presence of plaques, the Aß fibrils that compose plaques as well as NFTs. Neuroinflammation correlates with the severity of symptoms, and precedes the appearance of plaques and NFTs (Lue 1996). Neuroinflammation and pathological protein aggregates are key features of AD as well as PD and may contribute to progressive degeneration and functional loss in both diseases.

1.5.1b Parkinson’s disease PD is characterized by a progressive motor impairment that includes symptoms such as resting tremor, bradykinesia, the inability to initiate movement, and eventual akinesia. Individuals with PD often suffer cognitive loss that may progress to dementia. The motor symptoms of PD are attributed to loss of dopaminergic neurons in the substantia nigra (SN) pars compacta region (SNpc) and the loss of its inhibitory input to the striatum in the basal ganglia motor loop. The SN is a brainstem nucleus of dopaminergic neurons in the basal ganglia, a distributed system that plays a role in movement (Figure 8). The SN projects to the striatum and modulates the response of the striatum to input from the motor cortex. Glutamatergic input from the cortex to the striatum diverges into two pathways, both ultimately influencing inhibition from the globus pallidus upon the thalamocortical projections that facilitate movement. Generally, the direct motor pathway increases motor output and the role of the SN is to facilitate the influence of this pathway, while the indirect motor pathway decreases motor output and the SN diminishes its activity. Through the direct pathway, projections from the striatum inhibit the inhibitory projection from the GPi to the thalamus, resulting in increased motor output. Through the indirect pathway, striatal output strengthens inhibition from the GPi upon the thalamus, resulting in decreased motor output. The SNpc facilitates striatal output to the direct pathway and reduces striatal output to the indirect pathway, skewing the basal ganglia motor loop toward the release of GPi inhibition upon the thalamus, and facilitation of motor output through the thalamocortical projections. 37

The loss of the SNpc dopaminergic projections bias the motor loop toward the indirect pathway and overall inhibition of motor output, leading directly to some of the motor symptoms seen in PD such as rigidity, bradykinesia, akinesia and postural instability.

Figure 8. Effect of SN loss in PD on the basal ganglia motor pathway The dopaminergic (DA) neurons of the substantia nigra pars compacts (SNpc) project to the striatum where they positively modulate the direct pathway (solid lines) through the globus pallidus external (GPe) and suppress the indirect pathway (dashed lines) which synapses in the GPe and the subthalamic nucleus (STN). SNpc suppression of the inhibitory motor pathway releases the thalamus of inhibition from the internal globus pallidus (GPi) and facilitates stimulatory output to the motor cortex. Loss of dopaminergic projections from the SNpc biases this circuitry toward the indirect pathway and inhibition of thalamocortical projections.

Symptoms of PD occur after a threshold of 80% loss of SNpc neurons; demonstrating the capacity of this system to compensate for the loss of dopamine. Interestingly, while dopaminergic neurons degenerate in the SNpc, little or no loss is seen in the adjacent dopaminergic neurons of the ventral tegmental area. In fact, this distinction is used to validate animal models. The reasons why the SNpc, but not the VTA are vulnerable to degeneration are not well understood. Neuroinflammation is a factor that may contribute to loss of dopaminergic neurons in the SNpc. Like AD, neuroinflammation in PD interacts with pathological protein aggregates. Like NFTs and plaques in AD, protein aggregates called Lewy bodies (LBs) are a common feature of PD. LBs are formed by fibrils of inappropriately cleaved α-synuclein protein and produce intracellular aggregates similar to NFTs formed by hyperphosphorylated tau. The function of α-synuclein is unknown, but it may be a microtubule associated protein similar to tau. LB inclusions, like NFTs, have the potential to directly cause cell death, although it is unknown whether the formation of LBs is a trigger for or an event subsequent to the initiation of neuronal cell death. There is an interesting overlap between the protein aggregates associated with AD and PD and cognitive deficits (Figure 9). LBs are found in PD with dementia, and the related Lewy Body Dementia expresses both LBs and plaques. AD primarily expresses plaques and NFTs, though occasionally LBs are found (Kaufer & Tröster 2008; Tröster 2008). This supports the idea that the propensity to host inappropriately processed protein aggregates is tied to the development of neurodegeneration, particularly the development of dementia. Neuroinflammation is commonly associated with pathological protein aggregates that may facilitate their development. Neuroinflammation is also triggered by pathological protein deposition as well as the components that comprise plaques, NFTs and LBs and may be one mechanism by which protein aggregates lead to neurodegeneration and loss of function in AD and PD. These pathologies are not restricted to the hippocampus and SN, but are also observed in other brain regions, including the locus coeruleus (LC) and raphe nucleus which are affected in both AD and PD.

Figure 9. Protein aggregates in AD and PD related dementias Protein aggregates including amyloid ß plaques, tau neurofibrillary tangles and Lewy bodies are observed in Alzheimer’s disease, Parkinson’s disease with dementia and related Lewy body dementia (Tröster, 2008).

1.5.1c Locus coeruleus and raphe nuclei in AD and PD Neuronal degeneration in AD and PD is not limited to the brain regions discussed above. Loss of cells in other regions, including noradrenergic cells in the locus coeruleus (LC) and serotonergic cells in the raphe nuclei have been observed (Marcyniuk et al. 1986a; Marcyniuk et al. 1986b; Marcyniuk et al. 1988; D’Amato et al. 1987; Bertrand et al. 1997). Degeneration of these brainstem nuclei may contribute to dysregulation of sleep and arousal, a common symptom of AD and PD, as well as depression which shows a high comorbidity with AD and PD and often precedes other disease symptoms. When postmortem tissues of patients with AD, PD and AD/PD were studied, NFTs and LBs were observed in the LC and raphe as well as cell loss and the reduction of serotonin receptor binding in the raphe (D’Amato et al. 1987). The raphe nuclei has been shown to degenerate in AD, however data is lacking, in part because the raphe nucleus has not been included in postmortem staging systems such as Braak and Braak (Simic et al. 2009). Simic et al (2009) review the role in and susceptibility of the dorsal raphe nucleus to degeneration in AD, noting that the raphe is one of the first areas to degenerate in AD and hypothesizing 40 that degeneration in the dorsal raphe proceeds degeneration in the entorhinal cortex and hippocampus. Because the SN is easily seen, the loss of the SN is so severe, and because the connection to motor symptoms is well understood, it has been extensively studied in PD, while the LC and raphe have been frequently overlooked. The LC, raphe, hippocampus and SN share neuroinflammation as a common factor in AD and PD. The following sections will describe evidence of inflammation in AD and PD as well as ways in which neuroinflammation may mediate disease processes.

1.5.2 Neuroinflammation in AD and PD The pathology and symptoms of neurodegenerative diseases differ, but they share a common component, neuroinflammation. Neuroinflammation has been noted in a vast range of neurodegenerative diseases, including AD, PD, multiple sclerosis, prion disease, viral dementia (AIDS dementia), age-related macular degeneration, traumatic brain injury, amyotrophic lateral sclerosis and neuropathic pain (Wood, 2003). Increased levels of neuroinflammation are often detectable prior to the onset of symptoms and other disease features, such as the development of plaques and tangles in AD (Cagnin et al. 2001; Cagnin et al. 2004; Cagnin et al. 2006). It is important to investigate the ways in which neuroinflammation may contribute to disease onset and progression. While the factors that initiate and perpetuate neurodegenerative diseases are varied, the possibility of limiting a shared common factor, such as inflammation, has broad implications for the prevention and treatment of all neurodegenerative diseases. The following section will concentrate on clinical data from human participants, review evidence that neuroinflammation is elevated in AD and PD, summarize epidemiological evidence that limiting neuroinflammation reduces the risk of developing AD or PD, and discuss interventional trials with anti-inflammatory treatments.

1.5.2a Factors that initiate neuroinflammation in AD and PD Activated microglia increase in number with aging and in the brains of patients with AD and PD; correlating more strongly with cognitive and motor impairments than other disease pathology (McGeer & McGeer 1998; Imamura et al. 2003; Edison et al. 2008b; Griffin et al. 1998; Sheng et al. 1998; Sheffield & Berman 1998; Szpak et al. 2001; Xiang et al. 2006). Activated microglia in AD and PD brains are frequently found near

41 pathological protein aggregates, such as Aß plaques, hyperphosphorylated tau NFTs and α-synuclein LBs, as well as in the vicinity of dying and dead neurons. Increased microglia activation precedes these protein aggregations and neuronal death. In addition, pathological protein aggregates, their components and dying/dead neurons promote microglia activation. Protein fibrils and aggregates found in AD and PD brains including Aß, tau and α- synuclein initiate microglia activation (Lue et al. 2001; Shen et al. 2001) and phagocytosis (Lue et al. 2001; Lue et al. 2001; Shen et al. 2001; Sheng et al. 1997; Strohmeyer & Rogers 2001; Rogers 2002; Lewandowska et al. 2004). The receptor for advanced glycation endproducts (RAGE) is found on both microglia and neurons in the hippocampus and entorhinal cortex and is increased in AD brains. RAGE can be activated by Aß, and the consequences include transcription of nuclear factor κB (NFκB) and the subsequent production of pro-inflammatory factors (Lue et al. 2001). Microglia in culture activate upon exposure to α-synuclein in a dose-dependent manner, which leads to degeneration of dopaminergic cells in mesencephalic cultures (Zhang et al. 2005) and microglia are activated in the brains of mice that over-express α-synuclein (Su et al. 2008). These studies demonstrate that microglia become activated in response to pathological fibrils and aggregates of pathological protein species characteristic of AD and PD; microglia also become activated in response to dying or dead neurons in these diseases. Neurons participate in a dialogue with microglia, constitutively expressing signals that indicate health and keep microglia in a quiescent state, and actively releasing pro- inflammatory messengers upon distress (Lynch 2009). Surface markers found on neurons that maintain microglia quiescence by interaction with receptors on microglia include fractalkine and CD200 (Lyons et al. 2009; Koning et al. 2009; Lyons et al. 2007; Cox et al. 2011; Pabon et al. 2011). Dying cells reduce neurotransmitter production, release signals of distress and ultimately become cellular debris, all of which interact dynamically with microglia, initiating activation and phagocytosis. Necrotic hippocampal neurons initiate activation of cultured microglia, demonstrated by surface markers and release of pro-inflammatory factors including TNFα, cyclooxygenase-2 (COX2), and iNOS; and microglial activation leads to the subsequent degeneration of cultured cerebellar neurons (Pais et al. 2008). Dopaminergic neurons in the SN have a high content of the pigment

42 neuromelanin and release this pigment upon cell death. Neuromelanin activates microglia and induces microglia production of the ROS/RNS species NO, superoxide and H2O2 as well as pro-inflammatory cytokines. Extracellular neuromelanin leads to neurodegeneration, both in vitro and in vivo, that is dependent upon the presence of activated microglia (Zhang et al. 2009).

1.5.2b Neuroinflammation observed in AD and PD The number of activated microglia increases with aging, increases to a greater extent in AD and PD, occurs in regions that later degenerate, precedes pathological protein accumulation and correlates with cognitive decline. Increased numbers of activated microglia may predispose vulnerable regions to later express pathology and degenerate in AD and PD. The importance of microglia in AD was first described by Alois Alzheimer in 1910. He noted that microglia appear in proximity to degenerating neurons and conjectured that they work to solubilize debris and “clear the nervous tissue from waste” (Kettenmann and Ransom, 2005; Figure 10). Twenty years later, Pio del Rio Hortega also proposed a role for microglia as phagocytes, and more than 100 years later, we are still striving to understand this relationship. Whether microglia protect the brain through phagocytosis of plaques (Lee & Landreth 2010; Simard et al. 2006; D’Andrea et al. 2004; Liu et al. 2005; DeWitt et al. 1998), or become senescent with age and fail to protect the brain from plaques (Hickman et al. 2008; Streit et al. 2009) is still debated. Although the relationship between microglia and Aß is not concrete, the observations that microglia activation precedes plaque pathology and correlates better with dementia symptoms than plaques, in combination with epidemiological evidence that preventing neuroinflammation is protective against developing AD, suggests that neuroinflammation is a more proximal cause of AD, and a more important pharmacological target. The number of activated microglia increases (20-35%) in the frontal, temporal, parietal, occipital and cingulate cortices of individuals with age and more so in those diagnosed with AD as revealed by PET scan (Edison et al. 2008a; Cagnin et al. 2001); and activated microglia are distributed mostly in the hippocampus followed by the cortex (Sheffield et al. 2000; Xiang et al. 2006). Moreover, neuroinflammation in the hippocampus and entorhinal cortex precedes the deposition of Aß and formation of NFTs, 43 with the most pathology in the regions of greatest microglia activation (Edison et al. 2008a; Cagnin et al. 2001; Sheffield et al. 2000; Sheffield & Berman 1998; Sheng, et al. 1998). Furthermore, dementia and cognitive impairment correlates with neuroinflammation when evaluated by PET or in post-mortem tissue and compared to clinical evaluations (Xiang et al. 2006; Edison et al. 2008a; Cagnin et al. 2001).

Figure 10. Plaques and Tangles Plaques surrounded by microglia as first described and drawn by Alois Alzheimer in 1911 (Kettenmann and Ransom, 2005).

Activated microglia are present in aging, AD, Lewy Body Dementia, Parkinson’s with dementia and PD; with more activated microglia in the hippocampus of those with symptoms of dementia, in the SN of those with motor impairment, and frequent overlap in distribution of activated microglia between these regions and diagnoses (McGeer et al., 1988, Szpak et al., 2001, Imamura et al., 2003). Activated microglia in the SN and striatum of PD brains increase as neurodegeneration progresses, are associated with LBs but also are present in the absence of LBs, and are located near dopaminergic, serotonergic and α- synuclein positive neurites (Imamura et al., 2003). Work presented in this thesis will confirm the presence of activated microglia surrounding the dopaminergic neurons of the SN, the serotonergic neurons of the raphe as well as the noradrenergic neurons of the LC.

Pro-inflammatory cytokines are released by activated microglia, are elevated in aging, and are risk factors for cognitive decline. In older adults, the pro-inflammatory cytokines TNFα and IL-1ß are negatively correlated with total brain volume in (Jefferson et al., 2007). A high level of serum pro-inflammatory IFNƴ is associated with delirium in older adults; oppositely, a low level of anti-inflammatory IL-1Ra is associated with normal cognitive function (Adamis et al., 2009). Polymorphisms in the TNFα and IL-1ß genes are associated with processing speed and general cognitive performance in normal aging (Benke et al., 2011; Sasayama et al., 2011; Tsai et al., 2010; Baune et al., 2008); a connection that is not surprising given the role of IL-1ß in regulation of LTP (see Section 1.4.2). Pro-inflammatory cytokine levels are increased in the serum, CSF and post-mortem brain tissue of AD and PD patients, indicating that neuroinflammation is involved in the progression of the disease process. AD plasma, CFS and brains are characterized by increased expression of pro-inflammatory cytokines such as TNFα, IL-1ß, IL-6 and IL-18 (Tan et al., 2007, Di Bona et al., 2009, Ojala et al., 2009; Forlenza et al., 2009; Malaguarnera et al., 2006; Olgiati et al., 2010), and decreased expression of the anti- inflammatory messenger CD200 (Walker et al. 2009). Similarly, increased plasma levels of IL-6 and RANTES are correlated with decreased function in PD patients (Rentzos et al., 2007, Hofmann et al., 2009). Polymorphisms in cytokine genes correlate with increased susceptibility to cognitive impairment as well as AD and PD. Certain TNFα polymorphisms increase the likelihood of developing dementia, particularly in those that carry another genetic variable related to AD, ApoE4 (McCusker et al., 2001, Ma et al., 2004, Lio et al., 2006, Di Bona et al., 2009, Yang et al., 2009), and some of the same polymorphisms are related to increased risk of early onset PD (Wu et al., 2007b, Bialecka et al., 2008). Similarly, IL-1 polymorphisms are correlated with increased microglia in AD that interacts with ApoE4 genotype and age-dependent increased risk of AD and PD (Hayes et al., 2004; Sciacca et al., 2003, Licastro et al., 2004, McCulley et al., 2004, Rainero et al., 2004, Nishimura et al., 2005, Wang et al., 2005, Wu et al., 2007a, Di Bona et al., 2008; Arman et al., 2010). Like TNFα and IL-1, polymorphisms of IL-6 and IL-18 are associated with increased risk of AD and PD (Licastro et al., 2003, Zhang et al., 2004, Håkansson et al., 2005, Bossù et

45 al., 2007). Polymorphisms in genes for cytokines do not necessarily imply that expression of those cytokines is elevated or that their function is increased, but these polymorphisms in AD and PD do connect changes in pro-inflammatory cytokines with disease risk. Genetic variance in COX2, like the pro-inflammatory cytokines discussed above, also impacts risk of AD (Abdullah et al., 2006). COX is an enzyme that synthesizes prostaglandins from arachidonic acid. COX1 is constitutively expressed in most cells and COX2 is induced by iNOS in neurons as part of the inflammatory process. Expression COX2 is increased in the frontal cortex and temporal lobe of AD brain and correlates with amyloid plaque density in the hippocampus (Pasinetti and Aisen, 1998, Ho et al., 1999).(Hoozemans et al., 2001). COX2 levels increase with normal aging in the hippocampal CA3 region, subiculum and entorhinal cortex, and increase in the CA1 hippocampal sub-region of patients with AD compared to non-demented controls (Fujimi et al., 2007). The COX enzyme is inhibited by non-steroidal anti-inflammatory drugs (NSAIDs), and epidemiological evidence supports the hypothesis that reducing inflammation is protective against the risk of developing AD or PD.

1.5.2c Epidemiology of neuroinflammation in AD and PD Neuroinflammation alone is not likely to cause AD or PD; however, it is regarded as a risk factor that might contribute to the onset and progression in individuals whom are predisposed to develop these diseases. In at least one example, neuroinflammation is strongest associated etiological factor of dementia symptoms: dementia pugilistica is a syndrome that develops in some boxers after years of head trauma. These athletes present a spectrum of symptoms that including plaques, NFTs, dementia and pseudo-parkinsonism. Neuroinflammation was fist considered as a potential risk factor for neurodegenerative diseases based upon epidemiological evidence indicating that individuals who experience excessive neuroinflammation, such as the boxers described above, are prone to develop neurodegenerative diseases, while those that had a history of anti-inflammatory use were less likely to develop disease symptoms. The observation that individuals with rheumatoid arthritis who take relatively high doses of NSAIDs have a lower occurrence of AD and PD than the general population seeded a field of investigation into the role of neuroinflammation in AD and PD (McGeer et al., 1996, McGeer virtual

46 article). The following sections will summarize retrospective and protective studies that evaluate the protection of anti-inflammatory use against the development of AD or PD.

Epidemiological evidence of of protection against AD by anti-inflammatory use Epidemiological evidence shows that the use of NSAIDs and other anti- inflammatory agents during mid-life reduces the risk of developing AD (McGeer et al., 1996, Stewart et al., 1997, Zandi and Breitner, 2001, Etminan et al., 2003, Launer, 2003, Gasparini et al., 2004, Szekely et al., 2004). These studies indicate that protection by NSAIDs is independent from the ability of some NSAIDs to reduce Aß (Szekely et al., 2008a, Szekely et al., 2008b) and an Aß-lowering drug, tarenflurbil, did not slow cognitive decline in a recent clinical trial (Green et al., 2009). Taken together, these results support the hypothesis that neuroinflammation contributes to the disease process and preventing inflammation is an effective way to reduce the incidence of AD and PD. Retrospective studies in cohorts and twin pairs have found an inverse correlation between the use of NSAIDs and corticosteroids with the risk of developing AD (Breitner et al., 1994, Landi et al., 2003, Nilsson et al., 2003, Yip et al., 2005). Duration of NSAID use was evaluated in 3,227 participants in the Cash County Study and it was found that early use of NSAIDs, but not recent use of NSAIDs, protected individuals from AD and this protection was increased with duration of use, especially greater than 2 years; suggesting that reducing inflammation is important early in the development of disease and may be ineffective once the disease had progressed past a certain point (Zandi et al., 2002). Prospective studies that follow anti-inflammatory drug use and clinical diagnosis, sometimes including retrospective data as well, have supported the idea that use of anti- inflammatories reduce risk of developing AD. Use of NSAIDs in 4,615 cognitively normal individuals resulted in fewer diagnoses of AD after 5 years (Lindsay et al., 2002). The Rotterdam Study, a prospective study initiated in 1990, has reported that NSAID use reduces the risk of AD, particularly if the duration of use was at least 6 months (Andersen et al., 1995, Veld et al., 1998). It was recently reported in this cohort that statins, lipid- lowering drugs that also exhibit anti-inflammatory properties, reduced the incidence of AD, in contrast, non-statin cholesterol lowering drugs were without benefit, further supporting the role of anti-inflammatory use in AD prevention (Haag et al., 2009). The Baltimore Longitudinal Study of Aging reported similar results in cohort of 1,686 over a follow-up 47 period of 15 years, noting that AD risk decreased significantly in those that reported 2 or more years of NSAID use, however, no such association was found with acetaminophen which does not have strong anti-inflammatory properties and does not cross the blood- brain barrier (Stewart et al., 1997). Genotype can interact with the degree of protection of NSAID use against AD. ApoE is a protein involved in lipoprotein metabolism and persons with AD are more likely to be carriers of the allele for the ApoE4 isoform. The Cache County Study showed that carriers of ApoE4 gained the most benefit from NSAID use in combination with vitamins E and C; these treatments decreased decline in their cognition more than non-carriers (Fotuhi et al., 2008). Similarly, the Cardiovascular and Health Study showed that NSAIDs significantly reduced the risk of AD only in carriers of ApoE4. In contrast, other epidemiological studies have found no significant correlation between the use of NSAIDs and reduced risk of AD. NSAID use 3 years prior to the onset of AD symptoms did not correlate with decreased risk of AD in the Canadian Study of Health and Aging (Wolfson et al., 2002). A recent prospective study of the Adult Changes in Thought cohort followed 2736 non-demented elderly persons for 12 years, and reported that heavy NSAID use of ≥ 500 mg/day increased the incidence of AD. The authors postulate that heavy NSAID use delays the onset of AD in younger cohorts, but does not prevent the eventual development of AD, which would explain why more diagnoses may be appear in their cohort which was older than most studied (Breitner et al., 2009). These negative findings emphasize the need to distinguish between classes of anti- inflammatories, specific compounds within each class, dose and duration of use as well as to distinguish between characteristics of the subjects, for example, presence of the ApoE4 allele and both age and diagnosis during anti-inflammatory drug use or administration.

Epidemiological evidence of protection against PD by anti-inflammatory use Epidemiological evidence for anti-inflammatory use and protection against PD is not as strong as that for AD, possibly because there have been fewer epidemiological evaluations and prospective studies. A recent meta-analysis of 11 studies found a slight protective effect of ibuprofen but no protection from NSAIDs in general (Samii et al., 2009).

In a retrospective study, non-aspirin NSAIDs were reported to reduce the risk or PD in men but increase risk of PD in women in a cohort of over 1,258 PD and 6,638 controls from General Practice Research Database found that (Hermann et al., 2006). No protective effect of NSAID use was found in three recent retrospective studies (Etminan et al., 2008), although one reported a correlation between non-aspirin NSAIDs, cortisone and prednisone use and reduced PD risk, but this correlation did not reach significance (Bower et al., 2006). In contrast, another study reported that aspirin but no other NSAIDs correlated with reduced risk of PD, results which also did not reach significance (Ton et al., 2006). A retrospective study that examined statin use reported no protective effect (Becker et al., 2008). Prospective studies report conflicting results of protection by NSAIDs upon PD. No correlation between NSAID use and PD risk was found in a large, long-term prospective study with a cohort of 6,512 and average follow-up of 9.4 years (Bornebroek et al., 2007). However another large scale prospective study of 142,902 individuals from the Health Professionals Follow-up Study and Nurses' Health Study cohorts found that those who reported regular use of non-aspirin NSAIDs had a 45% lower risk of PD after a follow-up of at least 14 years, and this effect did not differ between men and women (Chen et al., 2003). Taken together, these studies suggest that reducing inflammation may protect against PD. Epidemiological studies of the effect of anti-inflammatory use upon the risk of PD share the same weaknesses with similar studies in the AD field. As we have seen, it is important to consider the drug type, dose, duration and timing of use as well as subject variables including genetic profiles and age. Although the epidemiological evidence of anti-inflammatory use and PD risk is equivocal, reducing neuroinflammation remains an attractive therapeutic target. In PD as well as AD, neuroinflammation is evident early in the disease process, may be driven by disease pathology such as protein aggregates and dying cells, and may contribute to the process of cell death and loss of function.

1.5.2d Neuroinflammation as a pharmacological target Current treatments for both AD and PD are largely palliative, and do nothing to modify disease progression. AD is primarily treated with acetylcholinesterase inhibitors (AchEIs), which block the metabolism of acetylcholine, prolonging and enhancing the 49 effects of acetylcholine in the cortex. These drugs do not rescue degenerating neurons. Similarly, pharmacological treatments for PD enhance or prolong the effects of dopamine from the SN. They do so by increasing dopamine production from the precursor L-DOPA or by preventing dopamine catabolism by the monoamine oxidase (MAO). Drugs for PD do not slow the loss of dopaminergic neurons in the SN, but simply enhance the function of the remaining neurons. If neuroinflammation contributes to cell death in these neurodegenerative diseases, then it represents a pharmacological target that may slow neuronal loss and disease progression. There is some evidence that AchEIs have anti-inflammatory properties as well, and may potentially act through acetylcholine at receptors on lymphocytes to reduce cytokine release (Borovikova et al., 2000). Donepezil, an AchEI, has been shown to reverse high levels of the pro-inflammatory cytokines IL-1ß, IL-6 and TNF and increase levels of the anti-inflammatory cytokine IL-4 produced by peripheral blood mononuclear cells after treatment in AD patients for one month (Gambi et al., 2004, Reale et al., 2004, Reale et al., 2005, Reale et al., 2006). Whether anti-inflammatory activity of AchEIs are a proximal effect of the drug or an effect related to improved neuron health and subsequent neuron- glia communication is unknown. Interventional trials with NSAIDs in AD have not been promising, and underline the need for compounds that reduce neuroinflammation through a mechanism other than COX inhibition. An early clinical trial of the NSAID indomethacin in mild-moderate AD demonstrated significantly less decline over a 6 month period, however, 50% of the non- responders dropped out of the study (Rogers et al., 1993). Later trials of indomethacin and triflusal, a platelet aggregation inhibitor with anti-inflammatory properties, showed small protective effects over a 12-13 month period that did not reach clinical significance due to small sample size (de Jong et al., 2008, Gómez-Isla et al., 2008). A recent clinical trial of the NSAID ibuprofen with the gastroprotectant esomeprazole given to patients with mild to moderate AD for one year showed decreased cognitive decline only in those individuals who carried the ApoE4 allele (Pasqualetti et al., 2009), which is in agreement with some epidemiological reports showing an interaction between NSAID use and the ApoE4 genotype (Fotuhi et al., 2008, Szekely et al., 2008a).

Some clinical trials of anti-inflammatories have shown no improvement and some poor side effects, including trials of aspirin (Bentham et al., 2008), the anti-malarial and anti-inflammatory hydroxychloroquine (Van Gool et al., 2001), the corticosteroid prednisone (Aisen et al., 2000) and the NSAIDs rofecoxib, naproxen, and diclofenac (Scharf et al., 1999, Aisen et al., 2003, Reines et al., 2004, Thal et al., 2005). In these trials drugs were tested for 4 years or less and in patients diagnosed with mild to moderated AD. The Alzheimer’s Disease Anti-inflammatory Prevention Trial (ADAPT) compared the efficacy of the NSAID naproxen and the selective COX2 inhibitor celecoxib to prevent onset of AD in a high risk group and showed that treated groups tended to have worse mental scores. The trial was terminated early because of concerns about cardiovascular risk due to celecoxib and showed significantly increased risk (~60%) of cardiovascular incident in patients treated with naproxen (Martin et al., 2002, Martin et al., 2006, ADAPT Research Group et al., 2007, Martin et al., 2008, Meinert et al., 2009). Although the ADAPT trials were terminated prematurely, further investigation into this cohort found that those who elected to use statins had a significantly reduced risk of AD (Sparks et al., 2008), and that NSAIDs are protective before the onset of AD, but harmful after the development of AD (Breitner et al., 2011) and the efficacy of NSAIDs may depend upon the rate of decline (Leoutsakos et al., 2012). It is possible that the ADAPT study showed poor outcome due to drug choice and because the cohort, advanced in age and high risk, likely had a well established inflammatory response years prior to the intervention with NSAIDs in these clinical trials, and duration of neuroinflammation is a critical factor determining disease onset (Bardou et al., 2013). Results from the epidemiological and interventional studies are consistent with the idea that early changes occur during a pre-clinical phase that lead to the development of AD and PD, and that interruption of certain processes during this period may prevent disease development, whereas targeting these same systems later in the disease state may be ineffective (Stewart et al., 1997, Zandi et al., 2002, Martin et al., 2008, Pasqualetti et al., 2009). Intervention during the pre-clinical stages of AD and PD is becoming more realistic as we develop improved screening techniques that utilize genetic tests, evaluation of inflammatory factors and pathological protein levels in serum and CSF, scanning for regional hypofunction, and PET imaging of the distribution of Aß and activated microglia

(Perrin et al., 2009). In the following section, I will provide evidence that glutamate is elevated in late-middle age and, paradoxically, decreased in aged individuals. I propose that this time-dependent course from a hyperglutamatergic to a hypoglutamatergic state explains the increased efficacy during early disease stages and relative loss of efficacy thereafter.

1.5.3 Glutamate dysregulation in AD Dysregulation of glutamatergic signaling is the likely mechanism by which neuroinflammation promotes cognitive impairment, and one of a collection of mechanisms by which neuroinflammation leads to neuronal cell death. Furthermore, the relative level of glutamatergic activity, i.e. network hyperexcitability vs. hypoexcitability, may influence the efficacy of anti-inflammatory interventions. Glutamate dysregulation is deleterious, and its consequences include poor cognitive processing, particularly in the hippocampus, metabolic failure and cell death. A prolonged hyperglutamatergic state may result in compensatory responses or the accumulation of cell death over the course of neurodegenerative disease, in either case. Therefore, hyperglutamatergic state may be followed by a homeostatic or hypoglutamatergic state. This may interact with anti- inflammatory treatment, and must be understood to approach pharmacological therapy targeting the glutamatergic system. This section will detail evidence that glutamate is dysregulated in AD.

1.5.3a Evidence of dysregulated glutamate in AD The important function of GLT1 to clear synaptic glutamate may be lost in AD, as GLT1 is reduced in AD brain tissue (Jacob et al., 2007), splice variants of GLT1 present in AD are associated with less glutamate clearance when assayed in vitro (Scott et al., 2011), and a detergent-insoluble form of GLT1 increases that correlates with degree of cognitive impairment (Woltjer et al. 2010). Interestingly, GLT1 is colocalized with tau pathology and amyloid plaques (Jacob et al. 2007; Sasaki et al. 2009; Thai 2002; Simpson et al. 2010), and while GLT1 is lower overall in AD brains compared to controls, the regions with the most severe pathology, particularly the inferior frontal and inferior temporal cortices, have the highest levels of astrogliosis and GLT1; potentially a compensatory response to excessive glutamate activity (Hynd et al. 2004b). Highlighting

52 the protective quality of GLT1, loss of GLT1 protection in the hippocampus is reflected by upregulated expression in the cerebellum, a region that is spared in AD (Jacob et al. 2007). These conclusions are supported by the observation that cognitive decline accelerated in an AD mouse model with genetic deletion of one GLT1 allele (Mookherjee et al. 2011). Glutamate clearance in AD fibroblasts is reduced to 40% of fibroblasts from controls, and this reduction is potentiated by exposure to excitotoxic stress at levels that do not change the dynamic of glutamate clearance in fibroblasts from controls (Begni et al. 2004). The decreased function of GLT1 in AD is accompanied by evidence from AD animal models that suggests changes in xCT (reviewed by Lewerenz et al. 2012), although there is a lack of clinical investigations associating cystine exchange with AD. Although GLT1 and its function are attenuated in AD, some AD brains also maintain normal levels of GLT expression (Beckstrøm et al. 1999). This may highlight the compensatory role of astrogliosis, underscore the need to look at splice variants, or simply suggest that glutamatergic dysregulation through loss of GLT1 is not necessary for symptomatic AD. There are many mechanisms by which glutamate may be dysregulated; likewise, there are many potential etiologies of AD. Neuroinflammation, in combination with oxidative stress and glutamatergic dysregulation, by loss of GLT1 or another mechanism, combined create an environment conductive to the development of AD. Reduced GLT1 function can cause neuronal degeneration. Knock-out of GLT1 on astrocytes by continuous i.c.v. administration of an antisense oligonucleotide over one week elevated extracellular glutamate from a mean of approximately 1 µM to 17 µM, a 32- fold increase, measured by microdialysis. This loss of GLT1 function also increased evidence of neurodegeneration and caused progressive paralysis (Rothstein et al. 1996). Prevention of the synthesis of GLT1 by an antisense antinucleotide in rats leads and excitotoxic neuronal damage (Rothstein et al. 1996). Reduced clearance of glutamate from the synapse by inhibition of GLT1 with DL-threo-β-benzyloxyaspartate (TBOA) or WAY- 855 leads to excitotoxic neuronal death (Selkirk et al. 2005; Bonde et al. 2005). These studies suggest that enhancing glutamate transport may be protective under pathological conditions. Like astrocytic regulation, interesting changes also occur in NMDARs in AD patients. A polymorphism in NR2B is the only genetic characteristic identified out of over

53 half a million genes investigated by ADNI that correlates with atrophied temporal lobe volume and poor MMSE performance (Stein et al. 2010). This genetic risk factor is present throughout the lifespan, and the predominant extrasynaptic position of NR2B is associated with excitotoxicity from excessive glutamate; together, these observations suggest that AD is strongly associated with glutamatergic hyperfunction. Yet, post-mortem AD brain tissue reveals that NR2B as well as NR2A NMDAR subunit protein expression is reduced in degenerating regions (Hynd et al. 2004a), and transcription of a particular NR1 subunit splice variant is also reduced in degenerating regions in AD patients (Hynd et al. 2004d). We have observed a similar decrease in NR1 protein expression in an animal model of chronic neuroinflammation, with no observations of cell death apart from decreased NR1 subunits (Susanna Rosi et al. 2004). Similarly, magnetic resonance spectroscopy studies find a decreased glutamate levels in AD patients (Fayed et al. 2011; Rupsingh et al. 2009; Lin et al. 2003; Antuono et al. 2001). These observations suggest the possibility that NMDAR subunits and glutamate decrease over time as a compensatory response, or that their expression is decreased due to loss of neurons. Either way, this establishes a scenario in which an early hyperexcitatory environment is followed by a hypoexcitatory environment (Table 1). The balance of glutamatergic function is further complicated by the observation that synaptic glutamatergic activity can be suppressed (i.e. loss of LTP and enhanced LTD) at the same time that network excitatory activity can elevate and synchronize to elicit seizures (Palop & Mucke 2010). Animal studies also report changes in glutamatergic regulation that tend toward decreased glutamate activity across age. TNFα-evoked glutamate release from astrocytes in hippocampal slices from APP-overexpressing transgenic mouse model of AD is unchanged at 4 mo. compared to age-matched controls, but decreased in 12 mo. mice (Rossi et al. 2005). Anti-inflammatory treatment use is preventative against the development of AD, but anti-inflammatory treatments generally fail during clinical trials. Aspirin and the NSAIDs ibuprofen and indomethacin not only reduce neuroinflammation, but also prevent astrocytic glutamate release by ~80% after application of TNFα. This suggests that their protection against AD may be attributed to modulation of glutamate, and that earlier timing may be beneficial by reducing glutamate release in a

54 hyperglutamatergic environment, whereas later use may depress already reduced glutamatergic synaptic activity. The possibility that glutamatergic activity changes with age is supported with evidence from Stephens et al. (2009) which demonstrated that KCl-evoked glutamate release was increased in the DG of middle-aged rats followed by late-age increase in DG glutamate clearance and a decrease in KCl-evoked glutamate release from CA3 measured in vivo with a glutamate-sensitive microelectrode array. A similar age-dependent decrease in KCl-stimulated glutamate release occurs in wild-type and AD transgenic mice (Minkeviciene et al. 2008). Similar results were found using the microelectrode array in transgenic mice that overexpress glutamate dehydrogenase 1 (GLUD1) and sustain a hyperglutamatergic state. These mice expressed increased basal levels of glutamate and elevated depolarization-induced glutamate, but reduced LTP (Bao et al. 2009) followed at later age by loss of dendritic spines, nerve terminals and neurons (Michaelis et al. 2011). The authors describe this model as creating a “lifelong, pulsatile, moderately excessive” increase in glutamate at neuronal synapses; similar to what may occur under chronic neuroinflammatory conditions. Indeed, this animal model is also associated with compensatory responses, including the elevation of glutamate receptors and GLT1. Furthermore, these authors found significant changes in expression level of groups of genes involved in neuroinflammation (including chemokines, oxidative stress pathways and TLRs) and recovery (such as neurotrophins and cell survival pathways) (Wang & Michaelis 2010; Michaelis et al. 2011); suggesting the presence of both M1 and M2a activated microglia. Importantly, the microglia activation state in this model is likely derived from proximal M1-initiating events, such as neuronal death, as well as concurrent M2a anti-inflammatory processes; a mixture of events that has been observed in AD (Colton et al. 2006).

Young Middle-Aged Aged Subject/ Reference Model ↓ NR1, NR2 AD patients (Hynd et al. 2004b; Hynd et al. 2004a) ↓ GLT1, AD patients (Scott et al. 2011; clearance Woltjer et al. 2010; Jacob et al. 2007; Begni et al. 2004) ↓ Glutamate AD patients (Fayed et al. 2011; Rupsingh et al. 2009; Lin et al. 2003; Antuono et al. 2001) ↑ KCl-evoked ↓ KCl-evoked Normal F- (Stephens et al. 2009) glutamate glutamate 344 rats release release ↑ clearance No change resting levels TNFα-evoked ↓↓ TNFα- APP AD (S. Rosi et al. 2005) astrocytic evoked mice glutamate astrocytic release and pre- glutamate synaptic release and pre- potentiation synaptic potentiation ↓↓KCl-evoked APP/PS AD (Minkeviciene et al. glutamate and wild- 2008) release, VGlut1 type mice ↑ basal Neuronal loss GLUD1 (Bao et al. 2009; glutamate, hypergluta Michaelis et al. 2011) induced matergic glutamate mice release ↓ LTP Table 1. Hyper- and Hypoglutamatergic states across aging and in AD

Like AD, PD may be associated with glutamatergic dysregulation, but there is little clinical evidence linking PD directly to a dysregulation to glutamate. Nevertheless, the dopaminergic neurons lost in PD are particularly susceptible to oxidative stress, and oxidative stress is associated with glutamate dysregulation. NMDARs are present on nearly all neurons in the CNS and may make these cells vulnerable to excitotoxicity (Hynd et al. 2004c). This includes the dopaminergic neurons of the SN that receive glutamatergic

56 input from the cerebral cortex and subthalamic nucleus (Figure 8). Activation of NMDARs in the SN leads to nNOS activation and may result in production of peroxynitrite (Beal, 1998). Dopaminergic cells are particularly vulnerable to oxidative stress, and the PD brain tends to produce less GSH to combat oxidative stress (Maher 2005). Joining these ideas, memantine is used to treat PD patients who have declining cognition or Lewy body dementia (Ballard et al. 2011; Olivares et al. 2011; Emre et al. 2010; Burn 2010).

1.5.3b The glutamatergic system as a pharmacological target NSAIDs have an impressive record of preventing AD and PD in epidemiological evaluations, but clinical trials mostly failed (see Sections 2.5.2c and 2.5.2d). It is possible, that prevention of AD by anti-inflammatories is due, in part, to their ability to reduce both neuroinflammation and astrocytic glutamate release. For example, aspirin and the NSAIDs ibuprofen and indomethacin prevent astrocytic glutamate release by ~80% after application of TNFα in vitro (Rossi et al. 2005). Similarly, caffeine reduces both neuroinflammation and glutamate release, and is protective against the development of AD (see Section 2.2.1). However, neither NSAID use nor caffeine consumption are efficacious treatments for AD. Glutamatergic signaling is a pharmacological target that might influence cognitive impairment directly. Currently memantine (Nemenda), a non-competitive NMDAR antagonist, is the only glutamatergic drug approved for use in Alzheimer’s disease memantine is able to block excessive NMDAR activity, but it’s fast on-off kinetics preserve normal synaptic activation of NMDARs (Rogawski & Gary L Wenk 2003; Parsons et al. 2007a). The efficacy of memantine underscores the influence of glutamatergic dysregulation in AD (Parsons et al. 2007a). No other glutamatergic drugs are approved for use in AD, but one, Levetiracetam, is currently in clinical trials. Levetiracetam (Keppra) is an anti-epileptic that suppresses synchronized network activity. Seizure activity is common, though not often recognized, in subsets of AD patients and in AD mouse models. Compared to other antiepileptics, Levetiracetam is associated with a good cognitive profile in patients (Cumbo & Ligori 2010; Lippa et al. 2010). In an AD mouse model, Levetiracetam reduced abnormal excitation, preserved hippocampal structure and function, and improved performance in memory tasks (Sanchez et al. 2012). In patients with mild cognitive impairment (MCI), Levetiracetam reduced hippocampal hyperactivity and improved cognitive performance 57

(Bakker et al. 2012; Belcastro et al. 2007). These data support the use of drugs that regulate glutamatergic activity in AD. Reduced glutamate transport may contribute to disease progression and pathology. Dysregulation of GLT1 occurs in several neurodegenerative disorders including amyotrophic lateral sclerosis (ALS, Sasaki et al. 2000; Shobha et al. 2007; Heath et al. 2002; Lin et al. 1998), Huntington's disease (Behrens et al., 2002; Beurrier et al., 2010; Faideau et al., 2010), and multiple sclerosis (Vallejo-Illarramendi et al., 2006). Interactions between glutamate transporters and neurodegenerative disease were thoroughly reviewed by Sheldon and Robinson (2007) and Beart and O’Shea (2007). Two drugs recently identified as able to increase glutamate clearance, Ceftriaxone and Riluzole, have potential therapeutic value in AD. Ceftriaxone (Rocephin) is a third generation ß-lactam antibiotic currently used in humans that increases GLT1 gene transcription, protein expression and function in human fetal astrocytes and an ALS mouse model (Rothstein et al. 2005). Ceftriaxone produces increased glutamate clearance in animal models (see Appendix E). Ceftriaxone reduced glutamate measured by microdialysis by 40-50%, and that reduction was sustained up to 20 days after the termination of treatment (Rasmussen et al., 2010). This suggests that ceftriaxone may also have long-lasting effects upon GLT1 in humans as well, thus allowing for intermittent treatment and avoidance of poor side-effects from continued antibiotic use. Ceftriaxone is protective against excitotoxicity and symptoms of neurodegeneration in animal models. Ceftriaxone elevated GLT1, reduced excitotoxicity and improved memory performance in a model of hypoxia and stroke (Chu et al., 2007; Hota et al., 2008; Verma et al., 2010). Ceftriaxone also produced positive outcomes in animal models of the degenerative diseases ALS (Guo et al., 2003) and Huntington’s (Sari et al. 2010; Miller et al. 2008). The antibiotic ceftriaxone is FDA approved, has been used in humans for two decades and is currently entering clinical trials for treatment in the degenerative disease ALS (phase III, Massechusetts General Hospital and National Institute of Neurological Disorders and Stroke). Augmentation of glutamate transport by ceftriaxone has been studied in models of neurodegenerative disease including ALS and Huntington’s (Guo et al., 2003; Miller et al., 2008; Rothstein et al., 2005; Sari et al., 2010). However, the relationship between

58 pharmacologically elevated glutamate transport and inflammation, an important component of neurodegenerative disease, or aging, a predominant risk factor for neurodegenerative disease, have not been explicitly examined. Ceftriaxone is currently used in humans as an antibiotic and, if it is effective in aged animals, it has potential to move relatively quickly from experimental research to clinical trials in MCI and AD. Like Ceftriaxone, Riluzole (Rilutek) increases glutamate clearance (Azbill et al. 2000; Mu et al. 2000a; Mu et al. 2000b; Izumi et al. 2002; Dunlop et al. 2003; Frizzo et al. 2004; Vidwans & Hewett 2004; Sameul et al. 1992), and may also decrease pre-synaptic glutamate release and reduce post-synaptic activity at NMDARs (see Appendix F). Riluzole decreases concentrations of extracellular glutamate and produces positive cognitive outcomes in animal models of AD and other inflammation-associated disorders (Malgouris et al. 1989; Abarca et al. 2000; Vorwerk et al. 2004; Banasr et al., 2010; Corderre et al., 2007; Chowdhury et al., 2008; Dzahini et al., 2010; Hassanzade et al., 2010; Hassanzade et al., 2011; Gourley et al., 2012). Riluzole is a therapeutic currently used for ALS, and a clinical trial is currently recruiting participants to test the use of Riluzole in AD. If it is efficacious, Riluzole could transition into use as an AD therapeutic quickly.

1.6 Chronic neuroinflammation model Chronic, low-level neuroinflammation, similar to that seen in AD and PD can be experimentally induced by the chronic infusion of LPS into the IVth ventricle of young rats. Chronic, infusion is more likely to be representative of the type of inflammation seen in the disease state than a single injection or microinfusion. Intracerebroventricular (i.c.v.) infusion is preferable to intraparenchymal infusion because it avoids tissue damage and compromise of the blood-brain barrier in the region of interest. Volume overload to the brain tissue is minimal using this procedure because the 0.25 μl/hr administered contributes only about 0.2% of the total cerebral spinal fluid (CSF) volume produced by the rat each hour and is only 0.12% of the rat’s total CSF volume. With this model, many regions are exposed to circulating lipid-soluble LPS, which highlights the propensity of some brain regions to develop inflammation over others. Increased neuroinflammation and behavioral deficit can also be produced by intraperitoneal (i.p.) injection of LPS, which may mimic a peripheral infection. Chronic intraventricular infusion does not induce a robust peripheral

59 immune response and allows the study of neuroinflammation that likely occurs in AD and PD in the absence of peripheral infection.

1.6.1 Mechanism by which lipopolysaccharide induces inflammation LPS is a component of the gram-negative bacteria cell wall and included in a group of highly conserved microbial components that activate the immune system called pathogen associated molecular patterns (PAMPs). LPS is recognized by a pattern recognition receptor (PRR) in the toll-like receptor family, toll-like receptor 4 (TLR4), which is present on microglia. LPS is composed by a lipid A anchor which is recognized by CD14, a component of the TLR4 receptor complex, and also a core and an oligosacharride chain which is recognized by TLR4. Once the TLR4 receptor complex is engaged, a signaling cascade of messengers such as IκB (inhibitor of κB), IRAK (IL-1 receptor-associated kinase), and MKK (mitogen activated protein kinase) can lead to the activation and translocation to the nucleus of transcription factors including NFκB, JNK (c-Jun N-terminal kinase), the MAPKs (mitogen activated protein kinases) ERK (extracellular signal-regulated kinases) and p38, and the IRF (interferon regulatory factor) family. These transcription factors regulate genes involved in immune response (i.e., inflammation, cytokine production), cell survival and cellular proliferation (Figure 11). Furthermore, activation of TLR4 by LPS drives a process which parallels those that occur in neurodegenerative disease. Polymorphisms of TLR4 and CD14 have been connected to AD risk (Minoretti et al., 2006, Balistreri et al., 2007, Balistreri et al., 2008, Rodriguez-Rodriguez et al., 2008), and a polymorphism of CD14 increases risk of PD in women (Lin et al., 2006). Necrotic hippocampal neurons can increase microglia production of glutaminase, which can contribute to increased extracellular glutamate, by a TLR- and MyD88- (myeloid differentiation primary response gene 88) dependent mechanism (Pais et al., 2008). MAPKs have been shown to be elevated with aging (Lynch and Lynch, 2002) and in the AD brain (Wood, 2003). Proteins (Aß-42), cytokines (IL-1, IL-6 and TNFα) and molecular signals and debris from dying and dead neurons such as mRNA, HSPs (heat shock proteins) and HMGB1 (high-mobility box group 1) known to be elevated in AD and PD have all been shown to activate TLR4 or other pathways that lead to activation of pro-inflammatory transcription factors (Lehnardt, 2010). For example, Aß activates TLR4 as well as RAGE and leads to the activation MAPK and transcription 60 factors such as NFκB, a pathway and outcome similar to stimulation of microglia by LPS (Greenberg et al., 1994, Yan et al., 1996, Pyo et al., 1998, Lue et al., 2001b, Liu et al., 2005, Vukic et al., 2009). Neuromelanin released from dying dopaminergic cells can also initiate the NFκB pathway (Wilms et al., 2003) and potential neuroinflammation.

Figure 11. Toll-like receptor 4 signaling pathway Schematic representing simplified toll-like receptor 4 (TLR4) signaling. The TLR4 receptor complex is activated by lipopolysaccharide (LPS), some cytokines that are downstream results of its activation and factors that are elevated in AD including amyloid ß42 (Aß42) and intracellular components released by cells during the neurodegenerative process including heat-shock protein (HSP) and mRNAs. Activation of the TLR4 receptor complex leads to a cascade of events that result in the activation of transcription factors (ISREs, NFκB and AP1) and the transcription of cytokines. 62

1.6.2 Chronic i.c.v. LPS reproduces components of AD and PD LPS activates the TLR4 pathway and initiates the production of factors that are known to be elevated in AD and PD. Chronic intraventricular infusion of LPS can reproduce molecular, cellular, and behavioral changes that share characteristics with AD and PD pathology (see Appendix D). The model of chronic intraventricular LPS administration was shown to reproduce some components of AD, including regional microglia activation in temporal lobe regions including the hippocampus and EC (Hauss-Wegrzyniak et al., 1998a, Hauss-Wegrzyniak et al., 1998b) as well as increased astrogliosis (Hauss-Wegrzyniak et al., 1998a, Rosenberger et al., 2004, Richardson et al., 2005). Activated microglia are first present near the site of infusion, later increased microglia activation is observed throughout the parenchyma and is not defined by proximity to the ventricles. After 6 days of chronic LPS infusion, activated microglia are found around the ventricles and pia but not in the brain parenchyma (Rosenberger et al., 2004). However, after 4 weeks infusion, activated microglia are present both near and distant to the ventricles, particularly in the EC and the CA3 and DG sub-regions of the hippocampus (Hauss-Wegrzyniak et al., 1998a, Hauss- Wegrzyniak et al., 1998b). LPS binds to the CD14 and TLR4 receptors which leads to the activation of MAPKs, transcription factors and eventually the production of some of the same pro- inflammatory mediators that are indicated in AD and PD. One week after a single injection of LPS into the lateral ventricle expression of CD14 mRNA and TLR4 receptor immunostaining is increased (Xia et al., 2006, Huang et al., 2008), implying that activation of the CD14/TLR4 complex by LPS may have self-amplifying effects. Activation of the CD14/TLR4 complex is followed by a cascade of events at least as early as one hour and as long as 2 weeks duration that includes increased activation of NFκB and IκBα which traditionally lead to the production of pro-inflammatory cytokines. ApoE4 genotype, a risk factor for AD, potentiates NFκB expression in mice after i.c.v. LPS injection (Ophir et al., 2005). These data illustrate a connection between a risk factor for AD and the elevation of pro-inflammatory cytokines. Cytokines which are elevated in AD and PD and have polymorphisms related to disease risk. TNFα, IL-1α, IL-1ß and IL-6, are elevated within as little as 15min of LPS

63 infusion and may remain elevated as long as 4 weeks (Sanna et al., 1995, Hauss- Wegrzyniak et al., 1998a). There is also an elevation in COX2 and prostaglandins that begins within 3 hours and lasts at least 7 days after a single injection (Albin et al., 1992, Zhang and Rivest, 2001). In another study, COX was not elevated between 6 hours and 6 days of chronic LPS infusion although prostaglandins were elevated during the same period (Rosenberger et al., 2004). This may be an artifact of dosing; those animals whom received LPS for 6 consecutive days received only 0.25µg spread throughout the entire day, whereas the single dose was 0.50µg once. It is possible that a single higher dose elicits a different inflammatory response. Infusion rate differentially effects the neuroinflammatory response to LPS infused into the pulmonary artery of pigs such that receiving the same dose of endotoxin more quickly resulted in higher levels of some inflammatory markers including TNFα and leukocyte number but not IL-6 (Lipcsey et al., 2008). A similar effect may be seen in the brain; if this is true, the timing and duration of LPS administration is critical. Oxidative stress may accompany the inflammatory state induced by i.c.v. LPS. Single i.c.v. injections of LPS have been shown to increase measures of oxidative stress (Zujovic et al., 2001), such as an in increase in lipid peroxidation (Montine et al., 2002) and an increase in iNOS in glia and SOD in neurons (Sugaya et al., 1998). Combined i.c.v. and i.p. LPS injections produced an increase in lipid peroxidation of neuronal membranes and reduced hippocampal dendritic branching, both of which were dependent upon NFκB and iNOS signaling (Milatovic et al., 2003). Oxidative stress may be one mechanism by which neurons die in AD and PD, and may be related to LPS-induced reduction of neuronal size and number. Cell shrinkage and possibly death may result from chronic neuroinflammation induced by i.c.v. LPS in regions where neuronal loss is observed in AD. Chronic LPS infusion into the IVth ventricle for 4 weeks leads to enlargement of the lateral ventricles and a diminution in size of the hippocampal formation (Hauss-Wegrzyniak et al., 2000a). The number of large neurons stained by cresyl violet and counted by stereology in layers II and III of the EC was diminished after 4 weeks of LPS infusion and indicates either shrinkage or loss of neurons (Hauss-Wegrzyniak et al., 2002). Whether neurons diminish in size or die, the functional outcome is likely to be loss of function.

Memory dysfunction related to that seen in AD is induced by chronic i.c.v. LPS infusion on both the electrophysiological level, i.e. LTP, and on the behavioral level. LTP, both NMDAR-dependent and NMDAR-independent, is reduced in the hippocampus of rats infused with LPS chronically for 4 weeks (Hauss-Wegrzyniak et al., 2002, Min et al., 2009). LPS-induced microglia activation is more extensive in CA3 compared to CA1 and correlates with the reduction of sparse encoding of Arc (activity-regulated cytoskeleton- associated protein), an immediate-early gene with function in learning and memory, as well as a reduction in the ability to accurately encode spatial information (Rosi et al., 2005b, Rosi et al., 2009). Spatial memory addressed by performance in the Morris water maze, a task dependent on hippocampal function, was impaired by as little as 3 weeks and as long as 10 weeks sustained i.c.v. LPS infusion (Hauss-Wegrzyniak et al., 1998a, Hauss-Wegrzyniak et al., 1999a, Hauss-Wegrzyniak et al., 2000b, Hauss-Wegrzyniak et al., 2000c, Marchalant et al., 2007, Min et al., 2009). Interestingly, LPS infused for 5 weeks followed by 5 weeks without endotoxin infusion resulted in a prolonged inflammatory response and memory impairment similar to that seen after 5 weeks infusion without a recovery period (Hauss- Wegrzyniak et al., 2000c). Memory deficits may reflect cell loss, although memory deficits have been seen after chronic neuroinflammation induced by intra-hippocampal injection of LPS without cell death (Tanaka et al., 2006). Drugs that indirectly prevent microglia activation, such as the NSAID NO-flurbiprofen, the partial NMDAR antagonist memantine and the endocannabinoid analog Win-55212-2 also prevent memory impairment resulting from chronic neuroinflammation (Hauss-Wegrzyniak et al., 1999a, Rosi et al., 2006, Marchalant et al., 2007, Rosi et al., 2009). These data support epidemiological evidence indicating that the reduction of neuroinflammation is protective against AD-like pathology and memory impairment. The following table in Appendix D: Summary of outcomes from i.c.v. LPS infusion summarizes the behavioral, cellular and molecular consequences of both i.c.v. injection and infusion of LPS described above. Within minutes of i.c.v. LPS injection, levels of pro- inflammatory cytokines are increased and remain so for a minimum of 4 weeks, and within 5 days COX2 is elevated. Oxidative stress, specifically lipid peroxidation of neuronal membranes, is observed within hours and persists for at least 3 days. Activated microglia

65 are visible in the vicinity of the ventricles and pia mater within 6 hours and increase in number throughout specific regions of the parenchyma, such as the CA3 and GD regions of the hippocampus, over a period of weeks. Within 4 weeks of chronic infusion of LPS, hippocampal volume and the number of large neurons in the EC are decreased while ventricular volume is increased. The functional outcomes of these LPS-induced changes within the hippocampus are impairments in glutamatergic signalling, LTP and spatial memory. Taken together, these data support the conclusion that inflammation is sufficient to induce AD-like changes in a variety of biochemical and behavioral indicators.

Chapter 2: Caffeine modulates microglia activation in the hippocampus of LPS- infused young rats

2.1 Brief Rationale: When I began as a graduate student in Dr. Wenk’s lab, I worked with a post-doc to investigate the use of a cannabinoid agonist to reduce pre-synaptic glutamate release and the consequent effects on neuroinflammation and memory performance in a model of experimentally-induced neuroinflammation and in natural aging. Based upon epidemiological evidence that caffeine consumption is protective against the development of neurodegenerative disease, evidence that caffeine reduces glutamate release, and presence of adenosine receptors on microglia, I designed my first set of experiments to investigate the role of caffeine in a model of neuroinflammation and in aging.

2.2 Introduction Caffeine may be useful in protecting against Alzheimer’s (AD) and Parkinson’s (PD) impairments related to chronic neuroinflammation. Epidemiological studies show that caffeine consumption reduces the risk of developing AD and PD, and animal models of AD and PD also demonstrate modest neuroprotection by caffeine. Caffeine may protect the brain by modulating microglia activation and by regulating glutamate release. In the current study, we evaluated whether caffeine administration is sufficient to prevent microglia activation and behavioral impairment induced by lipopolysaccharide (LPS).

2.2.1 Caffeine is protective against AD and PD Coffee consumption correlates with reduced prevalence of AD and PD in prospective and retrospective cross-sectional studies. Coffee consumption reduces the risk of AD (Lindsay et al. 2002; Maia & de Mendonça 2002; van Gelder et al. 2007; Hernán et al. 2002), and middle-aged (65-79 years) people who drink 3-5 cups of coffee per day have a 65% decreased risk of AD in late-life (Eskelinen et al. 2009). Similarly, coffee consumption reduces the risk of the developing PD (Ross et al. 2000; Ascherio et al. 2001; 67

Hu et al. 2007; Hernán et al. 2002; Hu et al. 2007), and those who drink 6 or more cups per day have a 25% reduced risk of developing PD (Powers et al. 2008). Furthermore, consumption of coffee combined with use of non-steroidal anti-inflammatories (NSAIDs) potentiates the protective effects against PD more than intake of either alone (Powers et al. 2008). Taken together, these data support the protective roles of both caffeine and the mediation of inflammation against the development of PD. The neuroprotective effect of coffee against AD and PD can be attributed primarily to the content of caffeine, an adenosine receptor antagonist (Ross et al. 2000; Ascherio et al. 2001). Adenosine is a purine that is used in cellular energy production. Adenosine is generally in the extracellular space at high levels only when there is cellular damage, therefore, the recognition of adenosine can trigger pro-inflammatory immune cascades. This conclusion made by epidemiological studies is substantiated by in vitro evidence and studies in animal models.

Caffeine and the selective adenosine A2A receptor (A2AR) antagonist SCH58261 protect cultured neurons and prevent memory impairment in animal models of AD that incorporate amyloid ß (Aß). Blocking the function of adenosine A2ARvia the antagonist SCH58261 or genetic inactivation attenuated mitochondrial dysfunction, synapse loss and memory impairment in a spontaneous alternation task 2 weeks after intracerebroventricular (i.c.v.) exposure to Aß (Canas et al. 2009). Similarly, deficits in inhibitory avoidance and spontaneous alteration in mice resulting from i.c.v. infusion of Aß were attenuated by treatment with SCH58261, as well as by administration of caffeine (30 mg/kg) in the drinking water or by daily intraperitoneal (i.p.) injection over 4 days and by (Dall’Igna et al. 2007). Treatment with caffeine water (0.3 mg/ml) also attenuates cognitive deficit it in a number of tasks and decreases levels of Aß in the brain of adult and aged transgenic mice that overproduce amyloid precursor protein (APP) (Arendash et al. 2006; Arendash et al. 2009; Cao et al. 2009). Most animals were pre-treated with caffeine or the caffeine was given concurrently with insult, and this is consistent with epidemiological evidence that consumption earlier in life, i.e. before the onset of disease, is protective.

Caffeine and antagonists of the A2AR (SCH58621, KW-6002, and CSC) also attenuate dopaminergic cell loss and improve motor performance in animal models of PD. Caffeine and CSC, directly protect vitality of rat mesencephalic neuronal cultures from 6- hydroxydopamine (6-OHDA) challenge (Nobre et al. 2010; Aguiar et al. 2006). Caffeine

68 and A2AR antagonists reduce the PD-like motor side effects (tremulous jaw movements, muscle rigidity and hypolocomotion) of anti-psychotic drugs and dopamine antagonists

(Trevitt et al. 2009; Wardas et al. 2001; Shiozaki et al. 1999). Likewise, caffeine, A2AR antagonists and genetic deletion of forebrain A2AR also protect against 1-methyl-4-phenyl- 1,2,3,6-tetrahydropyridine (MPTP)-induced inflammation in the SN and striatum, loss of SN dopaminergic neurons, reduction in striatal dopamine, striatal dopaminergic terminals and striatal dopamine transporter binding sites as well as hypolocomotion (Jiang-fan Chen et al. 2001; Shiozaki et al. 1999; Pierri et al. 2005). Furthermore, the adenosine agonist CGS21680 triggers catalepsy (Shiozaki et al. 1999). Like most studies of caffeine in AD models, caffeine in PD models is typically administered before or during the initial phases of pathology. This is consistent with the epidemiological evidence indicating that caffeine intake over a lifetime, prior to the development of disease symptoms, is protective. However, caffeine may not only protect against the onset of PD, but also improve symptoms when administered after the appearance of movement symptoms. Caffeine administered for 14 days after a unilateral intrastriatal injection of 6-OHDA attenuated the decrease of dopamine levels and apomorphine-induced rotations in a dose-dependent manner in rodents (Aguiar et al. 2006).

Similarly, the A2AR antagonist KW-6002 dose-dependently reduced motor deficits produced by MPTP in marmosets without causing stereotypy, hyperactivity or dyskinesia (Kanda et al. 1998). Marmosets injected with MPTP developed a PD-like motor impairment, and an antagonist of both A1R and A2AR (like caffeine) dose-dependently increased motor ability without significant side effects such as dyskinesias or dystonia (Mihara et al. 2008a; Mihara et al. 2008b). Moreover, caffeine administered to 16 patients with PD relieved the total akinesia aspect of freezing and of gait, but not trembling in place. These improvements lasted months, and a two-week caffeine withdrawal was sufficient to renew the clinical effect of caffeine (Kitagawa et al. 2007). Caffeine also has the potential to work synergistically with current PD treatments that augment dopamine, potentially lowering the dosage needed and decreasing dopamine- induced motor side effects. Striatal dopamine release increases in control and reserpine challenged rats after coadministration of L-3,4-dihydroxyphenylalanine (L-DOPA) with the A2AR antagonist CSC compared to L-DOPA alone (Golembiowska 2004). Adenosine

A2AR antagonists work synergistically with L-DOPA to improve motor deficits induced by 6-OHDA in mice and rats (Matsuya et al. 2007; Rose et al. 2007) and haloperidol-induced deficits in rats (Wardas et al. 2001). Similarly, subthreshold doses of L-DOPA and the

A2AR antagonist KW-6002 reduced catalepsy induced by either haloperidol or reserpine in mice (Shiozaki et al. 1999). Finally, the A2AR antagonist KW-6002 reduces PD-like motor symptoms and dyskinesias in rats and primates produced from L-DOPA or apomorphine treatment (Bibbiani et al. 2003). Antagonism of adenosine receptors is able to modify the outcome in AD and PD models, without contributing to the side-effect profile in PD, because it acts to reduce glutamate release and has anti-inflammatory properties (Ciruela et al. 2006; Deuchars et al. 2001; Malva et al. 2003; Martín & Buño 2003; Kalda et al. 2006; Rebola et al. 2011).

2.2.2 Caffeine mechanisms of action and protection Caffeine is a non-selective adenosine receptor antagonist, but acts primarily at the

A1 and A2A receptors (Sergi Ferré 2008). Adenosine receptors are G-protein coupled receptors expressed on neurons, microglia and astrocytes and the distribution of adenosine receptors varies across brain regions. Neurons express A1, A2A, A2B and A3 receptors.

Adenosine receptors A1 and A2A are expressed in the hippocampus in pyramidal neurons and glutamatergic nerve terminals (Rebola et al. 2005; Burnstock & Rev 2010; Cunha et al. 1996; Klausnitzer & Manahan-Vaughan 2008), as well as the basal ganglia (Svenningsson et al. 1999; Fredholm 2004; Fredholm & Svenningsson 2003; Mishina et al. 2007). Adenosine receptor subtypes A1 and A2A are located pre- and post-synaptically.

A1 and A2A receptors may interact and form heteromers, at which low levels of adenosine inhibit synaptic release through an A1R-driven mechanism and high concentrations of adenosine facilitate neurotransmitter release through A2AR-driven mechanism and suppression of A1R by A2AR (Ciruela et al. 2006; Quarta et al. 2004; Alloisio et al. 2004;

Halldner et al. 2004). Microglia express adenosine A1, A2A, and A3, but not A2B receptors and astrocytes express adenosine A2B but not A2A receptors (Fiebich et al. 1996; Daré et al. 2007; Hammarberg et al. 2003). The location of adenosine receptors on glutamatergic nerve terminals and immune cells contribute to its neuroprotective characteristics.

2.2.2a Caffeine reduces levels of extracellular glutamate Extracellular glutamate may contribute to cognitive impairment in AD, motor impairment in PD, and cell death. Glutamate is essential in the induction of long-term potentiation (LTP) and the formation of memory within the circuitry of the hippocampus (Rebola et al. 2008; Costenla et al. 2010; Fontinha et al. 2009; see Section 1.3.1), however, excessive glutamate can disrupt these processes. Glutamatergic signaling is likewise important for balanced function within the basal ganglia circuitry, but excessive glutamate can amplify the loss of dopaminergic signaling from the SN (see Figure 12). Furthermore, elevated glutamate may lead to excessive calcium entry through post-synaptic receptors, energetic failure, and excitotoxic cell death that detrimental to both regions and is potentiated in dopaminergic neurons that are particularly vulnerable to oxidative stress (see Section 1.4.1b).

Glutamate release can be modulated in both the hippocampus and striatum by A1 and A2A receptor interactions (O’Kane & Stone 1998; Ciruela et al. 2006; Hargus et al.

2009). The A2AR increases with age, while the A1R remains relatively constant, and skews the system toward facilitation of glutamatergic transmission (Rebola et al. 2003; Costenla et al. 2011). Adenosine is part of normal glutamatergic regulation, but adenosine is also released by neurons under stress, such as a neuroinflammatory environment or cell death, and these higher levels can elevate extracellular glutamate. Furthermore, elevated glutamate promotes a pro-inflammatory environment in the presence of an A2AR agonist

(Dai et al. 2010). By blocking the action of elevated adenosine at A2ARs, caffeine could inhibit glutamate release, and attenuate pathological increases in extracellular glutamate. Blockade of adenosine receptors can reduce levels of extracellular glutamate by attenuating presynaptic release of glutamate from neurons (Mendonça & Ribeiro 1993). Increased levels of extracellular glutamate can cause non-event-related activation of NMDARs, leading to increased Ca2+ entry and potentially excitotoxic cell death. Adenosine receptors are expressed on the pyramidal neurons of the hippocampus and glutamatergic terminals (Rebola et al. 2005), as well as within the striatum (Svenningsson et al. 1999; Ferré et al. 2001), locations that allow the potential modulation of glutamatergic release by caffeine within regions relevant to AD and PD. Therefore, we evaluated performance in tasks sensitive to damage in the hippocampus (spatial memory) and the

SNpc (gross motor behavior, balancing posture, and rearing) in a model on chronic neuroinflammation with and without caffeine treatment.

Figure 12. Potential targets for caffeine in the basal ganglia in Parkinson's Adenosine A1R and A2AR for heterodimers (yellow and green circles) potentially on any or all glutamatergic terminals throughout the basal ganglia. Low levels of adenosine may inhibit glutamate release through A1R, and high levels facilitate glutamate release + through A2AR. Caffeine could reduce glutamate release, contributing to both positive ( ) and negative (-) potential therapeutic outcomes. A2AR reduces the affinity of dopaminergic type 2 receptors (D2R) in the striatum and caffeine may allow those receptors to be more responsive to dopamine from the substantia nigra (SN). 72

2.2.2b Caffeine modulates immune activation Adenosine is important for cellular function, notably as part of adenosine triphosphate (ATP) and cyclic adenosine monophosphate (cAMP) that may stimulate microglia upon release from dead or dying neurons. Stimulation of adenosine receptors increases microglia activation, proliferation and the production of neuroinflammatory factors, while blockade of adenosine receptors generally has the opposite effect.

Antagonism of A2AR and activation of A1R, A2BR and A3R reduce neuroinflammation (Tsutsui et al. 2004; Rebola et al. 2011; Dai & Zhou 2011; Haskó et al. 2005; Koscsó et al. 2012), therefore, the anti-inflammatory effects of caffeine are likely due to antagonism of

A2ARs and suppression of A1R by A2AR in A1R-A2AR heteromers.

Activation of A1 and A2A receptors together induce microglia proliferation

(Gebicke-Haerter 1996). Systemic LPS induces A2AR upregulation and microglia contraction to amoeboid morphology, which can be reversed by administration of the A2AR antagonist SCH18261 (Orr et al. 2009). Adenosine receptor stimulation, particularly at

A2ARs, can lead to the production of COX2 mRNA and prostaglandin E2 synthesis by microglia (Fiebich et al. 1996) and caffeine is able to dose-dependently, albeit weakly, inhibit the synthesis of prostaglandin E2 in primary rat microglia culture stimulated by adenosine, adenosine analogues or LPS, perhaps by inhibiting the induction of COX2

(Fiebich et al. 2000). Stimulation of adenosine A2AR by a selective agonist potentiated nitric oxide (NO) release from microglia in mixed glia cultures, but not in cultures of microglia alone even though A2ARs were not found to be present on astrocytes (Saura et al. 2005); these results implicate a potential for caffeine to reduce NO production and attenuate the pro-inflammatory process Inflammatory factors are produced by microglia downstream of an intracellular cascade of events that includes the phosphorylation of mitogen-activated protein kinases

(MAPKs). The A2AR antagonist SCH58261 reduces the phosphorylation of the MAPK p38 in neuronal cultures exposed to Aß (Canas et al. 2009), prevents IL-1ß-induced phosphorylation of p38 as well as glutamate-induced post-synaptic Ca2+ entry and neuronal toxicity (Simões et al. 2012), and attenuates phosphorylation of p38 in vivo after cerebral ischemia (Melani et al. 2006). This mechanism may be one by which caffeine could directly reduce microglia activation and inflammation induced by LPS.

2.2.3 Hypothesis and approach The experiments described in this chapter were designed to address the following hypothesis: if the actions of adenosine at its receptors contribute to neuroinflammation and to the behavioral characteristics of AD and PD reproduced by LPS infusion or natural aging in rats, then the adenosine receptor antagonist caffeine should modulate the consequences of neuroinflammation and attenuate the disease-like characteristics produced by neuroinflammation and aging. To investigate the role that adenosine receptors play in the inflammatory process, we examined the potential of caffeine, an adenosine receptor antagonist, to prevent LPS- induced neuroinflammation, reduce preexisting neuroinflammation in aged rats and attenuate inflammation-related behavioral consequences. Caffeine reduces the risk of both AD and PD. If the mechanism by which caffeine is protective in both diseases is due to the ability to limit the consequences of neuroinflammation, then we should see less evidence of neuroinflammation and accompanying behavioral deficits in rats treated with caffeine. In order to investigate this, we administered a range of doses of caffeine by i.p. injection or in the drinking water to LPS-infused young rats during the development of neuroinflammation and to aged rats with naturally preexistent neuroinflammation. We predicted that caffeine would prevent LPS-induced neuroinflammation, reduce preexisting neuroinflammation in aged rats and attenuate inflammation-related behavioral consequences.

2.3 Methods

2.3.1 Subjects, surgical procedures and drug delivery Young (3 month) and aged (24 month) male F-344 rats (Harlan Sprague–Dawley, Indianapolis, IN) were singly housed in Plexiglas cages in a temperature-controlled room (22 ºC) on a 12/12 hr. light–dark cycle with lights off at 09:00 and given ad libitum access to food and water. All rats were given one week to acclimate to their new environment. Body weight, general health, movement and behavior were closely monitored throughout the study. Neuroinflammation was experimentally induced in young rats by the continuous infusion of LPS (0.25µg/hr) into the IVth ventricle (i.c.v.) over 2 or 4 weeks by an 74 indwelling cannula attached to an osmotic minipump (Alzet #2004, 0.25µl/hr). Controls were infused with artificial cerebral spinal fluid (aCSF; 140 mM NaCl, 3.0 mM KCl, 2.5 mM CaCl2, 1.0 mM MgCl2, and 1.2 mM Na2HPO4 adjusted to pH 7.4). Caffeine was administered to aged rats, and young rats infused with aCSF or LPS beginning the day after surgery, for either 2 or 4 weeks, and across a range of doses. Caffeine was given either in drinking water (0.1%) over four weeks to young rats, or by intraperitoneal (i.p.) injection (0.5-40mg/kg, 1 ml/kg) for either two or four weeks to aged and young rats. Controls received tap water or 0.9% saline injections, respectively. Experimental groups are outlines in Table 2. Caffeine was administered via the drinking water in order to model the typical method of consumption in humans and at a concentration of 0.1% to be consistent with previous studies (Popoli et al. 2000). The expected mean daily water consumption for a 250-350 g F-344 rat was 24-60 ml/day. Water bottles were weighed and changed every 3 days. Aged rats and LPS-infused rats that tend to consume less fluid each day; therefore, we used i.p. injections to assure that rats received a controlled dose of caffeine. Aged rats as well as young rats infused with LPS or aCSF for 2 weeks were injected daily for two weeks with either 0.9% saline or caffeine (40 mg/kg/d). Another subset of young rats was infused with aCSF or LPS for 4 weeks and injected i.p. with a range of doses of caffeine (0.5, 5, 10, 20 and 40 mg/kg/day). The maximal caffeine dose administered i.p., 40mg/kg/day, was chosen in order to reproduce the epidemiological findings for neuroprotection in humans. Most epidemiological studies indicate the consumption of 3- 5 cups of coffee per day, approximately 450 mg of caffeine, in order to see protection against the later development of AD or PD. To achieve an equivalent dose in a rat, caffeine was calculated as dose in mg/kg x [(0.3 kg rat)/(70 kg human)]0.33 (FDA, 2002). Injected caffeine solutions were prepared fresh daily and keep heated (35 ºC), as the highest concentrations of caffeine would precipitate at room temperature. The wake cycle of rats was observed, and caffeine injections were given approximately 2 hours after the beginning of the dark phase (their active phase).

Caffeine treated experimental groups Manipulation Caffeine Caffeine Caffeine Behavior Histology Biochemistry duration admin dose (n) (n) (n) aCSF 4w 4w Water Control 15 12 aCSF 4w 4w Water 0.1% 12 7 LPS 4w 4w Water Control 26 19 LPS 4w 4w Water 0.1% 16 8 Aged 2w i.p. 40 mg/kg 4 4 Aged 2w i.p. Control 4 5 aCSF 2w 2w i.p. Control 3 5 LPS 2w 2w i.p. 40 mg/kg 3 5 LPS 2w 2w i.p. Control 3 3 aCSF 4w 4w i.p. Control 5 aCSF 4w 4w i.p. 40 mg/kg 5 LPS 4w 4w i.p. Control 5 LPS 4w 4w i.p. 0.5 mg/kg 5 LPS 4w 4w i.p. 5 mg/kg 5 LPS 4w 4w i.p. 10 mg/kg 5 LPS 4w 4w i.p. 20 mg/kg 5 LPS 4w 4w i.p. 40 mg/kg 5 Table 2. Caffeine treated experimental groups

2.3.2 Behavior Young rats infused with aCSF or LPS for 4 weeks and administered 0.1% caffeine water or tap water were tested for spatial learning and memory in the Morris water maze task and observed for signs of hyperactivity in the open field. In the Morris water maze, rats were timed for the latency to find a hidden platform over 4 days, 6 trials per day and the total amount of time that they spent within the vicinity of the platform during a probe trial in which the platform was absent. Morris water maze method is further described in (Chapter 4, Section 5.3.2). In the open field, animals were observed for gross changes in motor behavior, particularly hyperactivity in caffeine treated rats. The open field method is described in Chapter 3 (Section 4.3.2a).

2.3.3 Histological procedures

2.3.3a Single peroxidase staining Rats were transcardially perfused with 4% paraformaldehyde and brains were sliced in coronal sections of 40 µm. Tissues were stained by standard avidin/biotin

76 peroxidase reaction with a primary antibody directed against the major histocompatibility complex II (MHCII) antigen (1:400, mouse monoclonal, Pharmigen, San Diego, CA, #22331D) in order to identify activated microglia. Primary antibody was followed by a biotinylated secondary anti-mouse antibody (1:200, Vector, Burlingame, CA, #BA-2001), and visualized with 0.05% diaminobenzidine (DAB; Vector, Burlingame, CA). Histological controls revealed no staining when primary antibody was absent. Tissues were rehydrated, counterstained with cresyl violet, dehydrated with serial dilutions of ethanol and coverslipped with Cytoseal (Allan Scientific, Kalamazoo, MI) mounting medium. Two representative slices of the anterior hippocampal region were examined using light microscopy. Density (cells per total area) of MHCII+ cells within the CA1, CA3 and DG subregions of the hippocampus and the corpus collosum (CC) was determined using MetaMorph imaging software (Universal Image Corporation, West Chester, PA).

2.3.3b Double-immunofluorescent staining Free floating sections were mounted on slides and air-dried, then processed as described previously (Rosi et al. 2005). Briefly, after washing in Tris-buffered saline

(TBS), the polyclonal rabbit anti-A1R (1:100, Sigma A268) was applied. After 24 hr. of incubation at 4 ºC, the sections were incubated for 2 hr. at room temperature with the secondary anti-rabbit biotinylated antibody (Vector, Burlingame, CA), followed by incubation with avidin/biotin amplification system (Vector, Burlingame, CA) for 45 min. The staining was visualized using the TSA fluorescence system CY3 (Perkin-Elmer Life Sciences, Emeryville, CA). After washing in TBS solution, the tissues were quenched, blocked again and incubated with primary anti-MHCII. Then, the biotinylated monoclonal secondary rat absorbed anti-mouse (Vector, Burlingame, CA, BA-2001) for 2 hr., the tissue was incubated with Avidin Biotin Blocking Kit (Vector, Burlingame, CA) for 30 min to block cross-reaction with the primary staining. Following treatment with the avidin/biotin amplification system (Vector, Burlingame, CA), staining was visualized with a TSA fluorescence system CY5 (Perkin-Elmer Life Sciences, Emeryville, CA) and the nuclei were counterstained with diamidino-2-phenylindole (DAPI; Molecular Probes, Eugene, OR). No staining was detected in the absence of the primary or secondary antibodies.

2.3.4 Biochemical procedures Rats were anesthetized and rapidly decapitated. Tissue from aged rats and young LPS- and aCSF-infused rats that were administered 40 mg/kg caffeine or 0.9% saline i.p. were sub-dissected and stored at -80 ºC. Tissues were homogenized and protein concentration of homogenates was determined by Bradford Assay. Western blots and a radio-binding assay were performed in order to determine the level of functional adenosine

A1R (A1R forms a heteromer with A2AR, and caffeine is an antagonist of both) and changes that may have occurred to the MAPKs, p-Erk and p-p38, with aging and upon LPS exposure.

2.3.4a Western Blots Samples and gels were prepared as described in (Chapter 3, 3.3.4a) and probed using phospho-ERK1/2 (1:1000, 43 kDA, Cell Signaling, 91015), phospho-p38 (1:500, 38 kDA, Cell Signaling, 91015) or adenosine A1R (1:2000, 36-40 kDA, Sigma A268) diluted in blocking buffer. Protein expression was visualized using chemiluminescence enhanced by SuperSignal West Pico Chemiluminescent Substrate (Thermo) and exposed to X-ray film (Amersham). Protein bands were quantified by densitometric analysis using Image J software (NIH). Protein load per lane was calculated based on total protein in the homogenate. Levels of p-Erk, p-p38 and A1R were normalized to constitutively expressed ß-actin (1:5000, 42kDa, mouse monoclonal, Clone AC-74, Sigma A5316) levels on the same. For each animal, samples were analyzed in triplicate, ratios were averaged, and then comparisons were made across experimental groups.

2.3.4b Binding Adenosine receptor activity was evaluated following protocols detailed by (Snell et al. 2000; Duarte et al. 2006). Briefly, total binding to the adenosine A1R was determined 3 by incubating samples with a radiolabeled selective A1R antagonist, [ H] 8-cyclopentyl- 1,3-dipropylxanthine (DPCPX; 1nM; Perkin Elmer, Net 974) for 30min in a 37 °C waterbath. Non-specific binding was measured in the presence of N6-cyclohexyl adenosine (CHA)

(10μM, Sigma C9901), a selective A1R agonist. The suspension was filtered under vacuum onto Whatman GF/B filters and placed into scintillation vials with 5 ml of scintillation fluid. Samples were stored at room temperature and radioactivity was measured the

78 following day using liquid scintillation spectrometry. All samples were run in triplicate. Specific binding was determined by subtracting non-specific binding from total binding. Values in decays per minute (DPM) were transformed to specific-binding in fmol/mg protein and averaged for each animal.

2.3.5 Statistics Data were analyzed by analysis of variance (ANOVA) followed by post hoc comparisons (Fisher LSD). Correlations were analyzed by Pearson Product method.

2.4 Results Caffeine water consumption was estimated by water bottle weight loss. Rats given 0.1% caffeine water consumed a mean of 82.6 mg/kg/day over 4 weeks, ranging from the lowest daily average 27.1 mg/kg to 128.08 mg/kg, and did not differ significantly between aCSF- and LPS-infused groups (t32 = -0.869, p = 0.391).

2.4.1 Behavior

2.4.1a Spatial learning and memory: Morris water maze Latency to locate the hidden platform in the water maze (Figure 13A) between rats infused with aCSF or LPS for 4 weeks and treated concurrently with 0.1% caffeine water was analyzed by 3-way repeated measures ANOVA, which revealed a main between- subjects effect of inflammation (aCSF vs. LPS; F1, 59 = 23.419, *p < 0.001), a main within- subjects effect of trial day (F3, 177 = 128.482, p < 0.001), an interaction between inflammation and trial day (F3, 177 = 13.878, p < 0.001), and an interaction between drug treatment (Sal vs. Caf) and trial day (F3, 177 = 2.934, p = 0.035). Paired t-tests indicate that all experimental groups improved from trial days 1 and 4 (p ≤ 0.010) in finding the platform in less total time. Swim speed (Figure 13B) was analyzed by 3-way repeated measures ANOVA, which revealed a main between-subjects effect of inflammation (F1, 59 = 0.884, p = 0.015) and a main within-subjects effect of trial day (F3, 177 = 37.401, p < 0.001). Thigmotaxis (Figure 13C) was analyzed by by 3-way repeated measures ANOVA, which revealed a main between-subjects effect of inflammation (F1, 59 = 10.909, *p =

0.002), a main within-subjects effect of trial day (F3, 177 = 86.418, p < 0.001), an interaction 79 between inflammation and trial day (F3, 177 = 8.702, p < 0.001), and an interaction between drug treatment and trial day (F3, 177 = 2.933, p = 0.035). Post hoc analysis indicates that LPS-infused rats spend more trial time in the pool perimeter than aCSF-infused rats (p ≤ 0.007), but that caffeine had no effect within aCSF or LPS conditions. A probe trial was conducted at the end of Day 4 (Figure 14) in which the platform was removed and the percentage time that rats spend within a 42 cm radius of the platform (25% of the water maze pool surface) was calculated within a one minute trial. Analysis by ANOVA revealed a main effect of LPS infusion (F1, 62 = 10.037, p = 0.002), but no main effect of caffeine treatment (F1, 62 = 3.40, p = 0.07). LPS-infused groups spent less time in the vicinity of the missing platform than did aCSF-infused groups (*p ≤ 0.03). Estimated mean daily caffeine consumption did not correlate with latency to find the hidden platform, swim speed or thigmotaxis on any of the trial days, nor with time spend in the vicinity of the absent platform during the probe trial in either aCSF- or LPS- infused rats (p > 0.05).

Figure 13. Water maze latency, swim speed and thigmotaxis Latency (A) swim speed (B) and thigmotaxis (C) were evaluated. LPS-infused rats performed took longer to find the hidden platform and spent more time in the pool perimeter that to aCSF infused rats (*p ≤ 0.007). 81

Figure 14. Water maze probe trial LPS-infused rats spent less time perseverating in the region of the missing platform than did their respective aCSF controls (*p ≤ 0.03).

2.4.1b Open field Distance traveled, mean speed, time spent moving, the frequency of grooming and the total time spent grooming were analyzed by ANOVA and yielded no main effects, suggesting that LPS did not induce hypolocomotion and that caffeine did not produce hyperactivity. The frequency of entrances between zones and the percentage of time spent within each zone of the open field was calculated as a measure of anxiety-like behavior and analyzed by ANOVA. There were main effects of zone entry frequency (F2, 230 = 156.10, p < 0.001) in which all animal groups entered the center zone least often (p < 0.001) and all groups except aCSF/Caf entered the mid-zone more frequently than the outer perimeter

(p ≤ 0.05). In addition, there was a main for duration of time spent in each zone (F2, 230 = 1803.40, p < 0.001) in which all experimental groups spent the most time in the mid-zone and the least in the center. Frequency of both fecal boli and urination were counted. There was an interaction in which the frequency of boli was increased in aCSF controls compared to aCSF/Caf (F1, 76 = 5.49, p = 0.002), and an interaction in which urination (F1, 76 = 7.61, p = 0.007) was increased in aCSF controls compared to both aCSF/Caf (p = 0.016) and LPS controls (p < 0.001).

There was a main effect of LPS upon both frequency of rearing (F1, 76 = 6.82, p =

0.011) and duration of rearing (F1, 76 = 9.97, p = 0.002) such that LPS-infused animals reared the least frequently and spent the least total time rearing (p ≤ 0.05).

Figure 15. LPS reduces rearing LPS-infused rats reared least frequently (A, p ≤ 0.05) and spent the least total time rearing (B, p ≤ 0.05).

Estimated average caffeine consumption over 4 weeks in LPS-infused rats did not correlate with the frequency between or amount of time spent within the perimeter, mid- zone or center of the open field arena, nor with average distance moved, average speed, time spent moving, rearing frequency or duration, grooming frequency or duration, number of fecal boli or frequency of urination (p > 0.05). Increased caffeine consumption in aCSF- infused rats correlated with spending more time in the outer perimeter of the arena (p = 0.009) and increased time spent grooming (p = 0.045). These animals spent less time in the mid-zone (p = 0.008), made fewer entrances into the mid-zone and outer perimeter (p < 0.021), traveled less distance (p = 0.001), spent less total time moving (p < 0.002), and moved with less speed (p < 0.002).

2.4.2 Caffeine reduces LPS-induced microglia activation in the DG MHCII+ microglia were detected immunohistochemically (Figure 16) and evaluated in three hippocampal subregions, CA, CA3 and DG (Figure 17) in rats infused with aCSF or LPS for 4 weeks and treated with 0.1% caffeine water. An ANOVA reveals significant main effects of experimental group (F3, 138 = 20.57, p < 0.001) and hippocampal subregion (F2, 138 = 14.49, p < 0.001) and an interaction between the two (F6, 138 = 4.92, p < 0.001). Both LPS-infused groups had a greater number of MHCII+ microglia in the CA3 and DG compared to CA1 (p < 0.001) and more MHCII+ microglia in the DG and CA3 than did aCSF-infused rats (*p < 0.001). There were no effects of caffeine water treatment, and estimated daily caffeine water consumption over 4 weeks did not correlate with the number of MHCII+ microglia in LPS-infused animals in any of the three subregions evaluated (p = 0.197)

Figure 16. Distribution of hippocampal MHCII+ microglia in rats treated with caffeine water 85

Figure 17. Quantification of hippocampal MHCII+ microglia in rats treated with caffeine water The number of MHCII+ microglia was elevated in LPS infused rats compared to aCSF infused in the DG and CA3 (*p < 0.001) and was not changed by 0.1% caffeine water treatment.

Naturally aged rats and young rats infused with LPS for 2 weeks were administered daily injections of caffeine (40 mg/kg/day) over a 2 week period. MHCII+ microglia detected immunohistochemically (Figure 18) and were analyzed in hippocampal subregions and within the corpus collosum by two-way ANOVA (Figure 19). There were main effects of region (F3, 90 = 43.52, p < 0.001) and experimental group (F4, 90 = 21.52, p < + 0.001) and an interaction (F12, 90 = 21.69, p < 0.001). Expression of MHCII microglia differed significantly across regions, overall, CC > DG > CA3 > CA1 (p < 0.001) with the exception of DG and CA3 which were not significantly different from each other. No regions differed significantly within aCSF controls or LPS/Caf. Within LPS controls, the DG expressed more MHCII+ microglia than any other region (‡p ≤ 0.025) and the CA3 expressed more than CA1 (†p = 0.002). Within both aged groups, the number of MHCII+ microglia was greater in CC compared to all other regions (Өp < 0.001) and in CA3 compared to CA1 (†p ≤ 0.022). Aged controls also expressed more MHCII+ microglia in

CA3 than DG (‡p = 0.010). LPS controls had more MHCII+ microglia in the DG than aCSF controls (*p < 0.001) and Aged controls (p < 0.001). LPS/Caf had fewer MHCII+ microglia in the DG than did LPS controls (§p < 0.001). Within the CA3 region, both LPS and Aged controls expressed greater MHCII+ microglia than aCSF controls (*p < 0.008). There are no significant differences between groups in the number of MHCII+ microglia in the CA1 region (p > 0.05). There was no correlation between the number of MHCII+ microglia in DG, CA3 or CA1 and latency to find the platform during the water maze task on Day 4 or with time spent perseverating in the vicinity of the missing platform during the probe trial (p > 0.05).

Figure 18. Distribution of hippocampal MHCII+ microglia of Aged and LPS-infused rats treated with caffeine (40 mg/kg/day i.p.)

Figure 19. Quantification of hippocampal MHCII+ microglia in Aged and LPS-infused rats treated with caffeine (40 mg/kg/day i.p.) Within Aged rats, the number of MHCII+ microglia was greater in CC compared to all other experimental groups and hippocampal regions (Өp < 0.001). Both Aged groups had more MHCII+ microglia in CA3 than CA1 (†p ≤ 0.022) and Aged controls also expressed more MHCII+ microglia in CA3 than DG (‡p = 0.010). Within LPS controls, MHCII+ microglia expression is higher in DG > CA3 > CA1 (‡†p ≤ 0.025). LPS/Caf had fewer MHCII+ microglia in the DG than did LPS controls (§p < 0.001).

MHCII+ microglia in the hippocampus of rats infused with either aCSF or LPS for 4 weeks and treated daily with 0.9% saline or caffeine (0.5-40 mg/kg/day i.p.) (Figure 20) were quantified (Figure 21) and analyzed by ANOVA. There were main effects of hippocampal subregion (F2, 112 = 84.88, p < 0.001) and experimental group (F7, 112 = 17.23, p < 0.001) and an interaction between them (F14, 112 = 4.01, p < 0.001). There was no significant difference between regions within aCSF-infused animals. Within all LPS- infused groups, DG and CA3 expressed more MHCII+ microglia than did CA1 (*p < 0.001). All LPS-infused groups expressed more MHCII+ microglia than aCSF-infused groups within the DG and CA3 subregions (*p < 0.001). Within the DG, LPS + 20 expressed fewer MHCII+ microglia than LPS controls and LPS + 10 (§p ≤ 0.037). 89

Figure 20. Distribution of MHCII+ microglia in the hippocampus of LPS-infused rats treated with caffeine (0.5-40 mg/kg/day i.p.)

Figure 21. Quantification of MHCII+ microglia in the hippocampus of LPS-infused rats treated with caffeine (0.5-40 mg/kg/day i.p.) MHCII+ microglia cell number varied significantly between hippocampal subregions of LPS-infused rats such that DG/CA3 > CA1 (*p < 0.001). Within the DG, LPS + 20 expressed fewer MHCII+ microglia than LPS controls and LPS + 10 (§p ≤ 0.037). 90

2.4.3 LPS infusion increased A1R binding + MHCII microglia and A1R were stained by double-immuno fluorescence (Figure

22). A1R appear to be on neurons and we did not observe colocalization of A1Rs and microglia, although microglia have been previously shown to express A1Rs (Fiebich et al. 1996).

Expression of the A1R protein levels (Figure 23A) were evaluated by Western blot.

An ANOVA, and yielded no significant differences between groups (F4, 23 = 0.60, p =

0.665). Evaluation of radioactive ligand binding of A1R (Figure 23B) revealed a significant main effect of experimental group (F4, 21 = 5.07, *p = 0.007) in which more functional A1R bound radioactive ligand in LPS controls than aCSF controls. This suggests that either more A1R is present after LPS infusion.

+ Figure 22. Adenosine receptor A1 and MHCII microglia in the CA3 A1R (red) is expressed by neurons in the hippocampus, including the CA3 region shown above. MHCII+ microglia (green) are dispersed throughout the CA3 region in LPS- infused and aged rats. The presence of A1Rs colocalized with MHCII on microglia (yellow) is not observed.

Figure 23. Adenosine receptor A1 protein expression and functional binding No changes in the level of A1R protein (A) in the hippocampus were observed, however, radioactive A1R ligand binding (B) was elevated in LPS controls compared to aCSF controls (*p = 0.007).

2.4.4 LPS elevates p-Erk and caffeine elevates p-38 Protein levels of the phosphorylated MAPKs Erk I/II and p38 (Figure 24) were measured and analyzed by ANOVA. There was a main effect of experimental group on p-

Erk expression (F4, 22 = 4.25, p = 0.013) wherein LPS-infused rats expressed higher levels of p-Erk protein than aCSF-infused rats and Aged controls (*p ≤ 0.016). There was also a main effect of experimental group upon p-p38 protein levels (F4, 22 = 3.69, p = 0.023) such that LPS/Caf and Aged/Caf expressed higher levels of p-p38 than aCSF and Aged controls (§p ≤ 0.035).

Figure 24. Protein levels of the phosphorylated MAPKs Erk and p38 Protein levels of p-Erk (A) increased in the hippocampus of rats infused with LPS for 2 weeks compared to aCSF and Aged controls (p ≤ 0.01*). Caffeine (40 mg/kg/day i.p.) treatment elevated phospho-p38 (B) protein in LPS and Aged groups compared to groups that were not treated with caffeine (§p ≤ 0.035).

2.5 Discussion

2.5.1 Caffeine reduces LPS-induced microglia activation Caffeine attenuated the number of LPS-induced MHCII+ microglia within the DG after treatment of 40 mg/kg/day i.p over 2 weeks and after 20 mg/kg/day i.p. over 4 weeks i.p., but not at 40 mg/kg/day i.p. over 4 weeks. Rats infused with LPS for 2 weeks were only treated with a dose of 40 mg/kg/day, therefore, we do not know if a lower dose would

94 also have attenuated microglia activation. It is possible that antagonism at the adenosine receptors causes a change in pharmacodynamics over time. If this were true, 40 mg/kg/day may be effective in rats infused with LPS for a period of 2 weeks, but that dose may induce tolerance or receptor internalization (Saura et al. 1998) that causes the response to the higher dose of caffeine to be ineffective over a 4 week period; i.e. a classic inverted-U dose-response effect. We have previously shown that A2BR expression on neurons is decreased after 4 weeks of continuous LPS infusion (Rosi et al. 2003). Although in the current study we do not see a reduction in A1R protein and observe an increase in A1R binding after 2 weeks of LPS infusion. Nevertheless, it is possible that 20 mg/kg/day caffeine is an optimal dose over a 4 week period; high enough to attenuate microglia activation but low enough to remain effective. We estimated that a dose of 40 mg/kg in a rat was roughly equivalent to 3-5 cups of coffee per day, a dose that is epidemiologically protective. However, if the metabolic rate of a rat is corrected for, 20 mg/kg is equivalent to approximately 4-6 cups of coffee and 40 mg/kg is the equivalent of 6-9 cups (Fredholm et al. 1999). Drinking 3 cups of coffee per day provided the optimal protection against cognitive decline, 4 cups was protective but less so, and drinking 5 or more cups per day was not protective in a large prospective study of the FINE cohort (Van Gelder et al. 2007). Another study showed that the highest caffeine consumption was the most protective in men, whereas moderate intake showed the most benefit to women; an inverted U-shaped response curve (Ascherio et al. 2001; Ascherio et al. 2004). Taken together, these data support the idea that a dose of 20 mg/kg/day may be an optimal dose in rats. Caffeine administered through the drinking water neither attenuated the LPS- induced increase in MHCII+ microglia nor improved water maze performance. LPS- infused groups consumed a mean of 85.5 mg/kg/day of caffeine (ranging 27.2-128.1 mg/kg/day), and variation in their consumption did not correlate with number of MHCII+ microglia or water maze performance. Although 0.1% caffeine water solution has been shown to reduce AD- and PD-like characteristics in animal models, it is possible that the dose was not appropriate to modulate the robust inflammation induced by i.c.v. LPS. It is also possible that this dose was too high, and did not attenuate microglia activation for the same reason that 40 mg/kg/day i.p. was not effective after 4 weeks of LPS infusion. An

95 alternative explanation is that because this dose was consumed over a 24-hr period instead of a single daily bolus, the resulting peak blood level was too low and therefore ineffective. Together, these confirm that caffeine reduces microglia activation in a part of the brain that is particularly vulnerable to LPS-induced neuroinflammation, and highlight the importance of optimal dosing.

2.5.2 A1 receptor expression and binding Changes in pharmacodynamics, due to prolonged exposure to the antagonist caffeine may influence the way in which caffeine interacts with its receptors and modulates inflammation over time. We observed strong A1R expression on neurons in the hippocampus, consistent with the distribution described by Ochiishi et al. (1999). Down- regulation and desensitization of A1Rs occurs after hypoxia and seizure activity in rats (Mendonça & Ribeiro 2000; Rebola et al. 2003). We have previously demonstrated that

A2BR expression on neurons decreases after 4 weeks of continuous LPS infusion (Rosi et al. 2003), consistent with loss of its anti-inflammatory influence (Vazquez et al. 2008; Koscsó et al. 2012; Rosi et al. 2003). However, the current results show that the protein levels of the A1R do not change after 2 weeks of exposure to daily i.p. injections of 40 mg/kg caffeine, nor do they change with normal aging or 2 weeks of continuous LPS exposure. This is consistent with the observation that levels of A1R and A2ARs do not change in the hippocampus or cortex of AD-transgenic mice after 5.5 months of caffeine administration through the drinking water (Arendash et al. 2006). Furthermore, our results show that A1R binding activity is increased in young animals infused with LPS for 2 weeks, suggesting that A1R levels increased after LPS exposure. If low-level adenosine is present after LPS infusion, then this would be consistent with anti-inflammatory action and inhibited glutamatergic function, but if high levels of adenosine are present after 2 weeks of continuous LPS infusion, then and increase in A1R may be associated with facilitation of glutamatergic signaling. After 4 weeks of continuous LPS infusion we see a hippocampal-sensitive memory impairment, loss of LTP and evidence of increased post- synaptic Ca2+ entry that is prevented with the non-competitive NMDAR antagonist memantine (Rosi et al. 2009; Rosi et al. 2006), suggesting that glutamatergic signaling is elevated and that the increase in A1R in young LPS-infused rats is associated with facilitating glutamatergic transmission. 96

2.5.3 LPS-induces increased protein levels of p-Erk and caffeine increases p-p38 LPS binds the CD14/TLR4 complex on microglia and phosphorylation of MAPKs, kinases which are part of a cascade leading to transcription of pro-inflammatory factors, may be elevated. We examined the phosphorylation of two MAPKs, Erk and p38,in aged rats and after 2 weeks of LPS infusion with concurrent administration of caffeine i.p. 40 mg/kg/day. Phospho-Erk was elevated after LPS infusion, and this may indicate a pathway by which chronic LPS infusion leads to an inflammatory state. If caffeine attenuates microglia activation by LPS, then we would expect to see a reduction of phospho-Erk in LPS-infused animals that received caffeine treatment; however, this was not the case. Protein levels of these two MAPKs were measured in whole hippocampus. It is possible that that caffeine treatment at 40 mg/kg/day for 2 weeks did reduce phosphorylation of Erk in the DG region of the hippocampus where it effectively reduced microglia MHCII expression, but that analysis of the entire hippocampus masked this change. It is also possible that caffeine reduces microglia activation through a mechanism that does not change the response of MAPK production induced by LPS exposure. Our data conflict with a study in which increased phosphorylation of Erk was seen after administration of the A2AR antagonist MSX-3 in a transgenic mouse model of dopamine deficiency, however increased phosphorylation of Erk may be a property of selective A2AR antagonism or an effect that occurs due to the loss of dopaminergic neurons which possess adenosine receptors (Botsakis et al. 2010). Unlike phospho-Erk, levels of phospho-p38 were only increased in caffeine treated rats, both LPS-infused and aged rats. The A2AR antagonist, SCH58261, reduced the activation of p38 in response to IL-1ß in neuronal culture (Simões et al. 2012). These results do not necessarily conflict with those found in the current study because we expect that the most robust phosphorylation of p38 occurs in microglia due to LPS-exposure and the results found in a primary neuronal culture are due to direct action of the adenosine antagonist caffeine upon neurons only. In other studies, caffeine protection has been associated with both reduced phosphorylation of p38 (Rebola et al. 2011; Yadav et al. 2012) and phosphorylation of p38 (Koscsó et al. 2012; Leshem-Lev et al. 2010).

2.5.4 Conclusions Caffeine attenuated the number of activated microglia in the DG of LPS-infused young rats over a period of 2 to 4 weeks. The results suggest that caffeine can modulate neuroinflammation, and this may be a mechanism by which caffeine offers neuroprotection seen in humans against AD and PD. Microglia activation within the DG induced by chronic LPS infusion has been previously shown to impair Morris water maze performance, and compounds that reduce microglia activation also reduce the spatial memory deficit. Taken together, these data imply that a dose and administration method of caffeine that is able to prevent microglia activation may also restore spatial memory. Epidemiological studies that show greater protection when caffeine is consumed early in life but not later near disease onset (Kandinov et al. 2007), suggest that caffeine might be more protective if animals are pre-treated before the infusion of LPS begins. This is not necessary, however, as we have shown that certain doses of caffeine initiated with the beginning of LPS infusion can prevent microglia activation. If there is an interaction between the response of adenosine receptors with the actions of caffeine over time, early exposure may prime the system so that caffeine is more efficacious. As we progress in identifying signs of AD and PD in earlier, pre-clinical stages (Perrin et al. 2009) caffeine may become a pharmacological tool to postpone symptom onset.

Chapter 3: Chronic neuroinflammation drives time-dependent changes that recover in the midbrain and brainstem

3.1 Brief Rationale: My previous work investigated the effects of neuroinflammation induced by 4 weeks of LPS on memory and included manipulation by pharmacological intervention. In this set of experiments (Chapters 3-5), I halved and doubled the duration of LPS infusion to investigate changes over time, and expanded my study to include an investigation of Parkinson’s disease (PD)- related changes in the midbrain and brainstem regions. The experiments presented in Chapters 3.5 were part of a large study, therefore, some of the methodology overlaps. This chapter investigates changes in the mid-brain and brainstem in the presence of chronic neuroinflammation, Chapter 4 is a pilot study testing delayed memantine treatment, and Chapter 5 focuses on the hippocampus and compensations to the presence of chronic neuroinflammation within the glutamatergic system.

3.2 Introduction Chronic neuroinflammation is involved in the onset and progression of neurodegenerative diseases such as PD (Hirsch et al., 2003; Tansey and Goldberg, 2010; Glass et al., 2010; Wahner et al., 2007). Pre-symptomatic and symptomatic PD are characterized by increased numbers of activated microglia in the SN (Bartels and Leenders, 2007; Gerhard et al., 2003, 2006; Iannaccone et al., 2012; Imamura et al., 2003; Ouchi et al., 2005;Ouchi et al., 2009), a region that later degenerates severely, and this indicates that neuroinflammation is an important early component of the disease process. In addition to the midbrain SN, the locus coeruleus (LC) and raphe are brainstem nuclei that also have increased microglia activation (Bertrand et al., 1997), degenerate in PD (D'Amato et al., 1987, Del Tredici & Braak, 2012; Bertrand et al., 1997; McMillan et al., 2011) and are related to symptom severity (Buchman et al., 2012). Aberrant extracellular proteins like α-synuclein activate microglia directly and both initiate and sustain a neuroinflammatory

99 response that potentiates SN neurodegeneration, which contributes to further accumulation of α-synuclein and the appearance of Lewy bodies (Gao et al., 2011; Zhang et al., 2011; Zhang et al., 2007; Zhang et al., 2005). Dopaminergic neurons are particularly vulnerable to the neuroinflammatory environment, and their death further stimulates microglia activation and neurotoxicity (Levesque et al., 2010; Gao et al., 2003). Prevention of microglia activation protects cultured midbrain neurons from MPP+- induced toxicity which targets neurons directly and indirectly activates microglia through neuronal death (Zhang et al., 2012; Ossola et al., 2011; Qian et al., 2011; Gao et al., 2011; Zhang et al., 2010; Wu et al., 2009; Yang et al., 2008; Chen et al., 2007; Qian et al., 2006), highlighting the role of neuroinflammation in progressive neurodegeneration. Similarly, epidemiological evidence shows that preventing neuroinflammation reduces the likelihood of developing PD. Neuroinflammation driven by an acute, peripheral immune challenge is sufficient to cause PD-like SN neurodegeneration in an animal model. One single peripheral injection of LPS in rats can initiate a central inflammatory response characterized after 10 months by sustained elevation of pro-inflammatory cytokines and selective loss of approximately 70% of dopaminergic neurons (Qin et al., 2007). Dopaminergic neurodegeneration was dependent upon presence of the tumor necrosis factor α receptor (TNFαR), demonstrating that continued upregulation of TNFα is a necessary component of SN pathology driven by neuroinflammation. This study demonstrates that the consequences of a strong peripheral immune response are gradual, progressive, long- lasting, self-propagating and damaging to the particularly vulnerable SN. PD could be triggered by an acute peripheral immune challenge, and can be exacerbated by an acute peripheral inflammation such as surgery. However, the inflammation that is generally associated with neurodegenerative diseases like PD is low- level, chronic, and thought to evolve from within the brain in response to pathology such as α-synuclein and components of dead neurons. It was our goal to reproduce a low-level and chronic neuroinflammatory environment similar to that characteristic of the diseased brain. We achieved this by continuous infusion of the gram-negative bacteria cell wall component lipopolysaccharide (LPS) into the IVth ventricle of adult male rats for a period of 2, 4 or 8 weeks. LPS directly activates an innate immune response by engagement with

100 the TLR4 receptor on microglia, similarly, microglia are activated by α-synuclein in a TLR4-dependent manner (Fellner et al., 2012; Zhang et al., 2005). Therefore, we believe that our model of slow, chronic LPS infusion is similar to the neuroinflammatory process in the PD brain. One of the strongest attributes of this inflammation model is that no brain region is targeted directly; regions in which activated microglia accumulate are self-selected and changes indicate vulnerability that distinguishes these regions from others. We have previously investigated time-dependent changes in other brain regions following LPS infusion, particularly the hippocampus, and documented corresponding loss of LTP and cognitive impairment (Hauss-Wegrzyniak et al., 1998; Wenk et al., 2000). In this study, we aim to characterize the response of midbrain and brainstem regions to low-level, continuous neuroinflammation and evaluate behavioral consequences as they relate to PD. We found the low-level, chronic neuroinflammation drives early changes in midbrain and brainstem, notably, transient declines in the expression of precursors to dopamine and norepinephrine, indicating that low-level neuroinflammation is sufficient to cause disturbances in neurotransmitter systems without significant cell loss.

3.3 Methods

3.3.1 Subjects, surgical procedures and LPS administration Young (4 month) male F-344 rats (Harlan Sprague–Dawley, Indianapolis, IN) were singly housed with ad libitum access to food and water in a temperature-controlled room (22 ºC) on a 12/12-h reverse light–dark cycle (lights off at 09:00). All rats were allowed 1 week to acclimate to their new environment. Body weights, general health, movement and behavior were closely monitored throughout the study. Artificial cerebrospinal fluid (aCSF; 140 mM NaCl, 3.0 mM KCl, 2.5 mM CaCl2, 1.0 mM MgCl2, and 1.2 mM Na2HPO4 adjusted to pH 7.4) or LPS (0.25µg/hr., 1.66¯¯ mg/ml dissolved into aCSF; E. coli serotype 055:B5, Sigma, St. Louis, MO, USA) were delivered by intracerebroventricular (i.c.v.) infusion continuously over 2, 4 or 8 weeks, creating six experimental groups (Table 3). A cannula was implanted into the IVth ventricle (-2.5 AP and -7.0 DV relative to lambda) and attached (via Tygon tubing, 0.06 O.D.) to an osmotic minipump (Alzet model #2006, 0.15µl/hr.; Durect Corp., Cupertino, CA, USA) as 101 previously described (Hauss-Wegrzyniak et al., 1998a). Calculations using the average fill volume of pump model #2006 and flow rate allow for release over an 8 week period. Post‐ operative care included lidocaine 1% solution and Neosporin (a topical anesthetic ointment) applied to the scalp as analgesics, and 2 ml of isotonic saline injected subcutaneously to prevent dehydration after surgery. All rats were sacrificed after 2, 4 or 8 weeks of i.c.v. infusion of aCSF or LPS.

Manipulation Duration Behavior (n) Histology (n) Biochemistry (n) 2w 10 5 5 aCSF 4w 23 12 11 8w 14 7 7 2w 16 8 8 LPS 4w 33 20 13 8w 18 10 8 Table 3. Animal groups Table shows distribution of animals across experimental groups and post-mortem analysis.

Week: 1 2 3 4 5 6 7 8 Surgery, 8w Behavior begin 4w Behavior infusion 2w Table 4. Schedule of infusion and behavioral testing

3.3.2 Behavioral testing In order to relate chronic neuroinflammation to motor deficits relevant to PD, we used the open field test to estimate gross motor deficit and stress-related behavior, and implemented a hanging task that requires forelimb strength and balance. PD-like motor symptoms (i.e. trembling, rigidity and slowness of movement) are displayed when approximately 80% or more of the cells in the SN are lost. We did not predict that our animals would cross this threshold and suffer PD-like motor deficits. We also acknowledge that cerebellar damage may contribute to poor performance. Behavioral testing was initiated 1 week prior to sacrifice, during the 4th or 8th week of aCSF or LPS

102 infusion (Table 4). No behavioral testing was conducted on animals with i.c.v. infusion of 2 weeks because rats were beginning post-surgical weight rebound and were likely still suffering the proximal effects of LPS, such as drowsiness. Each animal was acclimated to handling and to the experimenters on 3 days for 5 min each day during the week prior to testing. Prior to the hanging task and open field, rats completed 4 days of water maze testing described in Chapter 5 (5.3.2). All behavioral testing took place in the same room outside of the colony, under similar dim lighting conditions.

3.3.2a Open field Open field analysis was conducted to assess possible impairments in gross motor function and anxiety-related behavior. The open field was the drained and cleaned water maze pool with soiled litter from the colony sprinkled across the pool floor to obfuscate lingering odor from rats in previous trials and as an incentive to explore. The open field was divided into 3 zones: the outside perimeter (15.6 cm from the wall and 33.3% of open field arena), the mid-zone (20.3 cm wide and 33.3% of open field arena), and the center (49.05 cm radius, 33.3% of open field arena). Animals were placed into the tank at a position 20 cm from the wall, for a 15 min trial. Motor behaviors such as distance traveled and speed were calculated by EthoVision software (Noldus, Leesburg, VA) while rearing and grooming (both bouts and duration) were manually recorded using the EthoVision program. In addition, anxiety-related behaviors, frequency with which animals made entrances into each zone, the percentage of total trial time spent in each zone, freezing and grooming were analyzed.

3.3.2b Hanging Task In order to measure motor deficits that may be indicative of Parkinson-like symptoms in rats that are bilaterally impaired, we developed a novel task that requires animals to sustain balance in a perched position. A strong wire inside of rubber tubing was stretched and anchored taut across the length of the open frame of a plastic bin (80 cm depth x 30 cm wide x 40 cm long). Rats were placed on the wire with approximately 16 cm on both sides, 10 cm ahead of and behind them, and both hind legs flanking the forepaws. Latency to fall was recorded within a 60 sec trial. Two practice trials in which rats were allowed to use their tails along the edge of the bin for balance and support were

103 followed by 3 test trials. Trials were conducted for each rat directly after open field assessment, and the 5 total trials were consecutive with no rest periods.

3.3.3 Histological procedures Consequences of neuroinflammation were characterized in the SN, striatum, LC, hippocampus, raphe and cerebellum. Activated microglia, defined by expression of major histocompatibility complex II (MHCII), were quantified as an indication of the level of inflammation. Integrity of neurons in the SN and LC were investigated by staining tyrosine hydroxylase (TH), a rate-limiting enzyme in the production of dopamine and norepinephrine, and phosphorylated TH (pTH). In parallel, tryptophan hydroxylase (TrypH), an enzyme in the production of serotonin was evaluated in the raphe. Expression of TH in the striatum was investigated as an indicator of dopaminergic projections from the SN. Dopamine-ß-hydroxylase (DBH), an enzyme that produces norepinephrine from dopamine, was evaluated in this hippocampus were investigated as a reporter of projection strength. After behavioral testing was completed, rats were deeply anesthetized with isoflurane and transcardial perfusion (10 ml/min) was initiated with cold saline (0.9%, 80ml) containing heparin (1 U/ml), followed by 4% paraformaldehyde (dissolved in 0.1M phosphate buffer, pH 7.4, 120 ml). The brains were removed and post-fixed in 4% paraformaldehyde for 2-3 days at 4 ºC, transferred to 0.1M phosphate buffered saline (pH 7.4) and stored at 4 ºC. Serial coronal sections (40 µm) of regions of interest (Figure 25) were obtained using a vibratome (Leica Biosystems, Buffalo Grove, IL) and stored at -20 ºC in anti-freeze solution (0.5M phosphate buffer with glycol ethylene and glycerol).

104

Figure 25. Regions of interest Schematic representing the IVth ventricle as the cannula entry point and relative location of brain regions that were examined in this study including the SN (green), the LC (orange) and the dorsal raphe nuclei (red) as well as their projections to the striatum (green) and dorsal hippocampus (blue).

Tissues were stained with standard avidin-biotin peroxidase or fluorescence labeling methods. All reactions take place on an agitator and at 22 ºC unless otherwise noted and all rinses (3 x 10 min) are in PBS, PBS with 0.05% tween 20 (PBSt), TBS or TBS with 0.05% tween 20 (TBSt). Tissue were rinsed, quenched of native peroxidase activity with 0.3% H2O2 in 50% methanol for 1 hr., rinsed, blocked for non-specific binding with 5% NGS for 1 hr., and incubated in primary antibody diluted in 5% NGS over 1 night or 2 nights (TH, TrypH) at 4 ºC. Antibodies used include: anti-MHCII (1:400, mouse

105 monoclonal, Pharmingen, San Diego, CA, #554926), anti-TH (1:750, rabbit polyclonal, Millipore, Billerica, MA, #AB152), anti-pTH (1:500, rabbit polyclonal, Chemicon #AB152), anti-DBH (1:2000, rabbit polyclonal, AbCam, #ab43868) and anti-TrypH, (1:100, sheep, Millipore, #AB1541). Primary antibodies were identified with biotinylated secondary IgG (H+L) antibodies against the host species in which the primary was raised (1:200, Vector, Burlingame, CA, #BA-1000, BA-2001). No staining was detected in the absence of the primary or secondary antibodies. Confidence was also established for antibodies that produced a band of the correct molecular weight in Western blot analysis. Thereafter, peroxidase staining continued with incubation in a corresponding biotinylated secondary antibody for 1.5 hr., rinse, incubation for 1 hr. with avidin- biotinylated horseradish peroxidase (ABC kit, Vector, Burlingame, CA), rinse, and visualization by incubation with 0.05% 3,3-diaminobenzidine tetrahydrochloride (DAB, Vector, Burlingame, CA) or SG Blue (Vector, Burlingame, CA) as chromogen. Double peroxidase staining continued from this step. Sections were placed into another primary antibody and processed as described above. After a final rinse, sections were mounted onto gel coated slides and air-dried. Selected tissues were rehydrated and counterstained with cresyl violet. All peroxidase slides were then dehydrated with serial dilutions of ethanol and coverslipped with Cytoseal (Allan Scientific, Kalamazoo, MI) mounting medium. Fluorescence (described in Chapter 2 Section 2.3.3b) was used to quantify LC TH expression. Within the LC, TH chromogenic staining was so dense, and fluorescence intensity so high (causing bleed-through), that it prohibited the counting of MHCII+ microglia within the nucleus. Tissues were examined using light microscopy or florescent microscopy (as previously described in Chapter 2 Section 2.3.3). Stereological analysis of TH+ and MHCII+ cells in the SNpc was conducted on a group composed of 5 slices spaced every 6th slice across the SNpc, the counting frame spanned approximately 125,000 µm3 and sampled 30% of the region of interest. MHCII+ microglia were expressed on tissue from LPS groups but almost absent in the SN of aCSF infused animals, therefore, the parameters used to estimate TH+ cells in the SNpc by stereology were not appropriate for MHCII+ cells. Quantification of brainstem MHCII+ cells was carried out by measuring the density of staining per area with Image J (NIH, Bethesda, MD) on a region medial to each LC and

106 extending from the LC ventrally to the raphe. Estimates of total MHCII+ cell density in brainstem areas and cerebellum were also made between 2 blinded raters on a scale of 0- 6. Staining above threshold intensity was used to evaluate pTH in 5 evenly spaced SN slices, TH in the striatum and LC, and DBH in the hippocampus.

Figure 26. Regions of interest evaluated by histological analysis The SNpc (b), LC and raphe nuclei (c) are depicted. The raphe is located caudally to the section shown but included here for simplicity. The blue square indicated the brainstem region that was selected for densitometric measure of MHCII. DBH was evaluated in the CA1, CA3 and DG subregions of the hippocampus (a).

3.3.4 Biochemical procedures After behavioral testing was completed, rats were anesthetized with isoflurane and rapidly decapitated. Tissue was sub-dissected over ice, immediately transferred to dry ice, and stored at -80 ºC. The SN sends dopaminergic projections to the striatum, and the integrity and function of these projections were investigated. We used Western blot to measure TH expression as an indicator of dopamine production, and SNAP25, a synaptic

107 docking protein, as an indicator of synaptic stripping. Dopamine transporter (DAT) function was measured through radio-binding assay as an indicator of the function of the dopaminergic projections in the striatum. Calcineurin, a calcium-binding protein, and glial cell line-derived neurotrophic factor (GDNF) were evaluated in the brainstem by enzyme- linked immunosorbent assay (ELISA).

3.3.4a Western blot analysis The Western blot method follows that described in (Brothers et al. 2010) and Chapter 2 (Section 2.3.4a). Tissues were homogenized in lysis buffer (50 mM Tris-HCl, 50 mM NaCl, 10 mM EGTA, 5 mM EDTA, 2mM NaPP, 4 mM PNFF, 1 mM Na3VO4, 1 mM PMSF, 3.0% aprotinin, 2.0% leupeptin, and 0.1% SDS) and protein concentration was determined by Bradford Assay using BioRad protein assay buffer (BioRad, 500-0006). Samples of homogenates were prepared in loading buffer (0.5 M Tris-HCl, 25% glycerol, 10% SDS, and 0.5% bromophenol blue) and boiled for 5 min. Sample volumes containing 40 μg of protein were loaded onto a 10% SDS-PAGE gel and resolved by standard electrophoresis. Proteins were then transferred by electrophoresis onto nitrocellulose membrane (Hybond-C extra; Amersham) at 4°C with a constant 30 V and 0.9 A overnight. Membranes were blocked with blocking buffer (TBS with 5% dry milk and 5% sodium azide) and probed with primary antibodies diluted in blocking buffer. Primary antibodies used include: anti-ß-Actin (1:5000, 42kDa, mouse monoclonal, Clone AC-74, Sigma A5316), anti-TH (1:750, rabbit polyclonal, Millipore, Billerica, MA) and anti-SNAP25 (1:2000, 25kDa, rabbit polyclonal, AbCam ab5666). Confidence was established for antibodies that produced a band of the estimated molecular weight, and in some cases, for antibodies that produced a histological result that met expectations for distribution across tissue, location within tissue structures or morphology of cell type. After washing with PBSt, the membranes were incubated with horseradish peroxidase-conjugated secondary antibodies (1:100,000; Thermo) directed against the host species of the primary antibody and diluted in PBSt. Protein expression was visualized using chemiluminescence enhanced by SuperSignal West Pico Chemiluminescent Substrate (Thermo) and exposed to X-ray film (Amersham, Pittsburgh, PA). Membranes were stripped with Stripping Buffer (Thermo) and subsequently re-probed. The order of antibody application was monitored in effort to re-probe each membrane with antibodies 108 requiring different secondary antibodies or that produce bands at a different molecular weight. When membranes were re-probed with an anti-body to a protein with a similar molecular weight and the same secondary antibody as an antibody previously tested, caution was taken to first incubate the membrane in secondary antibody and develop with exposure of up to one hour to eliminate the possibility of the appearance of a residual band. These precautions allow us to be sure that the bands analyzed reflect the protein levels of the antibody being evaluated. Protein bands were quantified by densitometric analysis using Image J software (NIH). Protein load per lane was controlled for by evaluating protein expression in relation to levels of the constitutively expressed actin on the same membrane by creating a ratio in which actin was the denominator. For each animal, samples were ran in triplicate and averaged, and then comparisons were made across experimental groups.

3.3.4b DAT Binding The neuronal DAT is the primary mechanism through which dopamine is cleared from the synapse and taken into the pre-synaptic terminal. DAT binding was investigated following the methods described by Page et al., 2000 and similar to those described in Chapter 2 (Section 2.3.4b). Briefly, striatum from one hemisphere was homogenized and centrifuged. Binding to the DAT was determined in samples by incubating with a radiolabeled inhibitor of dopamine and norepinephrine transporters, [3H]GBR-12935 (20 ηM, Perkin Elmer, 35.5 Ci/mmol) at room temperature for 30 min. Non-specific binding was measured in the presence of nomifensine (100 μM, Sigma N1530) a dopamine reuptake inhibitor. The samples were vacuumed onto filters (Whatman GF/B) and placed into scintillation vials with 5ml of scintillation fluid. Samples were stored at room temperature and radioactivity was measured the following day using a liquid scintillation spectrometry reader (Packard Liquid Scintillation Analyzer 1900TR). Specific binding was determined by subtracting non-specific binding from total binding. Values in DPM were transformed to specific-binding in fmol/mg protein and averaged across the triplicate for each animal.

109

3.3.4c ELISA Brainstem (including the midbrain and pons regions) levels of calcineurin and GDNF were determined by ELISA in accordance with the manufacturer's instruction (Promega, Madison, WI). Samples were homogenized in lysis buffer (20 mM Tris, pH7.5, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% v/v Triton X-100, 2.5 mM sodium pyrophosphate, 1 mM B-glycerolphosphate, 1 mM Na3VO4, 1 µg/mL leupeptin, 1 mM PMSF) and centrifuged; after desalting to remove free phosphates, supernatants were diluted (50mM Tris pH 7.5, 1 mM DTT, 100 µM EDTA, 100 µM EGTA, 0.2% NP-40) and incubated (30 °C, 30 min) in reaction buffer. Okadaic acid (1 µM) was added to inhibit phosphatases PP1 and PP2A activity. A PKA regulatory subunit type II (DLDVPIPRFDRRV-pSer-VAAE) was used as substrate for the assay. Calcineurin activity was expressed in ηmol Pi released/mg protein and GDNF in pg/mg protein, relative to sample protein levels determined by Bradford assay.

3.3.4d Bradford protein analysis Protein content in homogenates used for Western Blot, DAT binding and ELISA were evaluated by Bradford analysis.

3.3.5 Statistics SigmaStat software (SysStat, San Jose, CA) was used to compare groups by ANOVA with Fisher LSD as the preferred post-hoc, and to perform Pearson correlation coefficients. Graphs are shown with SEMs represented by error bars. Control aCSF is shown in some graphs as one collapsed group, but aCSF groups were not collapsed for statistical analysis.

3.4 Results

3.4.1 Behavior

3.4.1a Open field analysis of gross motor and anxiety-related behaviors Gross motor performance, rearing and anxiety-related behaviors were monitored in the open field. Gross motor behaviors: distance traveled and percentage of total time spent moving, were not different between groups. Rearing is sensitive to the loss of dopamine neurons (Shi et al., 2004, de Meira Santos Lima et al., 2006). ANOVA revealed a main 110 effect of LPS treatment on rearing frequency (F1, 75 = 4.6, p = 0.035) and total trial time spent rearing (F1, 75 = 6.2, p = 0.015), such that LPS-infused animals reared less often and for less total time than aCSF groups (Figure 27A, B). Rearing deficits correlated with inflammation in the cerebellum (r = -0.526, n = 19, p = 0.020), brainstem (r = -0.732, n = 19, p < 0.001) and SNpc (r = -.381, n = 36, p = 0.021), in addition to TH+ cell number in the SNpc (r = -0.347, n = 36, p = 0.038). Time spent in the perimeter, passively avoiding the center can be interpreted as anxiety-like behavior (Figure 28). Frequency of entrances was analyzed by ANOVA and there was a main effect for experimental group (F3, 227 = 9.5, p < 0.001) and zone (F2, 227 = 150.4, p < 0.001). All groups made the least amount of entrances into the center of the arena (p < 0.001) and both aCSF 4w and LPS 4w made more entrances into the mid-zone than the outer perimeter (p ≤ 0.036). The mid-zone was also entered more frequently by LPS 4w than aCSF 4w (*p = 0.048). The mid-zone and outside perimeter were entered more often by aCSF 4w than aCSF 8w (p = 0.011) and entered more frequently by LPS 4w than LPS 8w (†p ≤ 0.009). All groups spent a greater percentage of trial time in the mid- zone than the outer perimeter, and within the outer perimeter than the center (p < 0.001) and there was no statistically significant difference between groups (F3, 227 = 1905.2, p = 1.000) in the percentage of time spent in each zone. There were also no significant differences between groups in other measures that relate to anxiety: freezing behavior, grooming frequency or time spent grooming.

3.4.1b Hanging task evaluation of balance and forelimb strength Latency to fall was averaged over the last 3 of 5 trials for each animal (Figure 27C).

There was a main effect of experimental group (F3, 66 = 27.0, p < 0.001). Both LPS 4w and LPS 8w fell more quickly than aCSF controls (*p < 0.001) and LPS 8w was more impaired than LPS 4w (p < 0.001). These data indicate that LPS infusion impairs performance on this task in a duration-dependent manner. Deficit in hanging performance correlates with inflammation measured in the cerebellum (r = -0.620, n = 19, p = 0.004) and SNpc (r = - 0.416, n = 32, p = 0.017), but not TH cell number in the SNpc (r = 0.060, n = 32, p = 0.743).

111

Figure 27. Open field rearing and hanging task results Both rearing and the hanging task require strength and balance. The frequency of rearing was not different between groups. LPS 4w spent less total time rearing than aCSF 4w (*p ≤ 0.035). In the hanging task, both LPS 4w and LPS 8w fell sooner than their aCSF controls (*p < 0.001).

112

Figure 28. Open field zone preference results LPS 4w made more entrances (A) into the mid-zone than aCSF 4w (*p = 0.048) and more entrances into the mid-zone and perimeter than LPS 8w (†p ≤ 0.009). All groups spent more time (B) in the perimeter and mid-zone than center and there were no differences in time spent in each zone between groups.

3.4.2 SN and projections to the striatum

3.4.2a Microglia activation and transient TH loss in the substantia nigra TH+ neurons and MHCII+ microglia were identified immunohistochemically in the SNpc (Figure 29) and their numbers were estimated by stereology (Figure 30). ANOVA performed on the estimated number of TH+ cells per area predicted by the stereological counting revealed a main effect of experimental group (F5,54 = 2.9, p = 0.025). The estimated number of TH+ cells was reduced in LPS 2w compared to aCSF 2w (*p = 0.023) 113 and both LPS 4w and LPS 8w (†p ≤ 0.009). We looked within the 2 week infusion groups, and found that the reduction in the number of TH+ cells in LPS 2w corresponded to a reduction in the density of staining of pTH after LPS infusion (*p = 0.025, Figure 31).

There was also a main effect of treatment (F1, 54 = 4.9, *p = 0.031) in which number of MHCII+ microglia was increased in LPS 4w and LPS 8w. There were no significant differences between aCSF groups and LPS 2w. There was no significant inverse correlation between the number of MHCII+ microglia and the estimated number of TH+ cells (p < 0.05). We also counted by stereology MHCII+ cells that took on a rounded, amoeboid morphology that appeared more like macrophages. There were few of these cells counted within the ROI (range 0-8, compared to 0-307 for total MHCII+ cells), making the optical fractionator method of stereology unsuitable to provide an appropriate estimate of total cell number (causing an overestimate). That withstanding, no statistically significant differences were driven by LPS infusion, but the number of counted MHCII+ cells with this morphology did, however, correlate negatively with expression of TH in the SNpc (r = -0.288, n = 47, p = 0.049).

114

Figure 29. Distribution of TH+ neurons and MHCII+ microglia in the SN TH+ neurons (blue) and MHCII+ microglia (brown), 10X left, 40X right. There is a transient reduction in TH+ neurons after chronic LPS infusion for 2 weeks. 115

Figure 30. Quantification of TH+ neurons and MHCII+ microglia in the SNpc The estimated number of TH+ cells in the SNpc (A) was reduced in LPS 2w compared to aCSF 2w (*p = 0.023) and both LPS 4w and LPS 8w (†p ≤ 0.009). Estimated MHCII+ microglia (C) were increased significantly in the SNpc of LPS 4w and LPS 8w (*p = 0.031).

116

Figure 31. pTH distribution in SN after 2 weeks infusion LPS infusion over 2 weeks reduced pTH staining density (*p = 0.025).

3.4.2b Integrity of SNpc projections to the striatum Dopaminergic projections from the SNpc innervate the striatum. Therefore, knowing that TH+ cell number in the SNpc is reduced after 2 weeks of LPS infusion, and in order to evaluate the integrity of these projections, we evaluated the expression of TH protein in the striatum as an indicator of dopamine levels and SNAP25, a protein that docks synaptic vesicles, as a marker of synaptic integrity. No changes in striatal TH were observed by either immunohistochemistry or Western blot analysis (p > 0.05), and the same

117 holds true for Western blot analysis of SNAP25 (Figure 32A-C). Function of the DAT was also evaluated by radioligand binding in the striatum and was analyzed as an indicator of the function of the dopaminergic projections from the SNpc (Figure 32D). These data taken together indicate that while TH+ cell number is reduced in the SNpc after 2 weeks of LPS infusion, the projections from the SNpc to the striatum may be intact and operational after 2 weeks, as well as 4 and 8 weeks of LPS infusion.

Figure 32. Striatum TH, SNAP25 and DAT binding do not change Protein levels of TH (A) and SNAP25 (B) detected by Western blot, TH detected by immunohistochemical staining (C) and dopamine transporter (DAT) binding in the striatum (D) do not change with LPS infusion.

118

3.4.3 Brainstem: LC and raphe nuclei Both the LC and the raphe are known to degenerate in PD, and may, in fact, precede degeneration of other areas such as the SNpc. In order to evaluate the effects of chronic LPS infusion on these brainstem nuclei, the LC was stained for the expression of TH, an enzyme that produces norepinephrine, and the raphe with TrypH, an enzyme that produces serotonin. Qualitative analysis of the structure of cells in these nuclei show that they are dysmorphic. Neurons in the LC and raphe of LPS-infused animals tend to look dystrophic, have more bends in their projections and localize closely with MHCII+ microglia (Figure 33. Distribution of MHCII+ microglia and neurons in the LC and raphe nucleih1-h2). Morphological changes seen in these nuclei do not seem to rebound with time although TH expression in the LC, like the SN, was reduced in LPS 2w compared to aCSF 2w (*p < 0.001) and is restored by 4 weeks of LPS infusion (Figure 34A, B). Activated microglia were additionally stained by double-peroxidase staining technique in order to see the way in which neurons in these nuclei and microglia in an activated state interact. TH staining within the LC was so dense (peroxidase) and so intense (immunofluorescent) that it prohibited the reliable counting of microglia within the nucleus. Therefore, an area between the LC of each hemisphere and ventral toward the raphe of each hemisphere was selected on three brainstem slices per animal and MHCII+ microglia were quantified by densitometry and expressed as density per area (Figure 34C).

ANOVA reveals a main effect for treatment (F1,27 = 6.0, p = 0.023) in which LPS-infused animals express more activated microglia per area, and LPS 4w express significantly fewer MHCII+ microglia than LPS 2w (†p = 0.028). Neurons and MHCII+ microglia appear to interact, and in some cases, MHCII+ microglia appear to wrap around neurons and their processes (Figure 33. Distribution of MHCII+ microglia and neurons in the LC and raphe nucleih1-h2). Noradrenergic projections from the LC were evaluated by examining DBH in hippocampus (Figure 35). There were main effects of experimental group (F5, 473 = 12.71, p < 0.001), hippocampal region (F2, 473 = 35.93, p < 0.001) and an interaction between the two (F10, 473 = 3.32, p < 0.001). Within experimental groups, there is less DBH staining in LPS 2w than aCSF 2w (p < 0.001), while LPS 4w and LPS 8w show greater staining density than their aCSF counterparts (p < 0.001) and LPS 8w > LPS 4w > LPS 2w (p ≤

119

0.029). LPS 2w has less DBH+ than aCSF 2w in both CA3 and DG (*p ≤ 0.007), while both LPS 4w and LPS 8w demonstrate more staining in CA3 than their aCSF controls (*p < 0.001). Within CA3, LPS 8w (‡) > LPS 4w (†) > LPS 2w (p ≤ 0.009) and LPS 8w has a higher density of staining in the DG than LPS 2w (†p = 0.028). DBH staining is more dense in CA3 that both DG and CA1 within aCSF 2w, LPS 4 and LPS 8w groups (p < 0.001) and within aCSF 8w CA3 DBH staining is more intense than CA1 (p = 0.046). Taken together, these data suggest that the LC projections to anterior hippocampus are effected by chronic neuroinflammation. Like TH in the brainstem and LC, DBH staining in the hippocampus corresponds with an increase in MHCII+ microglia (described in Chapter 5, Section 5.4.3). We observed no significant changes in brainstem levels of calcineurin or GDNF (p > 0.05).

120

Figure 33. Distribution of MHCII+ microglia and neurons in the LC and raphe nuclei MHCII+ microglia (brown) and TH+ neurons (blue) in the LC (a-d) and TrypH+ neurons (blue) in the raphe (e-h) are shown at 10X (a-h) and 60X images of MHCII+ microglia interacting with neurons in the raphe of a rat infused chronically with LPS for 8w are shown to the right (h1-h2). 121

Figure 34. LC TH and brainstem MHCII expression TH+ cells are visibly reduced (A, B) in the LC after 2 weeks of LPS infusion (*p < 0.001), but restored after 4 weeks infusion. MHCII+ microglia in the brainstem (C) were evaluated by densitometry in an area selected medial to the LC, ventral to the IVth ventricle and dorsal to the raphe. There is an overall main effect in which LPS-infused groups had greater estimated number of activated microglia (p = 0.023) than aCSF. LPS

122

4w express significantly fewer MHCII+ microglia than LPS 2w in the brainstem (†p = 0.028).

Figure 35. Hippocampal DBH Compared to controls, DBH staining intensity (A) is reduced in CA3 and DG of LPS 2w and elevated in CA3 of LPS 4w and LPS 8w (*p ≤ 0.007). DBH staining is more dense in LPS 4w CA3 and LPS 8w DG (†p ≤ 0.028) than LPS 2w in the corresponding regions. LPS 8w CA3 is more dense with DBH staining than both LPS 2w and LPS 4w (‡p ≤ 0.009).

123

3.4.4 Cerebellum microglia activation LPS is delivered into the IVth ventricle, thus the cerebellum is in close proximity to the entry point of LPS. While all infused rats tend to display an elevated MHCII+ microglia in the cerebellum compared to rats with no surgery (data not shown), animals infused with LPS express many more MHCII+ microglia (Figure 36). A visual rating scale of 0-6 was used to estimate MHCII+ microglia intensity in the cerebellum and averaged over two raters

(Figure 36B). Two-way ANOVA showed a main effect for treatment (F1, 31 = 12.7, p = 0.001) such that LPS-infused animals had more MHCII+ microglia at 4w (*p = 0.05) and 8w (*p = 0.007) in the area near the ventricle and intercalating between the layers of the vermis. Spearman rank order correlation reveals that the number of MHCII+ microglia cells in the cerebellum correlates with deficits seen in rearing frequency (r = -0.526, n = 19, p = 0.021), rearing duration (r = -0.580, n = 10, p = 0.009) and in the hanging task (r = -0.620, n = 19, p = 0.004); activities that require balance. Due to these correlations, no conclusions may be drawn relating deficits in rearing and the hanging task to integrity of the SNpc projections to the basal ganglia motor loop.

124

Figure 36. Cerebellar and brainstem MHCII+ microglia and TH/TrypH+ cells MHCII+ microglia (brown) and TH+/TrypH+ neurons (blue) are shown in the cerebellum and brainstem. MHCII+ microglia are observed after infusion of aCSF but are dramatically increased in LPS-infused brains and surround the cannula tract and the IVth ventricle. The cerebellum is populated with more MHCII+ microglia after 4w (*p = 0.05) and 8w (*p = 0.007) LPS infusion.

3.4.5 Data Summary Behavioral, cellular and molecular changes observed due to neuroinflammation induced by 2, 4 or 8 weeks of chronic LPS infusion (Table 5). LPS induced neuroinflammation causes a persistent reduction in rearing and impairment in the hanging task. MHCII+ microglia are highly expressed in SNpc as well as the brainstem and cerebellum. TH+ cells were reduced in SNpc and LC after 2w LPS infusion. No changes were observed in the striatum. DBH in the hippocampus originating from LC projections is reduced.

125

Striatum SN LC/Raphe Cerebellum

LPS: ↓ rearing, ↓ hanging Behavior No change: open field gross motor measures LPS: ↑ MHCII, LPS: ↑ MHCII, ↓ TH at 2w, ↓ TH and pTH at LPS: ↑ MHCII Histology No change TH ↓ DBH in 2w rebound by hippocampal 4/8w projection at 2w No change TH, Western blot SNAP 25 Binding No change DAT Table 5. Results summary

3.5 Discussion Genetics, exposure to environmental triggers, and chronic neuroinflammation are three elements widely accepted as factors in the etiology and progression of PD. Neuroinflammation is a relatively accessible target that may be vital for discovery of disease-modifying treatments. In order to better understand the effects of chronic neuroinflammation on behavioral and biochemical outcomes related to PD, we evaluated rats infused i.c.v. with LPS over three infusion periods: 2, 4 and 8 weeks.

3.5.1 Time-course of microglial activation Brains of PD patients and animal models of PD show exaggerated levels of activated microglia associated with Lewy bodies and degeneration (Imamura et al., 2003). The continuous i.c.v. LPS model is designed to resemble the chronic neuroinflammation characteristic of neurodegenerative diseases. Inflammation was experimentally defined as the presence of MHCII on microglia. LPS infused into the IVth ventricle circulates throughout the ventricular system, and also throughout the parenchyma due to its lipid solubility. Therefore, LPS could induce equivalent microglia activation in areas equidistant from the ventricles. Instead, however, activated microglia were distributed

126 heterogeneously throughout the brain and accumulated in regions that are recognized as vulnerable to neurodegeneration, including the SN. Microglia may become ‘primed’ by initial LPS exposure and subsequently show hyperresponsiveness to LPS (Henry et al., 2009). Alternatively, microglia may develop endotoxin tolerance to LPS and become less reactive (discussed more thoroughly in Chapter 5 Section 5.5.1). Whether microglia show endotoxin tolerance or become primed in response to LPS may depend on the dose and chronicity of exposure as well as the intensity of initial cellular responses such as cytokine production. However, because this model uses the slow and continuous infusion of LPS, microglia are likely to be exposed to and become activated by LPS gradually over the infusion period. Therefore, as microglia become activated one-by-one, they are likely to become tolerant at the same pace as they are activated, and, overall, maintain a low-level neuroinflammatory environment. Consistent with this postulation is evidence from another study recently conducted in our lab (Bardou et al., 2013) demonstrating that the general inflammatory profile of microglia in the brainstem was not changed in rats after continuous infusion of LPS over 3 or 8 weeks. This study found that LPS infusion over both 3 and 8 weeks induced an approximate 11- fold change in pro-inflammatory interleukin-1ß (IL-1ß) mRNA and 4-fold change in IL- 1ß protein. LPS infusion over 8 weeks, but not 3 weeks, produced significant elevations in protein expression of the pro-inflammatory cytokines IL-1α and IL-2. Similar to Il-1ß, LPS induced elevations in gene expression of pro-inflammatory TNFα, interferon-γ (IFNγ), IL-4, Il-5, IL-6, IL-12 and IL-13 as well as anti-inflammatory IL-4 and IL-10 that were not duration-dependent. Additionally, toll-like receptor 4 (TLR4) mRNA expression increased in the brainstem after both 3 and 8 weeks of LPS infusion. TLR4 is the receptor on microglia that recognizes LPS; this suggests an enhanced, not attenuated, ability of microglia to respond to LPS over the entire infusion period. Similar results were observed in the hippocampus as well (Bardou et al., 2013, described in Chapter 5, Section 5.5.1). The simultaneous and sustained expression of both pro- and anti-inflammatory cytokines supports the theory that microglia within the brainstem population are becoming activated and transitioning to an alternative activation state individually as they are each exposed to LPS over time; creating an environment of chronic, low-level neuroinflammation.

127

3.5.2 Compensatory recovery of aminergic systems during chronic neuroinflammation The number of dopaminergic neurons in the SNpc expressing TH and pTH are reduced after 2 weeks of LPS infusion, but restored after 4 weeks. Similarly, the number of noradrenergic neurons in the LC expressing TH and the density of noradrenergic projections to the hippocampus distinguished by DBA expression are reduced after 2 weeks of LPS infusion and restored after 4 weeks. Decreased availability of the rate-limiting enzymes TH and DBH indicates reduced production of dopamine and norepinephrine. Diminished production of neurotransmitters or, ‘luxury systems,’ may be a cellular defense strategy to divert energetic resources to survival mechanisms in a stressful environment. Interestingly, reductions in TH and DBH driven by the immune response to LPS are not sustained over 4 and 8 weeks, though the inflammatory environment persists or is elevated in the SNpc, LC and hippocampus. The decrease in TH+ cells and DBH+ projections after 2 weeks LPS infusion could be a response to the initial state of microglia activation that may be characteristically different from the activation profile after 4 or 8 weeks. For example, early innate immune activation is a pro-inflammatory and toxic state, but later microglia convert to an ‘alternative’ and anti-inflammatory immune profile characterized by tissue repair. We do not know the inflammatory profile of microglia in this study beyond their expression of MHCII, we can, however, predict their inflammatory profile from results of a more recent study conducted in our lab. Brainstem tissues express sustained elevation in pro- inflammatory cytokines (IL-1ß, TNFα, IFNγ, IL-4, Il-5, IL-6, IL-12 and IL-13) from 3 to 8 weeks of continuous LPS infusion, and a later increase of IL-1α and IL-2 after 8 weeks of infusion (Bardou et al., 2013). Similar results were found in LC projections to the hippocampus. These results suggest that microglia are in a pro-inflammatory state that is not appreciably attenuated between 3 and 8 weeks, however, we do not know if the levels were higher before or at 2 weeks LPS infusion when TH and BDH expression are reduced. Attenuated production of deleterious, pro-inflammatory factors, though not observed, could account for the rebound in TH and elevation of DBH. Conversely, a shift to an anti-inflammatory microglial profile or an increase in trophic factors could explain recovery in the aminergic systems as well. LPS infusion drives an increase in anti-inflammatory IL-4 and IL-10 in the brainstem after 3 weeks

128 infusion (Bardou et al., 2013). We do not know the expression levels of IL-4 or IL-10 before or at 2 weeks infusion, and it’s possible that the timing of their expression coincides with restored amine-producing function in the SN and LC between 2 and 4 weeks of continuous LPS exposure. We observed no changes in the trophic factor GDNF in this study, and no changes in BDNF were observed in the brainstem after 3 or 8 weeks of LPS infusion by Bardou et al. (2012). We did, however, observe an increase in gene expression of the glutamate-cystine anti-porter (xCT) after 3 and 8 weeks of LPS infusion that may indicate elevated production of the antioxidant glutathione by astrocytes. The SN is particularly vulnerable to oxidation (Floor and Wetzel., 1998; Venkateshappa et al., 2012; Zhang et al., 1999) and oxidation of dopamine augments excitatory currents at NMDARs in the SN (Wu and Johnson, 2011), therefore, increased antioxidant production would provide protection against the reactive oxygen and nitrogen species (ROS/RNS) released by activated microglia during the neuroinflammatory response (see Section 1.4.1b). In addition to particular susceptibility to oxidation, the SN and LC interact uniquely with the immune system due to anti-inflammatory properties of the monoamine neurotransmitters. Decreased dopamine and norepinephrine production after 2 weeks of LPS may further drive the inflammatory response. Neurons with reduced viability in pathological environments are likely to change their communication with the immune system, possibly by down-regulating the expression of constitutive markers such as CD200 and fractalkine and increasing the release of damage associated molecular patterns (DAMPs) like high mobility box group 1 (HMGB1). In addition, stabilization or increase of dopamine and norepinephrine production, respectively, after 4 and 8 weeks may help control the inflammatory process and have a protective effect (Gesi et al., 2000). LPS interacts directly and selectively with microglia and microglia activation initiates processes that may become self-perpetuating and lead to neurodegeneration. Yet, after prolonged exposure to neuroinflammation we see a rebound in enzymes involved in the production of aminergic neurotransmitters in the SNpc and LC. A single i.p. injection of LPS in a mouse or rat can cause PD-like cell loss in the SN (Qin et al., 2007; Zhang et al., 2009a). LPS given as a single bolus at a high concentration produces a peripheral immune response which is able to cause a CNS reaction, partially due to the travel of cytokines across the blood-brain barrier and to activation of the vagus nerve. It is possible

129 that a single i.p. bolus of LPS initiates a level of neuroinflammation that crosses a threshold which continuous low-level i.c.v. LPS infusion does not reach. After the initiation of a 2 week chronic infusion of 5ng/hr. LPS into the region directly above the SN in adult F-344 rats, significant loss of TH+ neuronal staining and loss of neurons in the SN were observed after 6 and 10 weeks, respectively, (Gao et al., 2002). Although this model was similar to ours, we did not observe continued loss of SN TH+ cells over time, and this may be due to the positioning of the cannula directly over the SN instead of the less direct IVth ventricle. However, another study from the same group in the same year reported that after unilateral SN LPS injection, TH+ cell number on the contralateral side was increased to “compensate for damage” (Hsieh et al., 2002), and we believe that this type of compensation is what we observe after i.c.v. LPS infusion. These studies differ in the chronicity and location of LPS infusion, but, taken together, indicate that a dose-dependent exposure to neuroinflammation in the SN is sufficient to drive reduction of TH+ cells and that the region is capable of compensatory up-regulation of TH+. A compensatory response in the LC has similarly been observed in post-mortem tissue from PD patients in which LC neurons were lost and remaining neurons had thinned dendritic arborization and irregular morphology (Balovannis et al., 2006; McMillan et al., 2011), similar to our observations. Within surviving LC neurons, TH expression per cell was similar to age-matched controls and DBH expression per neuron was increased (McMillan et al., 2011), supporting our findings of increased hippocampal DBH after 4 and 8 weeks of neuroinflammation. These findings suggest that compensatory responses by aminergic systems in a neuroinflammatory environment may be capitalized upon, in addition to reduction of inflammation, as potential pharmacologic therapy for PD.

3.5.3 Motor impairment due to neuroinflammation Motor symptoms appear in humans with PD and in animal models after a loss of dopaminergic neurons in the SNpc exceeding approximately 80%. We did not predict this level of injury in our model because we have previously observed unimpaired swimming in the Morris water maze. In the current study we evaluated motor behaviors sensitive to SN cell loss with tasks appropriate for bilateral damage that do not use food reward as a motivator, such as observations of rearing in the OF and performance in a hanging task requiring balance and fore-limb strength. LPS-infused rats reared less and were impaired 130 at the hanging task. These deficits correlated with neuroinflammation in the SN, but correlated more strongly with inflammation measured in the cerebellum and brainstem nuclei, which are also intricately involved in PD postural instability (Grimbergen et al., 2009). We also observed what may be considered hypermotility in rats after 4 weeks of continuous LPS infusion, as evidenced by more entrances between zones in the open field task and increased velocity after 8 weeks LPS infusion in the water maze described in Chapter 4 (4.4.1, Figure 38C). Hypermotility is consistent with behavior after bilateral intranigral LPS injections observed by Hsieh et al., (2002), which the authors hypothesize may be due to an increase in serotonergic or noradrenergic transmission. We saw reductions in SNpc and LC TH+ cells as well as hippocampal DBH after 2 weeks LPS infusion that were restored or elevated, respectively, after 4 weeks. Consistent with the interpretation by Hsieh et al. (2002), restoration and elevation in aminergic precursors after 4 weeks LPS of infusion may be a compensatory response to a neuroinflammatory environment.

3.5.4 Conclusions Continuous central infusion of low-level LPS induces a chronic neuroinflammatory environment that is more pronounced in self-selected regions including the SN and brainstem. Neuroinflammation drove a reduction in enzymatic precursors to dopamine and norepinephrine in the SNpc, LC and hippocampus. These reductions were transient, as the enzyme TH was restored in SNpc and LC, and DBH elevated past basal levels in the hippocampus when observed two weeks later. Moreover, the restoration of TH and elevation of DBH occurred despite the continuing infusion of LPS and evidence of a persistent pro-inflammatory environment. This phenomenon may be attributed to unobserved changes in the magnitude of microglia activation or the introduction of anti- inflammatory cytokines. Furthermore, up-regulation of TH and DBH may be a protective compensatory response to the neuroinflammatory process. This study shows that both neuroinflammation and the aminergic neurotransmitter systems interact, and that both are important targets for pharmacological therapy.

131

Chapter 4: Delayed memantine does not reverse LPS-induced microglia activation or memory impairment; A pilot investigation.

4.1 Brief rationale: We have previously found that memantine given at the same time as initiation of LPS infusion prevents the increase in MHCII+ microglia and improves hippocampal- dependent learning and memory (Rosi et al. 2006; Rosi et al. 2009). Unlike the cannabinoid agonists we tested (Marchalant et al. 2009a-c), memantine does not reduce the number of MHCII+ microglia nor improve spatial learning and memory in aged rats (unpublished results). These observations suggest that NMDAR antagonism is not a pathway to attenuate neuroinflammation and cognitive impairment in aged rats, or that memantine can prevent but not reduce microglia activation and cognitive deficit. Therefore, I set up a pilot study to test whether memantine is able to reduce, in addition to prevent, microglia activation in LPS-infused young rats. The experiments described in this chapter were part of the study conducted in Chapters 3 and 5. Therefore, there is some overlap in the procedures and results. However, the memantine treated groups in this pilot were not included in all of the biochemical and histological investigations, and are separated into this chapter for simplicity.

4.2 Introduction Memantine is currently one of few treatments for Alzheimer’s (AD), a disease characterized by progressive neurodegeneration and memory loss. Neuroinflammation is present early in the AD brain, prior to and predictive of the development of gross pathology and cognitive symptoms. Preventing neuroinflammation is protective against disease development; however, clinical trials of non-steroidal anti-inflammatories (NSAIDs) have failed to change disease symptoms when given later in disease progression. Memantine prevents microglia activation and cognitive impairment in an animal model of neuroinflammation, but not in aged rats in which neuroinflammation is already established

132

(unpublished observations). If reducing neuroinflammation is a beneficial property of memantine, in addition to its primary mechanism of action on the glutamatergic system, then it is important that we understand the optimal time to initiate treatment in order to best utilize memantine in clinical practice. Pre-symptomatic AD brain pathology initiates neuroinflammation that, in turn, drives glutamatergic dysfunction (Wenk 2006; Wenk et al., 2006). Microglia activation by soluble amyloid ß (Aß) oligomers, the primary AD pathology, leads to neuronal death in vitro that is dependent upon coincident production of inducible nitric oxide synthase (iNOS) and activation of both tumor necrosis factor alpha receptors (TNFαRs) and NMDARs (Floden et al., 2005). In addition, Aß interacts directly with glutamatergic NMDARs (Danysz and Parsons 2012; Texidó et al., 2011; Rammes et al., 2011). Furthermore, neuroinflammation itself can also drive excessive glutamatergic function, in the absence of irregular Aß, and conclude in seizure activity. Memantine is a non-competitive NMDAR antagonist with fast on/off kinetics that allows memantine to block the NDMAR channel and reduce Ca2+ entry at resting potential, but still permits activation of the NMDAR by action-potential-evoked release of pre- synaptic glutamate and allows the normal induction of LTP (Wenk et. al, 2006; Rogawski et al., 2003; Lipton 2004). Memantine also preferentially blocks currents at extrasynaptic NMDARs (Xia et al., 2010), which are more likely to be activated by elevated extracellular glutamate and lead to excitotoxic cell death than synaptic NMDARs (Hardingham and Balding, 2010). These properties explain why memantine does not interfere with LTP and also suggest that memantine may protect against excitotoxic cell death. Furthermore, Aß is elevated by extrasynaptic but not synaptic NMDAR activation, and this effect is attenuated by memantine (Bordji et al., 2010; Hoey et al., 2009; Rönicke et al., 2011), suggesting that blockade of extrasynaptic NMDARs by memantine will lead to a decrease in Aß, less pro-inflammatory stimulation of microglia and less disease pathology overall. We have previously shown that memantine, through modulation of glutamatergic activity, attenuates neuroinflammation and inflammation-induced cognitive impairment. Cultured microglia exposed to lipopolysaccharide (LPS) produce the pro-inflammatory cytokine TNFα and nitric oxide (NO), and this production is not attenuated by memantine, demonstrating that memantine does not have a direct anti-inflammatory effect on microglia

133

(Rosi et al., 2009). Neuroinflammation created by the chronic infusion of LPS into the IVth ventricle drives microglia activation as well the loss of sparse-encoding of the immediate early gene Arc in a novel environment and cognitive deficit in the Morris water maze; the latter two are indicative of glutamatergic dysregulation and all were attenuated by memantine treatment (Rosi et al., 2006; Rosi et al., 2009). Memantine is also protective against cytotoxicity due to experimentally-induced mitochondrial energy deficit (Wenk et al., 1996). Taken together, these results suggest that memory impairment under conditions of experimentally-induced chronic neuroinflammation or in AD is driven by an aberrant increase in the availability of glutamate at the NMDAR that memantine abolishes. Memantine does not, however, reduce either memory impairment or neuroinflammation associated with normal aging (unpublished observations), although it is able to do so in young rats with experimentally-induced neuroinflammation as discussed above. Furthermore, other drugs which reduce glutamate neurotransmission, cannabinoids and caffeine in particular, reduce cognitive deficit and neuroinflammation in aged rats and in young rats with LPS-induced chronic neuroinflammation (Marchalant et al., 2007; Marchalant et al., 2008; Marchalant et al., 2009; Brothers et al., 2010, Chapter 2). It is possible memantine does not produce ameliorative effects against neuroinflammation in normally aged rats, because treatment needs to begin before the onset of neuroinflammation. Therefore, we designed this pilot experiment to investigate whether delayed memantine treatment could reduce microglia activation and behavioral impairment. If memantine does not reduce neuroinflammation as effectively as it prevents neuroinflammation, then we need to consider this factor and initiate clinical treatment with memantine at a more advantageous time.

4.3 Methods

4.3.1 Subjects, surgical procedures and drug administration Young (4 mo.) male F-344 rats (Harlan Sprague–Dawley, Indianapolis, IN) were singly housed in Plexiglas cages in a temperature-controlled room (22 ºC) on a 12/12-hr. light–dark cycle (lights off at 09:00) with ad libitum access to food and water. All rats were given 1 week to acclimate to their new environment before beginning surgical

134 procedures. Body weights, general health, movement and behavior were closely monitored throughout the study. Artificial cerebrospinal fluid (aCSF, 140 mM NaCl, 3.0 mM KCl, 2.5 mM CaCl2, 1.0 mM MgCl2, and 1.2 mM Na2HPO4 adjusted to pH 7.4) or LPS (0.25µg/hr., 1.66¯¯ mg/ml dissolved into aCSF; Sigma, St. Louis, MO, USA E. coli, serotype 055:B5, TCA extraction) were chronically infused i.c.v. over 4 or 8 weeks. A cannula was implanted into the IVth ventricle (-2.5 AP and -7 DV relative to lambda) and attached (via Tygon tubing, 0.06 O.D.) to an osmotic minipump (Alzet model #2006, 0.15µl/hr., Durect Corp., Cupertino, CA, USA) as previously described (Hauss-Wegrzyniak et al., 1998a). Calculations using the average fill volume of pump model #2006 allowed for release over an 8 week period. Post‐operative care included application of topical analgesics (lidocaine 1% solution and Neosporin) over the sutures, and subcutaneous injection of 2 ml isotonic saline to prevent dehydration during recovery. All rats were sacrificed after 4 or 8 weeks of i.c.v. infusion of aCSF or LPS. Memantine (15 mg/kg/day) or isotonic saline (0.9%) were delivered over the last 2 weeks of aCSF or LPS i.c.v. infusion (weeks 3-4 of a 4 week infusion period or weeks 7-8 on an 8 week infusion period), creating a total of 8 experimental groups: aCSF 4w (n = 11), aCSF 4w/Mem (n = 7), LPS 4w (n = 18), LPS 4w/Mem (n = 13), aCSF 8w (n = 7), aCSF 8w/Mem (n = 8), LPS 8w (n = 10) and LPS 8w/Mem (n = 9). Memantine and saline were released continuously from an osmotic minipump (Alzet model #2Ml2) placed subcutaneously on the dorsal side (Rosi et al. 2006), and this placement did not inhibit normal movement.

4.3.2 Spatial learning and memory: Morris water maze In order to determine if delayed memantine treatment reduced inflammation- induced spatial memory deficit, we tested animals in the Morris water maze. Behavioral testing was initiated 1 week prior to sacrifice, during the 4th or 8th week of LPS/aCSF infusion and during the 2nd week of memantine/saline treatment. Each animal was acclimated to handling and to the experimenters on 3 days during the week prior to testing for 5 minutes each day. All behavioral testing took place in the same room outside of the colony, under similar dim lighting conditions. Spatial memory was evaluated as ability to find a hidden escape platform (10 cm diameter) that remained in a constant location, 135 submerged (2.5 cm below surface) within a dark grey water pool (85 cm radius) maintained at 20-25 ºC. The pool had a large black square above the rim as a proximal cue and was centered in the room with visual stimuli on the wall and a threshold as distal cues. Latency to find the platform, distance traveled, speed and other variables were tracked and recorded (Noldus EthoVision 3.1, Noldus, Leesburg, VA). On the first day, rats were placed upon the hidden escape platform for 30s. Subsequently, they completed 6 trials with 1 hr. between inter-trial-interval, and repeated this over the next 3 days (24 trials total). During each trial, the animal was placed into the pool from one of 6 entry points that were evenly spaced along the edge of the pool. Rats entered from each point once per day in the same order, and the order changed between days. Trial duration was 60 sec or ended what the rat found the hidden platform. The animal remained or was placed on the platform for an additional 30 sec following each trial. Latency to find the platform, distance swam, thigmotaxis and swim speed were analyzed. At the completion of 6 trials on the final testing day, animals completed a probe trial in which the platform was removed. The percentage of the 60 sec trial that they spent within the vicinity (42 cm diameter, 25% of pool area) of the original platform location was analyzed. Finally, to control for possible treatment- or drug-induced deficits in visual acuity, rats were tested on 2 visible platform trials in which the platform was moved to a new location, raised 2 cm above the water surface, and marked with a blue neoprene glove.

4.3.3 Immunohistochemistry The extent of inflammation was investigated by immunohistochemistry in the hippocampus. Activated microglia, defined by expression of Class II major histocompatibility complex (MHCII), were quantified. Tissues were prepared and stained with anti-MHCII (1:400, mouse monoclonal, Pharmigen, San Diego, CA) using the method described in Chapter 3 (3.3.3b/c). Two coronal slices of anterior hippocampus were examined using light microscopy. Number of MHCII immunopositive (MHCII+) cells within the CA1, CA3 and DG subregions of the hippocampus was determined using MetaMorph imaging software (Universal Image Corporation, West Chester, PA) as previously reported in detail (Rosi et al., 2005b). Briefly, two slices per animal were chosen and both the left and right hemispheres were counted for total cell number.

136

4.3.4 Statistics SigmaStat software was used to compare groups by two- and three-way Analysis of Variance (ANOVA) with Fisher LSD as the preferred post-hoc. Graphs are shown with SEMs. Differences between LPS and aCSF are indicated with *, differences between 4 and 8 weeks infusion duration are indicated with †, and differences between saline and memantine drug treatment are indicated by §.

4.4 Results Results in this pilot study are compared to results from Rosi et al., (2006) (Figure 37) in which 4 weeks of memantine treatment initiated at the same time as chronic LPS infusion prevented both inflammation-induced spatial memory impairment in the Morris water maze and LPS-induced microglia activation.

Figure 37. Memantine prevents LPS-induced spatial memory deficit and microglia activation (Rosi et al. 2006). Latency to find the hidden platform in the Morris water maze (A) is impaired by LPS infusion (*), and prevented by memantine (†). Similarly, number of MHCII+ microglia (B) is increased by LPS infusion, and the full extent of the increase is prevented by memantine treatment. 137

4.4.1 Morris water maze Continuous LPS infusion over 4 weeks, but not 8 weeks, impairs spatial learning and memory in the Morris water maze compared to aCSF controls, and memantine treatment over the last two weeks does not improve performance (Figure 38). Latency to find the hidden platform was evaluated by a 4-factor (inflammation condition, infusion duration, drug treatment and trial day) mixed repeated measures ANOVA. There were main between-subjects effects of inflammation condition (F1, 119 = 17.747, p < 0.001) and infusion duration (F1, 119 = 6.354, p = 0.013), a main within-subjects effect of trial day (F3,

357 = 367.964, p < 0.001), an interaction between inflammation condition and infusion duration (F1, 119 = 6.597, p = 0.011), an interaction between inflammation condition and trial day (F3, 357 = 7.913, p < 0.001), and an interaction between infusion duration and trial day (F3, 357 = 3.430, p = 0.017). Post-hoc analysis indicates that latency to the hidden platform is increased in LPS 4w and LPS 4w/Mem compared to all aCSF-infused groups (*p ≤ 0.004), that LPS 8w do not take significantly longer to find the hidden platform than any aCSF-infused group (p > 0.05) and find this platform in less time than LPS 4w and LPS 4w/Mem (†p < 0.001), and LPS 8w/Mem take longer to find the platform than aCSF 8w/Mem (*p = 0.014) and LPS 8w (§p = 0.006). LPS 4w and LPS 4w/Mem spent more time to find the hidden platform, if at all, and they spent less time looking in the correct area of the pool when the platform was removed during the probe trial (Figure 38B). There was a main effect of experimental group on the percentage of trial time spent within the quarter of the pool area surrounding the former platform location (F7, 125 = 3.116, p = 0.005). LPS 4w spent less time than aCSF 4w or LPS 8w (*†p ≤ 0.014). LPS 4w/Mem was not significantly different from LPS 4w, and spent less time in the area of the absent platform than aCSF 4w (p = 0.048), but not significantly less than aCSF 4w/Mem (p = 0.164). The aCSF 4w/Mem experimental group was intermediate between aCSF 4w and LPS 4w/Mem, and not significantly different from either (not shown on graph). Similar to the latency and distance results, LPS 8w was no different in probe trial performance from aCSF 8w controls, but LPS 8w/Mem were impaired compared to aCSF counterparts (*p = 0.002) and LPS 8w (§p = 0.011). Increased latency to find the platform could be explained by decreased velocity. A 4-factor mixed repeated measures ANOVA on velocity (Figure 38C). There was a main

138 between-subjects effect of inflammation condition (F1, 119 = 4.235, p = 0.042), a main within-subjects effect of trial day (F3, 357 = 75.934, p < 0.001), an interaction between inflammation condition and trial day (F3, 357 = 3.982, p = 0.008), and an interaction between infusion duration and trial day (F3, 357 = 3.985, p = 0.008). Post-hoc analysis indicates that all experimental groups swam similar speeds across days, except LPS 8w, which swam significantly faster than every other experimental group (*†§p ≤ 0.035). Finally, a 4-factor mixed repeated measures ANOVA was performed on the percentage of trial time spent in the pool perimeter (thigmotaxis, Figure 38D). There was a main between-subjects effect of inflammation condition (F1, 119 = 5.038, p = 0.027), a main within-subjects effect of trial day (F3, 357 = 294.228, p < 0.001), and an interaction between infusion duration and trial day (F3, 357 = 3.500, p = 0.016). Post-hoc analysis indicates that LPS 4w, LPS 4w/Mem and LPS 8w/Mem performed similarly (p > 0.05) and spent more time in the pool perimeter than LPS 8w (†§p ≤ 0.034). Continuous LPS infusion over 4 weeks impaired spatial learning and memory in the Morris water maze compared to aCSF infusion. This is demonstrated by increased time spent looking for the hidden platform, increased distance traveled, increased time in the pool perimeter and decreased time spent in the area of the missing platform during the probe trial. Memantine treatment over the last 2 weeks of the 4 week LPS infusion did not improve performance in any measure. Conversely, continuous LPS infusion over 8 weeks did not produce impairment in in the Morris water maze unless paired with memantine treatment during the last 2 weeks of the infusion period. Increased latency to find the platform in LPS 4w, LPS 4w/Mem and LPS 8w/Mem is not explained by slow swim speed, but can be attributed to more time spent swimming around the perimeter of the pool.

139

Figure 38. Morris water maze Figures mark main effects between groups across trial days with * between aCSF and LPS, † between 4 and 8 weeks infusion duration and § between saline and memantine treatment. LPS 4w and LPS 4w/Mem took longer to find the hidden platform (A) and spent more time in the pool perimeter (D) than aCSF-infused groups (*p ≤ 0.004) and LPS 8w (†p < 0.001). LPS 4w spent less trial time looking in the quarter of the pool from which the platform was removed during the probe trial (B) than aCSF 4w or LPS 8w (*†p ≤ 0.014). LPS 8w/Mem took longer to find the platform (A), spent more time in the pool perimeter (D), and less time searching in the area of the removed platform (B) than aCSF 8w (*p ≤ 0.006) or LPS 8w (§≤ 0.011). LPS 8w swam more quickly (p ≤ 0.002) than aCSF 8w (*), LPS 4w (†) and LPS 8w/Mem (§).

140

4.4.2 Microglia activation The number of MHCII+ microglia increased in the hippocampus of LPS-infused rats, but was unaffected by memantine treatment Figure 39. An ANOVA reveals main effects of inflammation condition (F3, 245 = 38.959, p < 0.001) and hippocampal region (F2,

245 = 26.883, p < 0.001), and an interaction between the two (F6, 245 = 9.547, p < 0.001), but no main effect of drug treatment (F1, 245 = 0.400, p = 0.528). Within all LPS infused groups, there were more MHCII+ microglia in CA3 and DG than CA1 (p ≤ 0.025). There are more MHCII+ microglia overall in LPS 4w and LPS 4w/Mem than aCSF 4w or aCSF 4w/Mem (*p ≤ 0.027). Similarly, there are more MHCII+ microglia in LPS 8w and LPS 8w/Mem than aCSF 8w or aCSF 8w/Mem (*p ≤ 0.001). Finally, more MHCII+ microglia are present in the hippocampus of LPS 8w (p ≤ 0.043) and in LPS 8w/Mem (†p ≤ 0.001) than both LPS 4w and LPS 4w/Mem. The number of MHCII+ microglia in the hippocampus correlate with poor performance in the Morris water maze. MHCII+ microglia in the CA1 (r = 0.347, p = 0.004) and CA3 (r = 0.298, p = 0.015) correlate with increased latency to the platform. Activated microglia in CA1 (r = -0.372, p = 0.002), CA3 (r = -0.364, p = 0.002) and DG (r = -0.271, p = 0.027) correlate less improvement in latency between the first and last trial days. Greater distance traveled is correlated with MHCII+ microglia in the CA3 (r = 0.265, p = 0.031), and time spent in the pool perimeter is correlated with MHCII+ in CA1 (r = 0.250, p = 0.041). Cerebellar inflammation could potentially cause difficulties with movement. Microglia in the cerebellum (Figure 36), however, do not correlate with WM measures such as latency to find the hidden platform (r = 0.162, p = 0.514) distance traveled (r = 0.194, p = 0.431), speed (r = 0.274, p = 0.266) or thigmotaxis (r = -0.078, p = 0.754), strengthening the idea that water maze deficits reflect changes due to inflammation in other regions, including the hippocampus.

141

Figure 39. Microglia activation

4.5 Discussion Continuous LPS infusion over four weeks drives an increase in the number of activated microglia in the hippocampus and a corresponding cognitive impairment. Memantine treatment overlapping with the four week LPS infusion prevents microglia activation and disruption in water maze performance (Rosi et al., 2006). Delayed memantine treatment, however, did not modify cognitive impairment in the Morris water maze or number of MHCII+ microglia induced by four weeks LPS-infusion. Similarly, memantine does not attenuate memory impairment of microglia activation in aged rats (unpublished observations). This may be confirmation that memantine is not effective when treatment is initiated after a neuroinflammatory environment is already established. The lack of improved performance and unchanged MHCII expression in this study could also be attributed to the shorter, two week treatment duration, compared to the four week treatment duration overlapping with LPS infusion used by Rosi et al., (2006). Rather than follow-up on this query, I became interested in defining a mechanism by which rats may 142 improve performance in the Morris water maze after 8 weeks LPS infusion and understanding why memantine impaired performance in this group. The LPS-induced water maze performance deficit observed after 4 weeks infusion disappeared by 8 weeks infusion with no additional manipulations. While LPS 8w performed no differently than aCSF controls in the water maze, this group was significantly impaired when treated with memantine. These results suggest that LPS 8w improved through a mechanism involving a change in glutamatergic function that was disrupted by delayed memantine treatment. Memantine is a non-competitive NMDAR antagonist with on/off kinetics that allow for normal synaptic transmission, but attenuate signals from excessive glutamate within the synapse (Parsons et al. 2007b). Because memantine prevents LPS-induced memory impairment (Rosi et al. 2006; Rosi et al. 2009), we hypothesize that chronic neuroinflammation elevates extracellular glutamate, which excessively depolarizes post-synaptic NMDARs and interferes with memory processing (Wenk et al. 2006; Rogawski & Wenk 2003; B. H. Hauss-Wegrzyniak et al. 2002). If extracellular glutamate is elevated by chronic LPS infusion and causes memory impairment after 4 weeks, but not after 8 weeks of LPS infusion, then it is possible that glutamate regulation returned a homeostatic balance. Additionally, if neuroinflammation initially drives a hyperglutamatergic state which produces a learning and memory impairment that is attenuated by memantine, and if glutamatergic homeostasis is restored over time, then later treatment with memantine may drive memory impairment through creation of a hypoglutamatergic state. Three potential mechanisms to restore glutamatergic balance are reduced glutamate release, attenuated post-synaptic response and increased glutamate uptake; these are explored in the following chapter. The initial mechanism by which LPS drives learning and memory impairment is through activation of TLR4 on microglia and stimulation of an immune response. In this study, the number of MHCII+ microglia continue to increase, although learning and memory are restored. It is possible that MHCII expression is a good marker of initial reactivity to LPS exposure, but that MHCII may remain on the microglia membrane past the initial MA-type activation period. As described in Section 1.2.1 and Appendix B: Microglia activation states and immune factors, microglia activation states vary across a large spectrum, and expression of MHCII alone is not sufficient to define the activity of

143 microglia. In the final chapter of this thesis, Chapter 6, I make an effort to thoroughly investigate microglia activation in our model. Furthermore, it is possible that microglia develop tolerance to LPS over time and that this contributes to attenuation of memory impairment in this study, an issue that is not supported by recent (unpublished) findings in our lab and further addressed in the following chapter.

144

Chapter 5: Glutamate transport may promote time-dependent recovery from inflammation-induced hippocampal-dependent spatial memory deficit

5.1 Brief rationale Previously, I investigated the effects of long-term chronic neuroinflammation produced by 4 weeks of LPS infusion on memory and used pharmacological interventions. In this chapter as well as Chapters 3 and 4, I shortened and lengthened the duration of LPS infusion to investigate changes over time. Interestingly, longer duration of LPS infusion resulted in resolution, not amplification, of cognitive impairment. Surprisingly, memantine caused a cognitive deficit in this group (4), suggesting that a compensatory adjustment occurred over time in the glutamatergic system that was then disrupted by memantine treatment. Here, I focus on possible compensatory mechanisms related to glutamate regulation in the hippocampus.

5.2 Introduction Chronic neuroinflammation is widely considered a risk factor for age-associated neurodegenerative diseases, including Alzheimer’s disease (AD) (Akiyama et al. n.d.; McGeer & McGeer 2002; Cooper et al. 2000; McGeer & McGeer 1999; for a review see Chapter 1 Section 1.6). Activated microglia are detectable many years prior to the onset of neuropathological and symptomatic changes in AD, are found in brain regions that ultimately show significant neuropathology, and correlate better with degeneration and cognitive impairment than other disease pathology such as amyloid ß (Cagnin et al. 2006; Edison et al. 2008; Okello et al. 2009; Hoozemans et al. 2006). A pro-inflammatory environment is not only present in AD, but also sufficient to drive cognitive impairment in animal models and is the primary event leading to the development of dementia pugilistica in humans. These observations suggest that neuroinflammation is an early and important contributor to cognitive deficit in AD.

145

Neuroinflammation appears early, is sufficient to drive memory impairment, and is predictive of later cognitive impairment, highlighting a lag between the initiation of the neuroinflammatory response and later cognitive impairment. Vulnerable brain regions are likely exposed to a pro-inflammatory environment for many decades (Agostinho et al., 2010; Bilbo, 2010; Heneka et al., 2010), yet do not show significant pathology until advanced age (Eikelenboom et al., 2010; Herrup, 2010; Hoozemans et al., 2006), suggesting that younger brains are able to compensate for the presence of continuous, low- level neuroinflammation and postpone the negative consequences. We have reproduced a similar low-level, chronic neuroinflammation in young male rats by continuous infusion of nanomolar levels of lipopolysaccharide (LPS) into the IVth ventricle, and have systematically documented a slowly evolving and region-specific series of time-dependent changes. Activated microglia are diffusely scattered throughout the brain after two days of LPS infusion, and, over four weeks, this number gradually decreases in most regions with some key exceptions: a large number of activated microglia concentrate within the dentate gyrus (DG) and CA3 subregions of the hippocampus (Wenk et al., 2000). The increased number of activated microglia in the hippocampus is associated with a 3-fold increase of Il-1ß mRNA (Hauss-Wegrzyniak et al., 1988), a reduction in the number of N-methyl-d-aspartate (NMDA) glutamate receptors within the DG and CA3 hippocampal areas without evidence of neuronal loss (Rosi et al., 2004), impaired long- term potentiation (Hauss-Wegrzyniak et al., 2002) and an impairment in spatial memory but not object recognition (Hauss-Wegrzyniak et al., 1998). While cell loss is not evident in the hippocampus, four weeks of LPS infusion is associated with the loss of pyramidal cells in layers II and III of cortex (Hauss-Wegrzyniak et al., 2002) as well as enlarged lateral ventricles and shrinkage of the temporal lobe regions identified by MRI (Hauss- Wegrzyniak et al., 2000). Taken together, these findings suggest that LPS quickly initiates a global inflammatory response that is later concentrated within the hippocampus and causes AD-like learning and memory deficits. Here, we evaluated the response to LPS exposure over 2, 4 and 8 weeks. The behavioral results indicate that chronic, low-level neuroinflammation drives a cognitive impairment from which young animals eventually recover despite the presence of continued LPS infusion and neuroinflammation. We concentrated on mechanisms related

146 to the regulation of glutamate because our previous work demonstrates that decreasing glutamate release with caffeine or cannabinoids (Brothers et al., __; Marchalant et al., __) or blocking NMDARs with the non-competitive antagonist memantine (Rosi et al., __) is sufficient to prevent spatial memory impairment; indirectly suggesting that neuroinflammation creates an environment of excessive glutamate. We found increased expression of glutamate transporter 1 (GLT1, excitatory amino acid transporter 2), a protein that sequesters glutamate away from the synapse, in young rats after prolonged neuroinflammation that correlates with cognitive recovery. Increasing glutamate transport may be a natural compensation to a neuroinflammatory environment that prolongs time before manifestation of cognitive impairment, and may be a pharmacologic target for prevention of neuropathology and symptom onset as brains age.

5.3 Methods

5.3.1 Subjects, surgical procedures and LPS administration Young (4 month aged) male F-344 rats (Harlan Sprague–Dawley, Indianapolis, IN) were singly housed in Plexiglas cages in a temperature-controlled room (22 ºC) on a 12/12 hr. reverse light-dark cycle (lights off at 09:00) with ad libitum access to food and water. All rats were allowed 1 week to acclimate to their new environment. Body weights, general health, movement and behavior were closely monitored throughout the study. Artificial cerebrospinal fluid (aCSF, 140 mM NaCl, 3.0 mM KCl, 2.5 mM CaCl2, 1.0 mM MgCl2, and 1.2 mM Na2HPO4 adjusted to pH 7.4) or LPS (0.25µg/hr., 1.66¯¯ mg/ml dissolved into aCSF; Sigma, St. Louis, MO, USA E. coli, serotype 055:B5, TCA extraction) were chronically infused i.c.v. over 2, 4 or 8 weeks (Table 3, Table 4). A cannula was implanted into the IVth ventricle (-2.5AP and -7DV relative to lambda) and attached (via Tygon tubing, 0.06 O.D.) to an osmotic minipump (Alzet model #2006, to deliver 0.15µl/hr.; Durect Corp., Cupertino, CA, USA) as previously described (Hauss- Wegrzyniak et al., 1998a). Calculations using the average fill volume of pump model #2006 allow for release over an 8 week period. Post‐operative care included topical analgesics (lidocaine 1% solution and Neosporin) applied to the incisions, and a subcutaneous injection of isotonic saline (2 ml) to prevent dehydration during recovery. All rats were sacrificed after 2, 4 or 8 weeks of aCSF or LPS infusion. 147

5.3.2 Spatial learning and memory: Morris water maze AD-like memory impairment was examined by performance in the Morris water maze, a test sensitive to disruption in the hippocampus and LPS-induced neuroinflammation (Hauss-Wegrzyniak et al., 1998a), as described in Chapter 4 (4.3.2). Rats were tested during the last week of the 4 and 8 week infusion periods. Rats were not tested after only 2 weeks infusion because they were recovering weight lost after surgery and were likely still suffering from proximal effects of LPS, such as sickness behavior, which is a confound to motor-dependent tasks. Briefly, spatial learning and memory was evaluated as ability to find a hidden escape platform within the water pool across 4 days with 6 trials per day. Latency to find the platform, distance traveled, thigmotaxis and velocity were evaluated. At the end of the final day of testing, animals completed a probe trial in which the platform was removed and each rat was placed into the pool. The time that they spent within the vicinity (25% of pool area) of the original platform location was analyzed.

5.3.3 Histological procedures Activated microglia cells were visualized with an antibody directed against Class II major histocompatibility complex (MHCII; 1:400, Pharmigen, San Diego, CA). Synaptosome associated protein 25 (SNAP25) is involved in vesicle docking, and was evaluated (1:750, AbCam, Cambridge, MA) and the glutamate transporter GLT1 was also evaluated (1:2000, Millipore, Billerica, MA). Tissues were prepared and stained using the standard avidin/biotin peroxidase method described in Chapter 3 (3.3.3b/c). Two coronal slices of anterior hippocampus were examined using light microscopy for both MHCII and GFAP. Density of MHCII+ cells within the CA1, CA3 and DG subregions of the hippocampus (Figure 26A) was determined using MetaMorph imaging software (Universal Image Corporation, West Chester, PA) as previously reported in detail (Rosi et al., 2005b). Briefly, two slices per animal were chosen and both the left and right hemispheres were counted for total cell number. SNAP25 and GLT1 were evaluated as an area percentage of staining above a threshold density using Nikon Elements (Nikon, Melville, NY).

148

5.3.4 Western blot analysis Western blots were conducted on the sub-dissected CA3/DG region of the hippocampus to investigate indicators of cellular integrity and function (calbindin, HSP 70, HSP90, SNAP25 and OxPhos IV), evidence of neuron-glia communication (HMGB1 and fractalkine ligand and receptor) and changes in NMDAR subunit expression (NMDAR1, pan-NR1, phospso-NR1, NMDAR2A, NMDAR2B). Hippocampi were prepared and evaluated by Western blot following the method detailed in Chapter 3 (4.3.4a). Actin, a constitutively expressed cytoskeletal protein was used as a measure of total protein expression (1:5000, 42kDa, Sigma). Anti-glial fibrillary acidic protein (GFAP), an intermediate filament protein found in astrocytes, was investigated as a measure of astrogliosis (1:1000, 51kDa, Millipore). NMDAR subunits were investigated using the following antibodies: NMDAR1 (1:2000, 120 kDa, Millipore), NR1 CT (1:1700, 130-150 kDa, Millipore), phospho-NR1 Ser896 and phosphor-NR1 Ser897 (both 1:1300, 120 kDa, Millipore), NMDAR 2A (1:750, 180 kDa, Millipore), NMDAR2B (1:750, 180 kDa, Millipore). Calbindin, a calcium binding protein found in neurons, was probed as a measure of post-synaptic calcium regulation (1:1500, 28kDA, Swant). Fractalkine (CX3CL1) is chemokine constitutively released by neurons that maintains the resting state of microglia through surface receptors. Anti-fractalkine ligand (1:100, 42/90kDa, eBioscience) produced a band at ~42kDa and one at ~90 kDa which is thought to represent the soluble form of this ligand (~40 kDa). Fractalkine receptors on microglia were evaluated using anti-fractalkine R (1:3000, 52 kDa, eBioscience). High-mobility box group 1 (HMGB1), a pro-inflammatory signal released from damaged neurons, was investigated (1:5000, 25-27kDa, AbCam). Release of heat shock proteins (HSPs) from neurons is enhanced with local heat stress. HSPs 70 and 90 were evaluated with anti-HSP 70 (1:1000, 70kDa, BD) and anti-HSP 90 (1:1000, 90kDa, BD). SNAP25, a synaptosome associated protein involved in vesicle docking, was evaluated for indications of synaptic stripping using anti-SNAP25 (1:2000, 25kDa, AbCam).

5.3.5 Nitric oxide (NO) release from BV-2 microglia In order to confirm that the LPS contained in the osmotic minipumps would still elicit an immune response after a period of 8 weeks at body temperature, we filled osmotic mini-pumps with LPS and incubated them in a 0.9% saline solution at 37º C for 4, 6 or 8 149 weeks. Freshly prepared LPS (100 ηg/ml) and samples of LPS from each mini-pump were investigated for their ability to induce the release of nitric oxide (NO) from a BV-2 microglia culture. Microglia cells were plated 100,000 cells per well in a 96-well plate and incubated with media (DMEM, 10% FBS, 1% PenStep and 1% Glutamax) or 200 µl of LPS sample for 24 hrs. NO release due to LPS exposure was examined using the Greiss Assay kit (Invitrogen, Carlsbad., CA).

5.3.6 Statistics SigmaStat software was used to compare groups by two- and three-way ANOVAs with Fisher LSD as the preferred post-hoc and to perform Pearson correlation coefficients. Graphs are shown with SEMs represented by error bars. Control aCSF is shown in some graphs as one collapsed group, but aCSF groups were not collapsed for statistical analysis.

5.4 Results

5.4.1 Behavior: Spatial learning memory in the Morris water maze Spatial learning memory was assessed by testing the performance of animals infused with LPS or aCSF for 4 or 8 weeks in the Morris water maze. A 3-way repeated measures ANOVA on latency to find the hidden platform revealed a main between- subjects effect of infusion duration (F1, 60 = 5.976, p = 0.017), a main within-subjects effect of trial day (F3, 180 = 240.124, p < 0.001), an interaction between inflammation condition and infusion duration (F1, 60 = 9.780, p = 0.003), an interaction between inflammation condition and trial day (F3, 180 = 5.005, p = 0.002), and a 3-way interaction between inflammation condition, infusion duration and trial day (F3, 180 = 5.281, p = 0.002). Post hoc analysis indicates that LPS 4w have a greater trial duration than both aCSF-infused groups (*p ≤ 0.002) and LPS 8w (†p < 0.001). Average swim speed (Figure 40B) was analyzed by 3-way repeated measures

ANOVA. There were main between-subjects effects of inflammation condition (F1, 60 =

4.901, p = 0.031) and infusion duration (F1, 60 = 4.255, p = 0.043), a main within-subjects effect of trial day (F3, 180 = 42.171, p < 0.001), and an interaction between inflammation condition and trial day (F3, 180 = 2.179, p = 0.092). Post hoc analysis indicates that LPS 8w swam more quickly than both aCSF-infused groups (*p ≤ 0.041) and LPS 4w (†p = 0.024);

150 indicating that the increased latency of LPS 4w to find the platform was not due to reduced swim speed. Thigmotaxis, defined as the percentage of trial time spent swimming within 10 cm of the pool wall, was analyzed (Figure 40C). Thigmotaxis was analyzed by 3-way repeated measures ANOVA, which revealed a main within-subjects effect of trial day (F3, 180 = 165.176, p < 0.001), an interaction between inflammation condition and infusion duration

(F1, 60 = 3.987, p = 0.050), an interaction between infusion duration and trial day (F3, 180 = 3.929, p = 0.010), and a 3-way interaction between inflammation condition, infusion duration and trial day (F3, 180 = 3.112, p = 0.028). Post hoc analysis indicates that LPS 4w spend a greater percentage of trial time in the pool perimeter than aCSF 4w (*p = 0.038) and LPS 8w (†p = 0.012). After the last trial on the fourth day of water maze, the platform was removed and the amount of time that rats spent swimming within a 42 cm radius of the missing platform (25% of pool area) was recorded (Figure 41). ANOVA revealed a main effect of experimental group (F3, 64 = 600.82, p = 0.032) in which LPS 4w spent less time within the radius of the previous platform position than did aCSF 4w (*p = 0.012) and LPS 8w (†p = 0.016). These data, like latency to find the hidden platform, indicate that rats were impaired after 4 weeks of LPS infusion, but were no longer impaired after 8 weeks of LPS infusion. Changes in motor ability could be erroneously interpreted as memory deficit. To rule out this possibility, we observed rats for gross motor impairments in open field, and compared water maze performance with inflammation in brain regions that may affect motor behavior. LPS infused animals did not show any gross motor impairment in the open field, and poor water maze performance did not correlate with inflammation in the cerebellum as did hanging task performance (Chapter 3, Section 3.4). Based upon these data we infer that LPS infused animals were as capable swimmers as aCSF infused animals, and that the poor performance in the water maze reflects a cognitive deficit.

151

Figure 40. Morris water maze latency, swim speed and thigmotaxis In the Morris water maze, LPS 4w take longer to find the hidden platform (A) and spend more time in the perimeter of the pool (C) than their aCSF 4w (*) and LPS 8w (†) counterparts. LPS 8w swim faster than aCSF-infused groups (*) and LPS 4w (†).

152

Figure 41. Morris water maze probe trial results LPS 4w spent less time within the radius of the absent platform than did aCSF 4w (*p = 0.012) and LPS 8w (†p = 0.016).

5.4.2 Hippocampal histology: MHCII, SNAP25 and GLT1 Microglia expressing MHCII are observed in regions distal to the IVth ventricle such as the anterior hippocampus (Figure 42) and are not limited by proximity to the ventricular system. Likewise, regions near the ventricles were not equally replete with activated microglia. For example, the subiculum and DG of the hippocampus are relatively equidistant to the third ventricle, yet, subsequent to LPS infusion, the DG displays numerous MHCII+ microglia which are almost absent from the subiculum. This suggests that some brain regions have a greater propensity than others to induce microglia activation or attract activated microglia.

153

Figure 42. Regional distribution of activated microglia after LPS infusion MHCII+ microglia are distributed heterogeneously across the brain, and regions that tend to have the most include the DG and CA3 subregions of the hippocampus, the corpus collosum, the SN and brain stem regions. Image shows microglia stained for MHCII (brown) and the monoaminergic cells of the SN and LC stained with tyrosine hydroxylase (blue).

MHCII+ microglia were identified immunohistochemically (Figure 38) and manually counted within the CA1, CA3 and DG subregions on two dorsal hippocampal slices per animal (Figure 44). An ANOVA revealed main effects of inflammatory treatment (F1, 171 = 37.09, p < 0.001), duration of treatment (F2,171 = 7.34, p < 0.001) and hippocampal region (F2,171 = 12.91, p < 0.001) as well as an interaction between duration and inflammatory treatment (F2, 171 = 6.52, p = 0.002) and an interaction between hippocampal region and inflammatory treatment (F2, 171 = 10.96, p < 0.001). There are no significant differences in the number of MHCII+ microglia between subregions in aCSF groups or LPS 2w (p > 0.05). In contrast, LPS 4w and LPS 8w both express more MHCII+ microglia in CA3 and DG than the CA1 subregion (p < 0.001), and LPS 8w have more MHCII+ microglia in CA3 than DG (p = 0.017). No groups are significantly different within the CA1 subregion (p > 0.05). Within the DG and CA3, LPS 4w and LPS 8w both express more MHCII+ microglia than respective aCSF controls (*p < 0.001) and LPS 2w (†p ≤ 0.002). LPS 8w has more MHCII+ microglia than LPS 4w within the CA3 subregion (‡p < 0.001). While rats infused with LPS for 4 weeks performed poorly in the Morris water maze task, there is no correlation (p < 0.05) between the number of MHCII+ microglia in any of

154 the hippocampal sub-regions investigated with latency to find the hidden platform or perseverance in the region of the missing platform during the probe trial, and this holds true for rats infused with LPS for 8 weeks and aCSF-infused animals as well. SNAP25 is a synaptic docking protein, and levels of SNAP25 may reflect the docking and release of vesicles containing glutamate. SNAP25 was quantified as density of immunostaining above a determined threshold per area (Figure 45). ANOVA revealed main effects of inflammation group (F1, 110 = 14.56, p < 0.001), duration of infusion (F2, 110

= 14.34, p < 0.001) and hippocampal region (F2, 110 = 83.74, p < 0.001) as well as an interaction between inflammation group and duration of infusion (F2, 110 = 10.77, p < 0.001). SNAP25 staining is more dense in the CA3 of LPS 4w than aCSF 4w (*p = 0.014). LPS 8w had a greater staining density than aCSF 8w in all three subregions (*p ≤ 0.036), than both LPS 2w and LPS 4w in the CA1 and CA3 (‡p < 0.001) and greater than LPS 4w in DG (‡p = 0.017). GLT1 staining density was determined within hippocampal subregions (Figure 46).

There was a main effect of inflammation group (F1, 121 = 6.72, p = 0.011). LPS 8w, which improved in the water maze compared to LPS 4w, had significantly more GLT1 in the CA3 than controls (*p = 0.039). Neither SNAP25 nor GLT1 correlated with WM performance on the final day of testing when compared within all samples. However, when only rats infused for 8w were analyzed, SNAP25 in CA1 correlated with impaired performance (latency: r = 0.805, p = 0.044; distance: r = 0.859, p = 0.003). SNAP25 in CA3 correlated with fewer entrances in the vicinity of the missing platform during the probe trial (r = -1, p = 0.005) and GLT1 in CA3 correlated positively (r = 0.938, p = 0.005) with more entrances. Unfortunately, due to exhaustion of tissue samples, SNAP25 and GLT1 could be stained only in a non- overlapping subset of subjects, prohibiting correlations between these two markers.

Figure 43. Distribution of MHCII+ microglia in the hippocampus Images show MHCII+ microglia (brown) and nuclei counterstained with cresyl violet at 10X (a-d) and 40X (a1-d1). 155

Figure 43.

156

Figure 44. Quantification of hippocampal MHCII+ microglia Number of cells per hippocampal subregion were evaluated (A) and expressed in LPS- infused rats relative to controls (B). LPS infusion increased the density of MHCII+ microglia in hippocampal subregions, CA3 > DG > CA1 in LPS 4w and LPS 8w. Within the DG and CA3, LPS 4w and LPS 8w had significantly more MHCII+ microglia than aCSF controls (*p < 0.001) and LPS 2w (†p ≤ 0.002). In the CA3, LPS 8w rats had significantly more MHCII+ microglia than LPS 4w rats (‡p < 0.001).

157

Figure 45. Distribution and quantification of hippocampal SNAP25 Hippocampi were stained for SNAP25 (A). Staining density was evaluated (B) and expressed in LPS-infused rats relative to controls (C). SNAP25 staining was more intense in LPS 8w than aCSF 8w (*p ≤ 0.036) and LPS 4w (‡p ≤ 0.017). For all groups CA3 > DG > CA1. SNAP25 staining intensified in the CA3 region of LPS 4w as compared to aCSF controls (*p = 0.014). LPS 8w expressed more SNAP25 than aCSF 8w in all regions (*p ≤ 0.036) and compared to LPS 4w in CA1 and CA3 (‡p < 0.001).

158

Figure 46. Distribution and quantification of hippocampal GLT1 GLT1 staining (A) density was evaluated (B) and expressed in LPS-infused rats relative to controls (C). GLT1 staining was more intense in LPS 4w and LPS 8w than aCSF controls (*p = 0.039).

159

5.4.3 Molecular and chemical analyses by Western blot None of the proteins evaluated within the DG/CA3 region of the hippocampus changed significantly with LPS infusion (p < 0.05). These include calbindin, fractalkine ligand, fractalkine ligand soluble form, fractalkine receptor, SNAP25, GFAP, NMDAR1, pan-NR1, phospso-NR1, NMDAR2A, NMDAR2B, HSP 70, HSP 90, HMGB1 and OxPhos IV.

5.4.4 Nitric oxide release from LPS-stimulated BV-2 microglia cell culture LPS stored in an osmotic mini-pump for 4, 6 or 8 weeks at 37º C in 0.9% saline bath elicited a significantly (F4,30 = 161.2, §p < 0.001) elevated and equivalent release of NO from a BV-2 microglia cell line as freshly prepared LPS solution (Figure 47). These results confirmed the potency of LPS stored within an osmotic minipumps at body temperature over 8 weeks, and also confirmed that the fill volume of a model #2006 allows for release over 8 weeks.

Figure 47. LPS maintains potency after 8 weeks incubation in osmotic minipump Cultured BV-2 microglia cells responded to fresh LPS as well as LPS that had been incubated in an Alzet osmotic minipump for 4, 6 or 8 weeks with the release of similar levels of NO, elevated compared to un-stimulated cells (§p < 0.001).

5.5 Discussion Chronic neuroinflammation is associated with many neurodegenerative diseases and contributes to disease onset and progression. In order to better understand the time- dependent relationship between chronic neuroinflammation and the etiology of AD, we used an animal model of chronic neuroinflammation created by the infusion of LPS into 160 the IVth ventricle for 3 durations (2, 4 and 8 weeks) and investigated effects in the hippocampus as they relate to AD-like memory impairment.

5.5.1 Microglia ‘activation’ and hippocampal vulnerability LPS directly activates toll-like receptor 4 (TLR4) on microglia, and elicits the expression of MHCII. Increasing numbers of microglia express MHCII over the 8 week LPS infusion period. Although LPS is lipid soluble and widely dispersed, not all microglia are activated simultaneously. This suggests that while microglia activation is progressively concentrated within certain brain regions over time (Figure 42), microglia activation within those regions is random and dependent upon exposure to LPS and the pro-inflammatory microenvironment. This pattern of microglial activation in the presence of LPS may recapitulate an important aspect of the AD brain, i.e. activation of individual microglia in vulnerable regions over time due to the accumulation of aberrant protein, such as amyloid ß or hyperphosphorylated tau, and presence of degenerating neurons. The distribution of activated microglia in response to chronic i.c.v. LPS infusion corresponded with brain regions that show the most intense pathology and degeneration in AD (McGeer et al., 1988, Sheng et al., 1998, Sheffield et al., 2000, Szpak et al., 2001, Imamura et al., 2003, Xiang et al., 2006, Edison et al., 2008a). The distribution of MHCII+ cells cannot be fully explained by vicinity to the ventricles where LPS enters. The subiculum and DG of the rat hippocampus are both relatively equidistant from the third ventricle, yet there is little to no MHCII+ microglia in the subiculum and a high level in the DG upon LPS infusion. A differential regional distribution of MHCII+ microglia is also seen in areas distal from the ventricles, although lipid-soluble LPS is likely to pass through most of the parenchyma. Brain regions such as the entorhinal cortex show many MHCII+ microglia, yet some layers of cortex show very few. The regional distribution of activated microglia is similar in the brains of mice injected with LPS into the cisterna magna and later analyzed by autoradiography for expression of the peripheral benzodiazepine receptor (Biegon et al., 2002). These data support the hypothesis that the propensity to develop neuroinflammation is more pronounced in regions vulnerable to degeneration in neurodegenerative disease. The hippocampal atrophy in AD is likely related to the manifestation of cognitive impairment and memory loss. Normal aging is associated with increased microglia 161 activation in the hippocampus, that is exaggerated in AD, and seen before other disease pathology develops (Cagnin et al., 2001). In rats, normal aging and chronic LPS-infusion are characterized by microglia activation in the hippocampus and impairment in the hippocampal-dependent Morris water maze task (Hauss-Wegrzyniak et al., 1998a, Marchalant et al., 2008; Hauss-Wegrzyniak et al., 1998a, Hauss-Wegrzyniak et al., 2000b). In this study, the number of MHCII+ microglia increases in the CA3 and DG of the hippocampus after both 4 and 8 weeks of continuous LPS infusion, and continues to increase significantly in the CA3 between 4 and 8 weeks. However, cognitive impairment observed in the Morris water maze and associated with microglia activation after 4 weeks of LPS infusion is attenuated after 8 weeks of LPS infusion, despite the increased number of ‘activated’ microglia. Microglial expression of MHCII remains elevated in the hippocampus despite the recovery of spatial memory, and there are several explanations for this phenomenon. It is possible that early responses to LPS initiate processes that are detrimental to cognition, but that later behavior of activated microglia is supportive to neurons and that MHCII is simply expressed residually on microglia which are no longer in an active, pro-inflammatory state. For example, the initial M1-phenotype may drive cognitive impairment while a later M2- phenotype may support neuronal function and repair (Wilcock, Colton, Landreth, Wyss- Coray). We have previously shown that after a recovery period of 5 weeks following the infusion of LPS for 5 weeks, the number of microglia expressing MHCII was not changed, although the immediate effects of LPS would have been absent (Hauss-Wegrzyniak et al., 2000c); suggesting either that MHCII remains on the cell surface long after exposure to LPS has been discontinued or that there is a continued, self-perpetuating inflammatory response that drives MHCII expression weeks after LPS infusion is ceased. MHCII is an indicator of a pro-inflammatory, M1-type microglial activation state, but because it is possible that MHCII surface expression may last beyond the transition from an M1 to an M2 phenotype, MHCII alone is not sufficient to confidently determine the nature of the microglial activation state. In another study we treated young rats with LPS for either 3 or 8 weeks and evaluated cytokine expression in the hippocampus (Bardou et al., 2013). LPS infusion did not induce changes in mRNA expression levels of the LPS receptor TLR4, pro-inflammatory TNFα or anti-inflammatory CX3CR1, CX3CL1 or

162

TGFß. Likewise, LPS infusion did not produce changes in protein levels of the pro- inflammatory cytokines IL-1α, TNFα, IL-6, IL-12, IFNγ, IL-2, Il-5 or anti-inflammatory IL-4 or IL-10. However, after 3 and 8 weeks of LPS infusion, there was a significant increase in IL-1ß mRNA and protein levels. Il-1ß is associated with an M1, pro- inflammatory microglia activation state and memory impairment (Wilcock, Colton, Griffin). Microglia may become less responsive to LPS infused over time due to a well- known phenomenon called endotoxin tolerance (Broad et al., 2006). If microglia simply react less to LPS exposure at the end of the 8 weeks LPS infusion, then this could explain the restoration of successful performance in the Morris water maze after 8 weeks LPS infusion, although it would not explain why memantine treatment was detrimental to performance in this group. Therefore, it is important that we take into consideration evidence of the presence or absence of endotoxin tolerance, before making conclusions about the degree to which other observed changes may be compensatory. Endotoxin tolerance has been studied mostly in macrophages and less in microglia, but it is likely to occur in microglia given that they are of the same cellular lineage as macrophages; the mechanism of TLR activation by LPS and induction of an immune- activating cascade are largely the same (see Figure 11). In vitro, primary microglia cultures stimulated with LPS over 1, 2 or 3 days expressed different functional states. Expression of pro-inflammatory TNFα, NO and prostaglandins was enhanced after the first LPS exposure, but only prostaglandin E2 levels were elevated by subsequent LPS exposures (Ajmone-Cat et al., 2003). Similar tolerance to repeated i.p. injections of LPS by glia in the spinal cord manifests as attenuation to LPS-induced hyperalgesia (Guo and Schluesener, 2006). Repeated peripheral injections of LPS in rats lead to endotoxin tolerance measured as reduced fever response, and this tolerance is attenuated by i.c.v. administration of a NOS inhibitor, indicating that NO, which is released by activated microglia, is important in a feedback mechanism that regulates continued response to endotoxin (Almeida et al., 1999). Endotoxin tolerance may endow an evolutionary benefit such that organisms exposed to an infection do not reach a level of immune activation that is in itself detrimental (Beurel and Jope, 2010). For example, cytokines alone have been shown to interfere with

163 memory function (He et al., 2007, Alkam et al., 2008, Moore et al., 2009, Taepavarapruk and Song, 2009) and possibly kill cells (Gromkowski et al., 1990, Kessler et al., 1993). Attenuation of the immune response to endotoxin may be a valuable regulation. One of the most-likely candidate mechanisms for endotoxin tolerance is the sequestration of LPS by increased levels of LPS binding protein in the blood; we did not examine this in our study. Other possible mechanisms include changes in the TLR4 receptor complex, the regulation of the intracellular cascade, decreased production of transcription factors and decreased production of pro-inflammatory cytokines in response to LPS binding. We have explored some of these possibilities, including expression of TLR4, MAPKs and cytokines and found a lack of overall changes as dynamic as those observed in an early immune response to a high, acute challenge, but found some prolonged changes indicative of an active immune response in the hippocampus. TLR4 mRNA expression is not different from aCSF control after 3 or 8 weeks of continuous LPS infusion in the hippocampus, but is significantly elevated after both 3 and 8 weeks of LPS infusion in the brainstem (Bardou et al., 2013). The MAPK p38 is upregulated in microglia cultures upon LPS exposure but the increase in expression is reduced following one or two previous exposures to LPS (Ajmone-Cat et al., 2003). The MAPK p38 is not changed after 2 weeks of LPS infusion, but the phosphorylated MAPK ERK is upregulated (Figure 24). We do not know, however, if phosphorylated ERK levels are still increased after 4 or 8 weeks of chronic LPS infusion. As described above, IL-1ß mRNA and protein remain at the same elevated level after both 3 and 8 weeks of LPS infusion (Bardou et al., 2013). Not much is known to characterize the response to low-level, continuous i.c.v. LPS infusion earlier than 3 weeks, leaving the possibility that a dynamic response of a high magnitude occurs within a time-frame that we have not directly observed. We postulate that individual microglia become activated upon LPS exposure at different time points throughout the 8 week infusion period and gradually become tolerant throughout that period at a similar pace. Therefore, a chronic neuroinflammatory environment is established not by a population of microglia that remain active, but by the activation of additional microglia over time. This perspective is consistent with our data indicating that there is an on-going pro-inflammatory immune response to chronic, low- level LPS. If the inflammatory response to LPS is sufficient to drive cognitive impairment,

164 but cognition is restored in the continued presence of a pro-inflammatory environment, then compensations in addition to regulation of the immune response must occur. An active immune response is present in AD and contributes to pathology, but there have been no successful clinical trials of immuno-therapy. Therefore, we are interested in understanding any compensatory mechanisms that restore cognition within an neuroinflammatory environment, because they would be viable targets for AD treatment.

5.5.2 Increased glutamate uptake may reestablish cognition A analogous effect to endotoxin tolerance occurs in young hippocampal cultures exposed to glutamate, NMDA or AMPA, in which initial excitotoxin exposure protects mature hippocampal cultures from excitotoxicity induced by a second exposure (Friedman and Segal, 2009). Our previous studies using chronic i.c.v. LPS infusion and pharmacological manipulation indicate that glutamate is elevated under conditions of chronic neuroinflammation and is involved in cognitive impairment. If neurons become less vulnerable to increased levels of extracellular glutamate over time, then this could account for the rebound in memory performance seen after 8 weeks of LPS infusion. In this study, NMDAR subunit expression was not affected by any duration of LPS infusion, although previous investigations have found a reduced number of NMDAR1 in the hippocampus after 4 weeks of LPS infusion. Increased glutamate clearance through the glutamate transporter GLT1 is a potential mechanism of tolerance to inflammation-induced elevation in extracellular glutamate, and we explored the relationship between GLT1 and attenuation of LPS-driven cognitive impairment. Chronic neuroinflammation may be associated with an increased availability of extracellular glutamate. Following four weeks of LPS infusion, rats demonstrated impaired accuracy of neural encoding, information processing, LTP and spatial memory that were restored by pharmacologically antagonizing glutamatergic-NMDA receptors with the partial un-competitive antagonist memantine (Rosi et al., 2005, 2006, 2009). Furthermore, a reduction of glutamate release by stimulation of endocannabinoid receptors or blockade of adenosine receptors by caffeine administration prevent LPS-induced microglia activation and spatial memory deficits (Brothers et al., 2010; Marchalant et al., 2007). The prevention of LPS-induced microglia activation and behavioral deficit by pharmacological reduction of glutamate activity suggests that neuroinflammation leads to aberrant 165 glutamate activity. The common underlying mechanism of memantine, cannabinoids and caffeine is their ability to dampen glutamatergic signaling, and this may be sufficient to prevent spatial memory impairment driven by chronic neuroinflammation. If spatial memory deficits due to LPS-infusion can be protected by the NMDA- antagonist memantine, then perhaps the mechanism by which animals recover memory function after 8 weeks LPS infusion is also due to a mechanism involving NMDARs. Down-regulation of NMDARs or reduced binding at NMDARs would reduce Ca2+ entry in response to elevated extracellular glutamate and could potentially restore memory, similar to blocking Ca2+ entry through NMDARs with memantine. Expression and binding of the obligatory NMDAR subunit NR1 are reduced in the hippocampus after chronic infusion of LPS for 4 weeks (Rosi et al., 2004). Neurons that have NMDARs are susceptible to glutamate excitotoxicity, and reduction of NR1 expression and binding may reflect a reduced number of neurons. However, reduced NR1 subunit expression and function may also be a mechanism by which neurons are protected from or become tolerant to excitotoxicity. Therefore, we investigated the possibility that compensatory changes in the expression of NMDARs may correlate with memory improvement after 8 weeks of chronic LPS infusion. We evaluated expression of NMDAR subunits by Western blot in DG/CA3 region of the hippocampus which also expresses the highest numbers of MHCII+ microglia. NMDAR1, pan-NR1, and phospso-NR1 were analyzed as well as NMDAR2A, NMDAR2B. No changes were observed between aCSF and LPS infused rats or between infusion periods of 2, 4 or 8 weeks in and of the NMDAR subunits. These data indicate that either our Western blots were not sensitive enough to detect the changes observed in (Rosi et al., 2004) or the rebound in spatial working memory seen with 8 weeks of LPS infusion may not be dependent upon changes in NMDAR expression in this study. Although we do not observe changes in NMDAR subunit expression, we do see changes in glutamate regulation, namely the increase in SNAP25 and GLT1, in young rats exposed to LPS. SNAP25 is part of the SNARE complex that allows exocytosis of neurotransmitter from vesicles and is implicated learning and memory. SNAP25 expression is reduced in elderly individuals with AD and Down syndrome, who also have AD-like pathology and cognitive impairment (Greber et al., 1999; Downes et al., 2008). SNAP25 expression

166 progressively increased between 4 and 8 weeks of LPS infusion and correlated with poor performance in the water maze probe trial; this can be interpreted in two ways. One interpretation is that increased SNAP25 in the glutamatergic terminals that abundantly populate the hippocampus may indicate enhanced glutamate release, a property that is consistent with the hypothesis that chronic neuroinflammation is associated with elevated extracellular glutamate. The increase SNAP25 after 4 weeks LPS precedes the increase in GLT1 by 8 weeks, and this timeline is consistent with the interpretation that increased SNAP25 under inflammatory conditions is representative of increased glutamate release and that elevated GLT1 is associated with a compensatory increase in clearance of extracellular glutamate that helps to restore glutamate balance and cognitive performance. Conversely, SNAP25 negatively regulates native voltage-gated calcium channels in glutamatergic neurons at glutamatergic terminals (Condliffe et al., 2010), suggesting that increased SNAP25 above basal levels may lead to decreased depolarization and neurotransmitter release from the pre-synaptic terminal. Therefore, the increased expression of SNAP25 may be a compensatory response to the pro-inflammatory environment that decreases glutamate exocytosis from glutamatergic neurons. This later interpretation is consistent with the blunted electrophysiological response of DG granule cells and failure to elicit long-term potentiation (LTP) after tetanic stimulation in vivo observed after 5 weeks of chronic LPS infusion (Hauss-Wegrzyniak et al, 2002), and also consistent with the negative correlation between SNAP25 expression and cognitive performance. GLT1 also increased in the hippocampus after 8 weeks of LPS infusion, but unlike SNAP25, GLT1 positively correlated with improved performance in the Morris water maze. Consistent with our finding, GLT1 in mice with no experimental manipulations correlated positively with memory performance in a T-maze task (Heo et al. 2011). GLT1 is the principle excitatory amino acid transporter responsible for approximately 80-90% of glutamate clearance from the extracellular space in hippocampal tissue (Maragakis et al. 2004; Selkirk et al. 2005). Reduction of glutamate transport and clearance, as well as variants of the GLT1 gene have been associated with aging and AD (Lauderback et al. 2001a; Massie et al. 2010; Rodriguez-Kern 2003; Montori et al. 2010; Dabir et al. 2006; Matos et al. 2008; Minkeviciene et al. 2008; Trotti et al. 1998; Woltjer et al. 2010; Hermans

167

2007; Pow & Cook 2009; Beart & O’Shea 2007). Increased GLT1 expression in the CA3 and DG of the hippocampus after 8 weeks of LPS infusion indicates that glutamate clearance may be enhanced. Increased glutamate clearance may compensate against the effects of chronic neuroinflammation and improve spatial memory by reducing synaptic ‘noise’ from elevated extracellular glutamate. Consistent with this interpretation, a compensatory increase in glutamate clearance rate is also seen in aged rats after elevated evoked-glutamate release in middle age (Stephens et al., 2009), and in a genetic model of consistent-low level lifetime elevated glutamate release (Bao et al., 2010, Michaelis et al., 2011). If an increase in glutamate transport occurred by 8 weeks of LPS infusion that promoted behavioral recovery by reducing the availability of extracellular glutamate and returning to glutamatergic homeostasis, then it is possible that administration memantine in an environment of increased glutamate clearance would drive a ‘hypoglutamatergic’ state and related memory impairment; consistent with the results observed in Chapter 4. These data support the possibility that increased glutamate clearance is a compensation the brain makes in the presence of a neuroinflammatory environment, like that of pre-clinical AD, that prolongs the period of cognitive ability before the manifestation of memory symptoms. These data further promote pursuing drugs that increase glutamate clearance for therapeutic use in AD.

5.5.3 Conclusions Following chronic infusion of LPS into the IVth ventricle, the regional distribution of MHCII+ microglia was heterogeneous and concentrated in the DG and CA3 regions of the hippocampus. Inflammation was associated with impairment in spatial memory after 4 weeks that is not sustained by continued LPS infusion despite increased number of MHCII+ microglia. Our results are consistent with the hypothesis that compensatory changes related to synaptic regulation of glutamate might underlie the recovery of spatial working memory in the continued presence of LPS and increased number of MHCII+ microglia. The unique mechanisms discovered in this study by which the young brain compensates against chronic exposure to an inflammatory environment, namely an increase of GLT1, shed light on the interaction between neuroinflammation, glutamatergic regulation and compensatory mechanisms that may prolong the period of health before the manifestation 168 of pathology and clinical symptoms in neurodegenerative disease. Taken together, these data indicate that naturally occurring changes that decrease the action of glutamate within the hippocampus of the young adult brain in an environment of persistent, low-level neuroinflammation represent a process of compensation that proceeds and persists throughout behavioral recovery in the young brain. These observations are consistent with the hypothesis that compensatory biochemical processes within the young brain are able to delay the clinical symptoms of degenerative diseases; with advanced age these processes may fail to provide the necessary compensation of function or respond positively to anti- inflammatory therapy (Hauss-Wegrzyniak et al., 1999). Therefore, if we can pharmacologically drive an increase in glutamate clearance, we may be able to facilitate a compensatory response that would be sufficient to attenuate inflammation-induced cognitive impairment and potentially be efficacious against the manifestation of clinical symptoms of AD; this is precisely what I endeavor to do in the next Chapter.

169

Chapter 6: Manipulation of glutamate transporters by Ceftriaxone and Riluzole in in aging and a model of neuroinflammation

6.1 Brief rationale I previously established that pre- and post-synaptic pharmacological interventions that decrease glutamate function (caffeine [Chapter 2, Brothers et al., 2010], cannabinoids [not included in this thesis; (Marchalant et al. 2008; Marchalant et al. 2009; Marchalant et al. 2009b; Bardou et al., 2012)] and memantine [Chapter 4, (Rosi et al. 2006; Rosi et al. 2009)] are protective against cognitive deficits driven by experimentally-induced neuroinflammation. I also demonstrated that an increase in glutamate transport may be one mechanism by which young animals compensate for the deleterious effects of chronic neuroinflammation on cognitive performance (Chapter 5). This chapter investigates whether a pharmacologically-induced increase in glutamate transport is sufficient to prevent/reverse neuroinflammation- or age-related spatial memory impairment. Furthermore, in this study I systematically characterize the activation state of microglia in our LPS model and in aged rats, because microglia activation states cannot reliably be inferred from histological analysis of MHCII alone. To do so, I investigated a number of markers (both mRNA and protein) that are used to distinguish between microglia activation states. This study is confluent with my previous work and fills a gap in knowledge by extending our investigations to the tripartite synapse. This work moves the field forward in understanding of the interaction between neuroinflammation and glutamate signaling, and by focusing on relevant pharmacologic interventions with drugs that are currently used in humans.

6.2 Introduction Neuroinflammation is a characteristic of Alzheimer’s disease (AD) that may promote neurodegeneration and cognitive decline through changes in glutamate regulation; it is our goal to understand this relationship. Chronic neuroinflammation involves

170 prolonged microglia activation and sustained release of immunomodulatory factors. Although immune activation in response to infection or injury is generally beneficial, protracted neuroinflammation may be detrimental and contribute to pathology and memory decline in natural aging and neurodegenerative disease. Neuroinflammation may contribute to AD brain pathology and cognitive symptoms through dysregulation of the glutamatergic system. Elevation of extracellular glutamate may cause memory impairment and ultimately lead to neuronal loss through energetic failure and excitotoxicity (Ankarcrona et al. 1995). Indirect evidence from our previous studies suggests that extracellular glutamate is elevated in the hippocampus under conditions of chronic neuroinflammation. Chronic neuroinflammation generated by the continuous intracerebroventricular (i.c.v.) infusion of the endotoxin lipopolysaccharide (LPS) over 4 weeks produces spatial memory impairment in the Morris water maze, a task dependent upon glutamatergic function in the hippocampus. Compounds that reduce glutamatergic activity in the synapse, despite different mechanisms of action, ameliorate spatial memory deficit and attenuate microglia activation in this model. These include two compounds that reduce glutamate release, caffeine and the endocannabinoid WIN55212-2 (Brothers et al. 2010; Marchalant et al. 2008), as well as the glutamatergic N-methyl-D- aspartic acid receptor (NMDAR) non-competitive antagonist memantine (Rosi et al., 2006). These studies support the conclusion that attenuating glutamatergic signaling is sufficient to prevent cognitive deficit induced by chronic neuroinflammation. Glutamate signaling is terminated primarily by sequestration of glutamate into astrocytes by transporters, and GLT1 (excitatory amino acid transporter 2, EAAT2; solute carrier family 1 member 2, SLC1A2) is the principle glutamate transporter within the hippocampus (Maragakis et al. 2004; Selkirk et al. 2005). We recently discovered that inflammation-induced spatial memory deficit is transient; performance is impaired after 4 but not 8 weeks of continuous LPS infusion, and elevated GLT1 expression correlates with the resolution of spatial memory impairment (Chapter 5). We interpret increased GLT1 expression to be a compensatory response to inflammation-induced extracellular glutamate elevation and to be directly related to cognitive recovery. Therefore we developed the following hypothesis: if experimentally induced or age-associated chronic neuroinflammation increases extracellular glutamate in the hippocampus and contributes

171 to spatial memory impairment, then the activation state of microglia will be related to changes in glutamate regulation and degree of memory impairment, and pharmacological augmentation of glutamate transport will attenuate spatial memory deficits related to chronic neuroinflammation. To address this hypothesis, we first assessed the microglia activation state induced by continuous i.c.v. LPS infusion over 4 weeks and in aging by investigating changes in mRNA and protein expression inflammation markers (Table 6, Appendix B: Microglia activation states and immune factors, and Section 1.2.1). Next, we tested the prediction that elevated GLT1 function is sufficient to attenuate spatial memory deficit induced by chronic neuroinflammation in LPS-infused young rats and impairment observed in aged rats by treating with the drugs Ceftriaxone (Cef, Rocephin) and Riluzole (Ril, Rilutek) to pharmacologically elevate GLT1. We chose these drugs because previous studies have demonstrated that they increase glutamate clearance, and we chose doses consistent with studies that reported beneficial effects in the CNS or in cognitive performance (see Appendix E and Appendix F). Finally, we evaluated the relationship between microglia activation state, glutamate transport and spatial memory.

6.3 Methods

6.3.1 Experimental groups Young (3 mo.) and aged (22 mo.) Fisher F-344 (Harlan and N.I.A., respectively) were housed in ventilated Plexiglas cages in a temperature-controlled colony room on a 12/12 reverse light/dark cycle (lights off at 09:00) with ad libitum access to rat chow and water. All rats were housed for one week before any manipulations. Young rats were divided into two groups, one that received LPS-induced neuroinflammation and artificial cerebral spinal fluid (aCSF) controls. Young rats and aged rats were treated with either Cef, Ril or vehicle control, creating 9 experimental groups: aCSF (n = 7), aCSF Cef (n = 6), aCSF Ril (n = 8), LPS (n = 7), LPS Cef (n = 7), LPS Ril (n = 8), Aged (n = 8), Aged Cef (n = 9) and Aged Ril (n = 10). All procedures were in compliance with the ethical standards of the IACUC.

172

6.3.2 Chronic neuroinflammation via continuous i.c.v. LPS Neuroinflammation was induced in young rats by continuous intracerebroventricular (i.c.v.) infusion over 3 weeks of LPS (0.25 µg/hr., 1 mg/ml dissolved in aCSF; E. coli, serotype 055:B5, TCA extraction; Sigma, St. Louis, MO), and controls were infused with aCSF (140 mM NaCl, 3.0 mM KCl, 2.5 mM CaCl2, 1.0 mM

MgCl2, and 1.2 mM Na2HPO4 adjusted to pH 7.4). A three-week infusion period was chosen to coordinate with the three-week drug infusion period. A small sub-group was infused over 4 weeks and not treated with drugs, for an experiment using flow cytometry. A cannula was implanted into the IVth ventricle (-2.5AP and -7DV relative to lambda) and attached (via Tygon tubing, 0.06 O.D.) to an osmotic minipump (0.25µl/hr., Alzet model #2004; Durect Corp., Cupertino, CA) as previously described (Hauss-Wegrzyniak et al. 1998). Post‐operative care included application of the topical anesthetics lidocaine (1%) and Neosporin to the incision, and sub-cutaneous injection of 2 ml of isotonic saline to prevent dehydration during recovery from surgery. Rats were given Tylenol in their drinking water to lessen pain beginning 3 days prior and ending 3 days after surgery .

6.3.3 Drug treatment to increase glutamate clearance: Ceftriaxone and Riluzole Treatment with Cef, Ril or vehicle began the day after surgery in young rats, or one week after arrival in aged rats, and continued daily for 3 weeks. Both drugs were prepared fresh daily and delivered by intraperitoneal (i.p.) injection at approximately 15:00. Cef (Sigma) 200 mg/kg/day i.p. was dissolved in ddH2O at 200 mg/ml. Because this solution reaches 302 mOsM, 0.9% saline (308 mOsm) was used for the vehicle control. Riluzole (Selleck Chemicals) 4 mg/kg/day i.p. was dissolved at 8 mg/ml in 50% PEG and the vehicle, 50% PEG, was used for controls. The drug vehicles 0.9% saline and 50% PEG were split evenly between aCSF, LPS and Aged controls. This was taken into account during analysis of each experimental measure before collapsing between vehicle controls. Drug preparation, dose, and delivery were chosen to be consistent with studies that reported positive effects on cognition in animal models of diseases with a neuroinflammatory component and based upon personal communication with some of the authors (see Appendix E, Appendix F and Section 1.5.3b).

173

6.3.4 Spatial learning and memory: Morris water maze Spatial learning and memory was tested in the Morris water maze, a task sensitive to both hippocampal damage and LPS-induced neuroinflammation (Hauss-Wegrzyniak et al., 1998a). Rats were tested over the last 4 days of the 3 week infusion/drug treatment period. Testing was initiated at 09:00 and completed before 15:00, so that animals were tested during their waking period and not subject to acute drug effects. Spatial memory was evaluated as ability to find a hidden escape platform (10 cm diameter) that remained in a constant location, submerged (2.5 cm below surface) within a dark grey water pool (85 cm radius) maintained at 20-25 ºC. The pool had a large black square above the rim as a proximal cue, and was centered in room with highly contrasted visual stimuli on the wall and a threshold as distal cues. Latency to find the platform, distance traveled, speed, thigmotaxis (time spent in the pool perimeter) and other variables were tracked and recorded (Noldus EthoVision 3.1, Noldus, Leesburg, VA). Rats were tested 6 trials per day over 4 days, with approximately 1 hour between trials. Immediately before the first trial on the first day, rats were placed on the hidden platform for 30 sec. All trials were 60 sec. or ended when the rat found the hidden platform. The rat remained or was placed on the platform for an additional 30 sec. Animals were placed into the pool from 6 evenly spaced entry points in a non-adjacent order that was consistent for all animals and changed each day. On the 4th day animals completed a probe (7th) trial, in which the platform was removed and the time that they spent within the vicinity (42 cm diameter, 25% of pool area) of the original platform location was analyzed. Finally, to control for possible deficits in visual acuity, rats were tested on 2 visible platform trials in which the platform was raised 2 cm above the water surface, moved to a new location, and marked with a white latex glove.

6.3.5 Biochemical analysis Animals were lightly anesthetized with isoflurane and rapidly decapitated. Brain regions were quickly subdissected on ice. One hippocampi from each rat was stored in - 80 ºC and later homogenized for RNA extraction and the other hippocampi was immediately homogenized and stored for later protein analysis with a BioPlex. Hippocampi from both hemispheres of a small separate group of rats infused with aCSF or

174

LPS over 4 weeks were homogenized immediately for flow cytometry. Markers of microglia activation state were investigated according to Table 6.

Method M1 M2a M2c Resting IL-1ß, TNFα, CX3CL1, rtPCR TLR4, GLT1, TGFß, BDNF CX3CR1 xCT IL-1ß , TNFα, IL-4, IL-13 IL-10 CX3CR1 BioPlex IFNγ, IL-6, IL- 12, IL-2, IL-5 GM-CSF Flow TLR4, MHCII, CX3CR1

cytometry GLT1 Table 6. Microglia activation state markers examined These markers are listed in categories that approximate their relationship to microglia activation state according to (Colton & Wilcock 2010b; Colton 2009; Cameron & Landreth 2010; Lucin & Wyss-Coray 2009b).

6.3.5a rtPCR mRNA analysis Gene expression in hippocampal samples was determined by quantitative reverse- transcription polymerase chain reaction (qrtPCR). The following genes were analyzed in sample triplicates: TLR4 (the receptor for LPS); the pro-inflammatory cytokines IL-1ß, TGFß and TNFα; the anti-inflammatory cytokine CX3CL1 and its receptor CX3CR1; brain-derived neurotrophic factor (BDNF); the glutamate transporter GLT1 and the glutamate-cystine anti-porter xCT; and the housekeeping gene glyceraldehyde 3-phosphate dehydrogenase (GAPDH). Briefly, tissues were homogenized PureZol RNA Isolation Reagent (BioRad) and NucleoSpin RNA II (Machery-Nagel, Allentown, PA). RNA was quantified using (Synergy HT, BioTek) and 1 µg RNA per sample was reverse transcribed (iScript Reverse Transcription Supermix, Bio-Rad) to create a complementary DNA (cDNA) library. Each sample was plated in triplicates with mastermix (SsoAdvanced SYBR Green Supermix, BioRad) with forward and reverse primers to identify specific target genes (primers from Integrated DNA Technologies; Table 7). Primers were validated and optimized by using a melting curve (65 ºC and progressing to 95 ºC in increments of 0.5 ºC for 5 sec.). The reproducibility and efficiency for each primer was

175 accounted for during gene analysis. The progression of rtPCR cycles (Bio-Rad, model CFX96, C1000 Thermal Cycler) began with 95 ºC x 30 sec., followed by 40 repeated sets of 95 ºC x 5 sec. and 60 ºC x 30 sec., and terminated with a melting curve. The cycle (Ct) at which expression levels crossed threshold were normalized to the Ct of the housekeeper GAPDH, producing ∆Ct with arbitrary units of total gene expression. All plates were evaluated with respect to –RT and H2O controls.

Primer Forward (5’-3’) Sequence Reverse (3’-5’) Sequence Accession # TLR4 TCTGCCCTGCCACCATTTACAGTT TGGTCTCAGGCAGGAAAGGAACAA NM_019178 IL-1ß ACCTGCTAGTGTGTGATGTTCCCA AGGTGGAGAGCTTTCAGCTCACAT NM_031512 TGFß TGATACGCCTGAGTGGCTGTCTTT TTTGCTGTCACAAGAGCAGTGAGC NM_021578 TNFα CTGGCCAATGGCATGGATCTCAAA AGCCTTGTCCCTTGAAGAGAACCT X66539.1 CX3CL1 ACTTCTGTGCTGACCCAAAGGAGA CACGCTTCTCAAACTTGCCACCAT NM_134455.1 CX3CR1 GTGCAAGCTCACGACTGCTTTCTT GTGTTGCACTGTCCGGTTGTTCAT NM_133534.1 BDNF AGAGCTTTGTGTGGACCCTGAGTT TGGACGTTTGCTTCTTTCATGGGC BC087634 GLT1 TCTTGCCAGCTTCCTGTTGTCTCA ACACCTTGTGTGGCTTGGTGTTTC AY_069978 xCT TGTATGACTGGGAAACCACAGCGA TACAGAGAAGCAGCTGGAAGCACA NM_001107673 GAPDH TGACTCTACCCACGGCAAGTTCAA ACGACATACTCAGCACCAGCATCA NM_017008 Table 7. PCR Primers

6.3.5b Protein analysis: Bradford Protein content of homogenates used in the BioPlex method were quantified using a Bradford protein assay. Briefly, homogenates were diluted in NaOH 1N (appropriate dilutions for homogenate sets were determined by trial) to fall within the linear portion of a standard curve of bovine serum albumin (BSA) ranging from 0.075 mg/ml to 1.5 mg/ml. Samples were plated in triplicate and incubated with Protein Assay Buffer (BioRad) for 10 minutes at room temperature and then read with a plate reader (Synergy HT, BioTek, Winooski, VT).

6.3.5c Protein analysis: BioPlex Hippocampal tissues were immediately homogenized with a glass/Teflon homogenizer in 1.5 ml Sucrose (320 mM)/Tris (4 mM) and an aliquot was stored at -20 ºC for analysis of protein content by Bradford assay. An aliquot of 600 µl was added to 120 µl of Extraction 6x (300 mM Tris-HCl, 6% BSA, 900 mM NaCl, 12 mM CaCl, 0.6 mM

176 bensethonium chloride, 12 mM EDTA NA2, 6% Triton 100X, 0.0004% aprotinin, 3 mM PMSF, 0.3% NaN3, pH 7.0) and centrifuged at 15,000 x g for 10 min at 4 ºC. Aliquots of the supernatant were stored at -20 ºC for evaluation of cytokines with a BioPlex analysis. Twelve cytokines, GM-CSF, IFNγ, IL-1α, IL-1ß, IL-2, IL-4, IL-5, IL-6, IL-10, IL- 12p70, IL-13, and TNFα, were examined in each sample simultaneously with a bead-based flow cytometric immunoassay (Bio-Rad, BioPlex Pro Rat Standard, 171-K1002M). Briefly, a mixture of 12 distinct capture beads (fluorescently dyed microspheres) each with a specific spectral address and conjugated to an antibody against one of the cytokines listed above were dispensed across a 96-well plate and protected from light. Samples (diluted in BioPlex Cell Lysis Buffer, Bio-Rad) and antigen standards were added in duplicate and incubated for 1 hr. at 300 RPM at room temperature; unbound materials were washed away (3x). Then biotinylated detection antibodies directed against each of the 12 cytokines were added for 30 min at 300 RPM at RT; unbound materials were washed away (3x). Each well was then incubated for 10 min at 300 RPM at room temperature with a reporter dye, streptavidin-phycoerythrin conjugate (SA-PE), which binds to the detection antibody; unbound materials were washed away (3x). Each well was then resuspended in assay buffer and shaken at 1100 RPM for 30 sec. Finally, the contents of each well was passed through a dual detection flow cytometer with a classification laser that distinguishes each of the 12 cytokines by color of its bound antigen-specific bead and a reporter laser that quantifies each cytokine based upon the fluorescence of bound antigen-specific SA-PE reporter dye. Values were standardized to protein content of the homogenate.

6.3.5d Protein analysis: Flow Cytometry Microglia surface expression of MHCII, TLR4 CX3CR1 and GLT1 was examined by flow cytometry in a small group of young aCSF- and LPS- infused rats, but not aged nor drug treated animals, using the protocol described by Wynne et al. (2009). Briefly, brains were immediately homogenized in Hank’s Balanced Salt Solution (HBSS, pH 7.4), filtered through a 70 µm cell strainer and centrifuged (500 g x 6 min.). The supernatant was discarded and cell pellets were resuspended in 70% isotonic Percoll (GE-healthcare, Uppsala, Sweden) at room temperature. Samples were added to a discontinuous Percoll density gradient (70%, 50%, 35%, and 0%) and centrifuged (2000 g x 20 min.). Microglia were harvested from the interphase between the 70% and 50% Percoll layers, washed and 177 re-suspended in sterile HBSS. The number of viable cells was determined using a hemacytometer and 0.1% trypan blue staining. Extractions yielded approximately 72,000 viable cells. Fc receptors were blocked with anti-CD16/CD32 antibody (eBioscience, San Diego, CA). The cells were then incubated with anti-CD11b/c-PE (BD Bioscience, San Jose, CA) to identify microglia and macrophages, and CD45-PerCP (eBioscience) to distinguish microglia from macrophages. Cells were additionally stained with anti- MHCII-FITC, primary anti-TLR4 (AbCam, Cambridge, MA) and secondary anti-mouse- APC (eBioscience), primary anti-CX3CR1 and secondary anti-rabbit-FITC (eBioscience), or primary anti-GLT1 (BD Bioscience) and secondary anti-mouse-APC. Expression of these surface antigens was determined using a Becton-Dickinson FACSCaliburTM four- color cytometer. Ten thousand events were recorded and microglia were identified by CD11b/c+ and CD45low expression (Ford et al., 1995; Nair and Bonneau, 2006). For each antibody, gating was determined based on appropriate isotype controls. Flow cytometric data were analyzed using FlowJo software (Tree Star, San Carlos, CA).

6.3.6 Statistics SigmaStat software was used to compare groups by analysis of variance (ANOVA) with Fisher LSD as the preferred post-hoc, and to perform Pearson correlation coefficients. Graphs are shown with SEMs represented by error bars. Control aCSF is shown in some graphs as one collapsed group, but aCSF groups were not collapsed for statistical analysis. Within treatment controls, differences between young aCSF, young LPS and Aged are marked * and differences between LPS and Aged are marked **. Within each inflammation group (aCSF, LPS and Aged), differences between control treated animals and drug treatments are marked §, with dark orange used to identify ceftriaxone and green used to identify Riluzole. Within experimental groups, differences across days in the water maze are marked †.

178

6.4 Results

6.4.1 Spatial memory Reduced latency to find the hidden platform in the Morris water maze is used as an indicator of spatial learning and memory (Figure 48). A 3-way repeated measures ANOVA revealed a main between-subjects effect of inflammation group (aCSF, LPS or Aged; F2, 56

= 76.005, p < 0.001), a main within-subjects effect of trial day (F3, 168 = 67.852, p < 0.001), and an interaction between inflammation group and trial day (F6, 168 = 10.416, p < 0.001). A 2-way repeated measures ANOVA between inflammation groups treated with vehicle revealed a main between-subjects effect of inflammation group (F2, 19 = 19.575, p < 0.001), a main within-subjects effect of trial day (F3, 57 = 29.898, p < 0.001), and an interaction between inflammation group and trial day (F6, 57 = 6.065, p < 0.001). Post-hoc analysis indicates that aCSF controls found the platform in less time than LPS controls and Aged controls (*p < 0.001). Main within-subjects effects of trial day were observed by 2-way repeated measures ANOVA between drug treatments within aCSF-infused (F3, 51 = 81.948, p < 0.001), LPS infused (F3, 51 = 14.999, p < 0.001) and Aged (F3, 66 = 6.872, p < 0.001) groups. Trend analysis of trial day revealed significant linear (F1, 17 = 119.545, p < 0.001) and quadratic (F1, 17 = 50.664, p < 0.001) effects within aCSF-infused groups and a linear effect within LPS-infused groups (F1, 22 = 15.464, p = 0.001). Within Aged groups, trend analysis reveals a linear effect of trial day (F1, 22 = 4.163, p = 0.029) and quadratic interaction between trial day and drug treatment within aCSF-infused groups (F1, 22 = 119.545, p < 0.001). Paired t-tests were conducted between the first and last trial day within each Aged group in order to identify the quadratic interaction between trial day and drug treatment show that Aged rats did not improve in finding the hidden platform (p >

0.05) unless treated with Ril (†t7 = 2.602, p = 0.035).

179

Figure 48. Latency to find the hidden platform Young aCSF controls found the platform quickly than LPS controls and Aged controls (*p ≤ 0.001). Aged rats improved across days (†p ≤ 0.035) only when treated with Cef or Ril.

Reduced path length is another indicator of improved memory performance. A 3- way repeated measures ANOVA of total distance swam (Figure 49) revealed a main between-subjects effect of inflammation condition (F2, 56 = 54.450, p < 0.001) and a main within-subjects effect of trial day (F3, 168 = 27.811, p < 0.001) as well as interactions between inflammation group and drug treatment (F4, 56 = 3.203, p = 0.019), inflammation group and trial day (F6, 168 = 15.209, p < 0.001), and an 3-way interaction between inflammation condition, drug treatment and trial day (F6, 168 = 1.825, p = 0.048). A 2-way repeated measures ANOVA within vehicle-treated control groups revealed a between- subjects effect of inflammation condition (F2, 19 = 8.194, p = 0.003), within-subjects main 180 effect of trial day (F3, 57 = 16.652, p < 0.001) and an interaction between trial day and inflammation condition (F6, 57 = 8.609, p < 0.001). A Fisher LSD post hoc test revealed that LPS controls swim more overall distance than aCSF controls and Aged controls (*p ≤ 0.036). Analysis by two-way repeated measures ANOVA reveal no significant between- subjects differences between drug treatments and vehicle treatment within aCSF-infused, LPS-infused or Aged groups (p > 0.05).

Figure 49. Distance Aged controls swam greater distance than aCSF controls (*p < 0.001) and less than LPS controls (*p ≤ 0.036).

Velocity was compared between groups because differences in velocity confound the interpretation of differences in latency and distance (Figure 50). A 3-way repeated 181 measures ANOVA revealed a main between-subjects effect of inflammation condition (F2,

56 = 88.367, p < 0.001), a main within-subjects effect of trial day (F3, 168 = 48.155, p <

0.001), and an interaction between inflammation condition and trial day (F6, 168 = 21.980, p < 0.001). A two-way repeated measures ANOVA between vehicle treated control groups revealed a main between-subjects effect of inflammation condition (F2, 19 = 25.626, p <

0.001), a main within-subjects effect of trial day (F3, 57 = 12.978, p < 0.001), and an interaction between inflammation condition and trial day (F6, 57 = 36.231, p < 0.001); and post hoc analysis indicates that Aged controls swim with significantly less velocity than young, aCSF- or LPS-infused controls (*p < 0.001).

Figure 50. Velocity Aged controls swam the more slowly than young controls (*p < 0.001).

182

Thigmotaxis is defined as the percentage of time spent in the outer perimeter of the pool. Rats may circle the wall of the maze instead of spending time in the area in which the platform can be found; therefore this is an estimate of time spent on a poor strategy. However, if a rat finds the platform quickly, this number can seem inflated because leaving the entry position near the pool wall takes a greater portion of the shorter trial time. Thigmotaxis was evaluated (Figure 51). A 3-way repeated measures ANOVA revealed a main between-subjects effect of inflammation condition (F2, 56 = 54.553, p < 0.001), a main within-subjects effect of trial day (F3, 168 = 89.923, p < 0.001), and an interaction between inflammation condition and trial day (F6, 168 = 8.013, p < 0.001). A two-way repeated measures ANOVA between vehicle treated control groups revealed a main between-subjects effect of inflammation condition (F2, 19 = 13.117, p < 0.001), a main within-subjects effect of trial day (F3, 57 = 24.837, p < 0.001), and an interaction between inflammation condition and trial day (F6, 57 = 5.989, p < 0.001); and post hoc analysis indicates that Aged controls spend a greater percentage of trial time in the pool perimeter than young, aCSF- or LPS-infused controls (*p ≤ 0.001).

183

Figure 51. Thigmotaxis Aged controls spent a greater percentage of time along the pool wall than aCSF- or LPS- infused young rats (*p < 0.001).

After the completion of 6 trials on the fourth day, the platform was removed and we analyzed the time each animal spent within a radius of the missing platform that was equal to 25% of the total area of the water maze pool (Figure 52). Thus, if the animal swam randomly throughout the pool, we would expect the percent of time spent within this radius to be approximately 25%. There was a main effect within inflammation group (F2, 64 = 51.82, p < 0.001). Young aCSF controls spent more time in the vicinity of the missing platform than LPS or Aged controls (*p < 0.001), which both performed similarly at approximately chance levels. No drug treatment had an effect on the amount of time spent near the missing platform.

184

Figure 52. Probe Both LPS and Aged groups spend less time in the vicinity of the missing platform than aCSF (*p < 0.001), and there were no effects of drug treatment.

6.4.2 Microglia activation phenotype and relationship with spatial memory

6.4.2a Inflammation-related mRNA expression We evaluated mRNA for markers associated with inflammation or resolution, including pro-inflammatory IL-1ß and TNFα and anti-inflammatory CX3CL1, CX3CR1, TGFß and BDNF. IL-1ß was the only marker that changed significantly between groups, and increased in all LPS-infused rats compared to aCSF-infused and Aged (**p < 0.001, Figure 53). The relationships between spatial memory and all mRNA and proteins measured in the hippocampus were evaluated by Pearson product-moment correlation. Only GLT1, xCT, TLR4, IL-1α and IL-1ß significantly correlated with water maze behavior on the final day of testing. TLR4 expression is necessary for the recognition of LPS, is generally increased after initiation of a pro-inflammatory state, and may decrease with the development of endotoxin tolerance. There were no significant differences in the expression of TLR4 mRNA between groups. TLR4 mRNA expression did, however, correlate (Table 8, Figure 54) with increased latency to find the hidden platform (r = 0.321, p = 0.010) and increased distance traveled (r = 0.358, p = 0.003) among all comparisons, and these relationships were stronger between aCSF and LPS. Additionally, there is a 185 significant relationship between TLR4 expression and thigmotaxis when only aCSF and Aged are analyzed (r = 0.319, p = 0.037).

Figure 53. Hippocampal inflammatory marker mRNA expression IL-1ß total mRNA expression (A) and expression relative to aCSF controls (B) is elevated in LPS infused rats compared to aCSF and Aged rats (**p < 0.001).

186

Table 8. Correlation between TLR4 mRNA expression and spatial memory Pearson correlation coefficients (r) are displayed for comparisons between all data points and individual group comparisons including and excluding drug treatments. Significant correlations (p < 0.05) are marked with an asterisk.

Figure 54. Increased TLR4 mRNA correlates with impaired spatial memory Scatterplot represents the relationship between TLR4 mRNA expression and latency to find the hidden platform in the Morris water maze on day 4. Data points are individual rats, and the dashed gray trend line and corresponding R2 value in the bottom left represent all data points. Black diamonds represent aCSF controls and black circles highlight LPS controls. The correlation between these two groups alone is marked with the black trend line and the bold R2 value bottom center.

187

6.4.2b Cytokine protein expression Pro-inflammatory and anti-inflammatory cytokine proteins were investigated within the hippocampus (Figure 55). Significant increases were observed in IL-1ß and IL- 1α with LPS treatment. There was a main effect of inflammation group on IL-1α and IL- 1ß (p < 0.001). LPS controls expressed more IL-1α and IL-1ß (**p ≤ 0.003) than aCSF or Aged controls and treatment with Cef or Ril did not have a significant effect. The differences in production of IL-1ß protein parallel those observed in mRNA expression (Figure 56). IL-1ß and IL-1α expression also correlate with poor performance on the last testing day in the Morris water maze (Table 9, Figure 57) when only aCSF and LPS groups are compared. Il-1ß mRNA expression, IL-1ß protein expression and IL-1α protein expression correlate with increased time to find the hidden platform, increased swim path, and reduced time spent in the vicinity of the missing platform during the probe trial (p ≤ 0.05). Il-1ß protein expression also correlated with increased time spent circling the pool perimeter (r = 0.352, p = 0.030). Unlike cytokines in the hippocampus, none of the same cytokines measured in the serum were elevated after four continuous weeks of LPS infusion. There were, however, main effects inflammation group (p ≤ 0.031) in which pro-inflammatory cytokines IL-1ß (*p = 0.007) and IL-2 (**p ≤ 0.042) and anti-inflammatory cytokines IL-4 (*p = 0.017) and IL-10 (**p ≤ 0.037) were elevated in Aged controls. Cef treatment in Aged reduced the expression of IL-1ß (§p = 0.041) and IL-10 (§p = 0.037) in the serum.

188

Figure 55. Hippocampal cytokine protein expression IL-1ß total protein expression (A) and expression relative to aCSF controls (B) is elevated in LPS infused rats compared to aCSF and Aged rats (**p ≤ 0.003).

189

Figure 56. IL-1ß mRNA and protein expression increase similarly after LPS IL-1ß mRNA and protein are both elevated approximately 4-fold in LPS treated animals compared to both aCSF and Aged (**p ≤ 0.003).

Table 9. IL-1α/ß mRNA and protein expression correlate with poor spatial memory Pearson correlation coefficients (r) are displayed for comparisons between all data points and individual group comparisons including and excluding drug treatments. Significant correlations (p < 0.05) are marked with an asterisk. Rows represent IL-1ß protein, IL-1ß mRNA and IL-1α protein. 190

Figure 57. Increased IL-1ß correlates with impaired spatial memory in LPS Scatterplots represent the relationship between IL-1ß protein expression and both latency to find the hidden platform in the Morris water maze on trial day 4 (A) and duration spent within the vicinity of the missing platform during the probe trial (B). Data points are individual rats, and the dashed gray trend line and corresponding R2 value in the bottom left represent all data points. Black diamonds represent aCSF controls and black circles highlight LPS controls. The correlation between these two groups alone is marked with the black trend line and the bold R2 value bottom center.

191

Figure 58. Serum cytokine protein expression Total protein expression (A) and expression relative to aCSF controls (B) is quantified. IL-1ß and IL-4 are elevated in Aged controls compared to aCSF controls (*p = 0.017) and IL-10 and Il-2 are elevated in Aged controls compared to both aCSF and LPS controls (**p ≤ 0.037). Cef treatment reduced expression of IL-1ß and IL-10 in Aged rats (§p ≤ 0.041).

192

6.4.2c LPS-induced TLR4, MHCII and GLT1 but not CX3CR1 on microglia Hippocampal microglia (CD11b/c+ cells) increased surface expression of the LPS receptor TLR4 and the pro-inflammatory markers MHCII and GLT1, but not the anti- inflammatory CX3CR1, after 4 weeks of continuous LPS i.c.v. relative to aCSF controls (*p < 0.05). GLT1 on microglia is consistent with an M1-type pro-inflammatory profile (Colton & Wilcock 2010; Colton, 2009) and increased GLT1 on LPS-activated microglia is consistent with previous observations (Persson et al. 2005). There is not yet agreement in the field on the role of GLT1 on microglia. It may serve to alert microglia to glutamate in the environment, perhaps as a signal from stressed or dying neurons, and activate microglia. Alternatively, GLT1 on microglia may serve as an additional form of glutamate sequestration in an inflammatory environment and augment the process already being carried out by astrocytes. We made an effort to investigate GLT1 on astrocytes alone, but were unable to establish a protocol to successfully isolate astrocytes for flow cytometry.

193

Figure 59. LPS induced MHCII and TLR4 and GLT1 but not CX3CR1 on microglia LPS 4w induces an increase in MHCII (A-B) TLR4 (C-D) and GLT1 (E-F) on a subset of the total microglia population (CD11b/c+) relative to aCSF controls (*p < 0.05).

194

6.4.3 GLT1 and xCT expression and relationship with spatial memory Glutamate transport and glutamate-cystine exchange were evaluated by investigating GLT1 and xCT mRNA expression in hippocampal tissue (Figure 60). There was a main effect of inflammation group (F2, 67 = 6.4, p = 0.003) in which Aged controls express less GLT1 mRNA than both aCSF and LPS controls (**p ≤ 0.014), unless treated with Ril (§p = 0.028). These data suggest that glutamate transport may be impaired and that Ril treatment may restore glutamate transport in aged rats. There were no differences in expression of xCT mRNA.

Figure 60. Hippocampal GLT1 and xCT mRNA expression GLT1 total mRNA expression (A) and expression relative to aCSF controls (B) is reduced in Aged rats compared to aCSF and LPS controls (**p ≤ 0.014), and reversed with Ril treatment (§p = 0.028). Total xCT mRNA expression (C) and relative expression (D) are not different between groups. 195

GLT1 mRNA correlated with reduced latency to find the hidden platform (r = - 0.295, p = 0.019) and increased time spent in the vicinity of the missing probe when all data points were compared (r = 0.323, p = 0.010; Table 10, Figure 61). These correlations were stronger when only aCSF and Aged groups were compared, and strongest when drug treated rats were not included. Time spent in the perimeter of the pool (thigmotaxis) is negatively correlated with GLT1 (r = -.0538, p = 0.0473), and positively correlated with xCT expression (r = 0.518, p = 0.048) when aCSF and Aged controls were examined.

Table 10. Correlation between GLT1 and xCT mRNA and spatial memory Pearson correlation coefficients (r) are displayed for comparisons between all data points and individual group comparisons including and excluding drug treatments. Significant correlations (p < 0.05) are marked with an asterisk. GLT1 correlated with improved water maze performance, and this relationship is strongest when evaluate between aCSF and Aged controls (highlighted), while xCT had an inverse relationship with thigmotaxis.

196

Figure 61. Reduced GLT1 and increased xCT correlate with impaired spatial memory Scatterplots represent the relationship between GLT1 mRNA expression and both latency to find the hidden platform in the Morris water maze on trial day 4 (A) and time spent within the vicinity of the missing platform during the probe trial (B). Time spent within the pool perimeter is compared with GLT1 (C) and xCT (D). Data points are individual rats, and the dashed gray trend line and corresponding R2 value in the bottom left represent all data points. Black diamonds represent aCSF controls and black squares highlight Aged controls. The correlation between these two groups alone is marked with the black trend line and the bold R2 value bottom center.

197

6.5 Discussion Neuroinflammation and aging are two of the strongest risk factors for AD and are both associated with impairment in spatial learning and memory. Inflammation may impair memory through direct modulation of the glutamatergic system (Rosi et al. 2006; Rosi et al. 2005; Marchalant et al. 2009b). My previous studies suggest that elevated GLT1 can attenuate inflammation-induced cognitive impairment (Chapter 5). Therefore, we treated young rats with experimentally-induced neuroinflammation and aged rats with two drugs that have been shown to increase glutamate clearance by GLT1, Cef and Ril (see Appendix E and Appendix F). We found that LPS-induced neuroinflammation in young rats is qualitatively different from age-associated neuroinflammation. LPS induces an M1/pro- inflammatory phenotype, whereas aging is not associated with a strong central pro- inflammatory response but is associated with a mixed peripheral immune phenotype. LPS- driven memory impairment in young rats correlates with elevation of IL-1 cytokines, and memory impairment in aged rats correlates with changes in glutamate clearance and processing in astrocytes. These two phenomena, however, may interact, as elevated IL-1ß attenuates glutamate clearance (see Section 1.4.2). The drugs Cef and Ril had no effect in young rats, but in aged rats they were associated with an attenuated immune response and improved spatial memory performance. Taken together, these data reinforce the connection between neuroinflammation, hyperglutamatergic activity and memory impairment. Moreover, this study highlights the therapeutic potential of enhancing glutamate clearance in AD.

6.5.1 Neuroinflammation and age are associated with similar cognitive impairments but qualitatively different inflammatory profiles Chronic neuroinflammation induced by continuous LPS over 3 weeks drove impairments in hippocampal-dependent memory that are similar to those seen in aged rats. LPS-infused and aged rats took longer (or failed) to find the hidden platform and spent less time swimming in the vicinity of the missing platform than young aCSF-infused rats. Despite similar cognitive deficits in the Morris water maze, the inflammatory profile differed between LPS-induced neuroinflammation and inflammation observed in aged rats. Inflammation generated by continuous LPS infusion was characterized by elevations in IL- 1α and IL-1ß protein in the hippocampus, as well as increased microglial surface

198 expression of MHCII, GLT1 and TLR4. IL-1ß mRNA expression was dramatically increased after LPS infusion, but LPS-induced changes in expression of other genes were not observed. These results are consistent with a previous study that reported a 3-fold increase in IL-1ß mRNA in the hippocampus of rats infused with LPS i.c.v. over four weeks, but no changes in TNFα or IL-6 mRNA (Hauss-Wegrzyniak et al., 1998). Elevated M1/pro-inflammatory markers (IL-1α, IL-1ß, MHCII and GLT1) in the absence of changes in M2/anti-inflammatory markers (IL-4, IL-10, TGFß, CX3CL1, CX3CR1) suggest that after weeks of continuous LPS infusion, the hippocampus remains in a pro-inflammatory state. Age-associated neuroinflammation is characterized by an increased number of hippocampal MHCII+ microglia, albeit fewer than in LPS-infused young rats (Marchalant et al. 2009; unpublished observations), and may be a response to a past or present immune challenge. MHCII+ microglia in the hippocampus of aged rats does not correlate with cognitive impairment (unpublished observations; VanGuilder et al., 2011). The cytokine profile of the aged hippocampus did not differ from young aCSF-infused rats, and suggests that the increased presence of MHCII+ microglia in the hippocampus does not necessarily indicate a pro-inflammatory environment. Age-related elevations in cytokines were only observed in the serum and included the pro-inflammatory cytokines IL-1ß and IL-2 in addition to the anti-inflammatory cytokines IL-4 and IL-10. The presence of both M1/pro- inflammatory and M2/pro-resolution markers in the serum of aged rats indicates a mixed immune phenotype. In order to understand what factors may be related to inflammation-induced and age-associated impairment in hippocampal-dependent spatial learning and memory, we analyzed correlations between all of the immune factors examined in the hippocampus with performance in the water maze. Expression of TLR4, IL-1α, IL-1ß, GLT1 and xCT correlated with cognitive behavior on the final day of testing. TLR4 surface expression on microglial increased after LPS infusion. Although TLR4 mRNA was not significantly elevated, TLR4 mRNA correlated with poor cognitive performance. In LPS-infused rats, a relationship between TLR4 and memory impairment was expected because direct activation of TLR4 by LPS is the first step in the immune cascade leading to cognitive impairment. The relationship between TLR4 and cognitive

199 performance is more interesting in aged rats, in which there was no experimental activation of TLR4 by LPS. In aged rats, more TLR4 could reflect a past or current immune challenge, and this is supported by the elevation of both M1/pro-inflammatory (IL-1ß and IL-2) and M2/pro-resolution cytokines (IL-4 and IL-10) in the serum of aged rats that were not elevated in young aCSF- or LPS-infused rats. TLR4 is necessary for LPS to elicit an immune response, and a decrease in TLR4 could diminish the response to LPS over time. However, surface expression of TLR4 on microglia increased after 4 weeks of continuous LPS infusion, suggesting that microglia were still able to respond to LPS. We did not see an LPS-induced increase in TLR4 mRNA after 3 weeks of LPS infusion, possibly because flow cytometric analysis was conducted after 4 weeks of LPS exposure or because rtPCR on whole hippocampal homogenates was less sensitive to changes in TLR4 than flow cytometry on individual microglia. While tolerance to LPS would confound the interpretation of our results and is important to consider, we have no evidence suggesting that tolerance occurs. Furthermore, if some degree of endotoxin tolerance did occur in LPS-infused rats, it was not sufficient to protect against cognitive impairment, and this may be attributed to continued elevation of pro- inflammatory IL-1α and IL-1ß.

6.5.2 IL-1α and IL-1ß correlate with LPS-induced cognitive impairment Impaired spatial learning and memory in young, LPS-infused rats is likely due to elevation of the IL-1 cytokines. LPS infusion over 3 weeks in young rats generated an approximate 2-fold elevation in IL-1α protein and 4-fold elevation in both IL-1ß mRNA and protein in the hippocampus. In young aCSF- and LPS-infused rats, IL-1α and IL-1ß were significantly correlated with poor water maze performance (i.e. increased latency, lengthened swim path, increased perseverance in pool perimeter, and less time searching for the missing platform in the correct area). A low physiological level of Il-1ß is required for the normal induction of LTP and successful performance in tasks sensitive to hippocampal damage (Avital et al., 2003; Goshen et al., 2007). Paradoxically, the elevated IL-1 levels associated with neuroinflammation produce impairment in these same tasks (Moore et al., 2009), and blocking the effects of IL-1 preserves memory performance (Terrando et al., 2010). Furthermore, treatment with the IL-1R antagonist in a triple- transgenic AD mouse model not only attenuates CNS inflammation and cognitive 200 impairment, but also reduces AD-associated Aß and tau pathology (Kitazawa et al., 2011). This demonstrates that immune modulation is sufficient to improve behavioral outcomes despite the presence of genetically-driven AD pathology. How IL-1ß impairs hippocampal-dependent memory processes is not well understood, but may be through dysregulation of the glutamatergic system (Figure 62). The perforant path efferent input to the hippocampus and the mossy fiber projections from the DG to the CA3 and CA1 are primarily glutamatergic; therefore, IL-1-driven hippocampal-dependent memory impairment is likely due to disruption of this circuitry. Glutamate opens the NMDAR, which allows Ca2+ influx and stimulates production of immediate early genes, such as Arc. This process is integral to the induction of long-term potentiation (LTP) and memory formation. Excessive glutamate, paradoxically, is detrimental to cognition. The strongest correlation between Il-1ß and memory performance was found in aged rats. The postulation that this relationship is mediated by changes in the glutamatergic system is supported by previous studies in aged rats that reported changes in hippocampal glutamatergic regulation and memory. These include changes in NMDAR activity and Ca2+ regulation, decreased number of synapses, loss of LTP, enhanced long-term depression (LTD), and poor performance in a number of memory tasks (Rosenzweig & Barnes 2003). IL-1 upsets the balance of glutamatergic signaling at the post-synaptic site. In this study and previous work, we demonstrated that three to four weeks of chronic i.c.v. LPS infusion in young rats stimulates an increase in IL-1ß, interferes with induction of LTP, disrupts post-synaptic Arc signaling in response to a novel environment, and impairs performance is the spatial water maze task (Hauss-Wegrzyniak et al., 1998). Moreover, drugs that reduce neuroinflammation and/or attenuate glutamatergic signaling restore cognitive performance. Consistent with our findings, innate immune activation and elevated IL-1ß disrupts Arc signaling, impedes induction of LTP and impairs performance in hippocampal-dependent tasks (Hein et al., 2010; Barrientos et al., 2009; Chapman et al., 2010; Frank et al., 2010; chen et al., 2008). IL-1 interacts with the glutamatergic function not only at post-synaptic receptors, but also at astrocytic transporters. IL-1ß dose-dependently increases surface expression of glutamate AMPARs (Lai et al., 2006) and increases Ca2+ entry through hippocampal

201

NMDARs by (Luca et al., 2003; Viviani et al., 2003; Wang et al., 1999). IL-1ß effects glutamatergic homeostasis not only through excitatory post-synaptic receptors, but also by changing glutamate processing within astrocytes. For example, IL-1ß increases the velocity of cystine-glutamate exchange by xCT, which increases cystine uptake and elevates extracellular glutamate levels in murine cortical cultures (Fogal et al., 2007). Furthermore, IL-1ß reduces GLT1 on astrocytes and dose-dependently decreases glutamate uptake (Sama et al., 2008; Prow and Irani, 2008; Hu et al., 2000; Ye and Sontheimer, 1996). Increasing activity of NMDARs, AMPARs and xCT or reducing activity of GLT1 could all lead to a hyperglutamatergic state. In fact, peripheral or central injection of IL- 1ß drives an increase in extracellular glutamate in rats (Kamikawa et al., 1998; Mascarucci et al., 1998; Chao et al., 1995) and IL-1ß concentration directly correlates with hyperexcitability of glutamatergic circuitry in multiple sclerosis patients in vivo (Rossi et al., 2012). These data are cohesive in illustrating the consequences of the pro- inflammatory cytokine on glutamatergic regulation. Taken together, these results suggest that IL-1 disrupts hippocampal-dependent learning and memory through intensification of glutamatergic neurotransmission, and support the use of drugs that modulate glutamate function to prevent inflammation- associated memory impairment. It is not clear if IL-1 acts directly upon NDMARs, AMPA, xCT and GLT1 or if the effects of IL-1 on one or some indirectly affect the others. Nevertheless, increasing glutamate clearance may be sufficient to ameliorate some of these effects on memory. Therefore, we treated rats with the drugs Cef and Ril to elevate the glutamate transporter GLT1. In the current study, the drugs did not produce an elevation in GLT1 mRNA, nor improve cognitive performance in LPS-infused young rats as they did at the same dose and regimen in aged rats and other models (see Appendix E and Appendix F). IL-1 reduces GLT1 expression and function (Sama et al., 2008; Prow and Irani, 2008; Hu et al., 2000; Ye and Sontheimer, 1996), and it is possible that neither Cef nor Ril were able to increase glutamate clearance in the presence of abundant IL-1. LPS- and aCSF-infused rats expressed similar levels of GLT1 which were not enhanced by drug treatment, suggesting that perhaps Cef and Ril are able to restore GLT if reduced, but not able to increase GLT1 expression above normal physiological levels. We predict, however,

202 that if GLT1 expression had been augmented, that memory function would have been protected from impairment induced by chronic neuroinflammation.

6.5.3 Elevated extracellular glutamate may drive cognitive impairment in aged rats In contrast to young rats, Ril and Cef treatment in aged rats prevented the age- related decrease in GLT1, inhibited the age-associated increase in serum IL-1ß, and improved performance in the spatial water maze task. GLT1 mRNA were reduced significantly in aged rats compared with young aCSF-infused rats, but were not reduced in young LPS-infused rats, and there were no changes in xCT mRNA. Moreover, GLT1 and xCT were the only markers examined in addition to TLR4 and IL-1α/ß that correlated with cognitive performance. Higher GLT1 levels directly correlated with successful water maze performance (decreased latency and thigmotaxis, increased time spent in the area of the missing platform), consistent with the relationship between GLT1 and cognitive performance described in the previous chapter. GLT1 levels may reflect glutamate clearance from the synapse, and aged rats that had the least GLT1 performed the worst in the water maze task. Likewise, xCT inversely correlated with successful water maze performance (increased thigmotaxis) when young aCSF-infused and aged rats were compared. Levels of xCT may reflect production of the protective antioxidant, glutathione, but also export of glutamate into the extracellular space. Rats with higher xCT expression performed more poorly. The opposing functions of GLT1 and xCT on synaptic glutamate likely explain their opposite relationships with cognitive performance. Less GLT1 in aged rats could result in a retardation and/or reduction in glutamate clearance, and potentially prolong the post-synaptic response to pre-synaptic glutamate release. Furthermore, reduced clearance by GLT1 in the presence of normal glutamate-cystine exchange by xCT may amplify accumulation of glutamate in the synapse. Excessive extracellular glutamate may depolarize the post-synaptic membrane and engage extrasynaptic NMDARs that express more NR2B subunits, which are more likely to lead to cell death then synaptically located NDMARs with the predominant 2A subunit (Hardingham & Bading 2010; Potier et al. 2010). All of these mechanisms are plausible explanations for the correlations between GLT1 and xCT with cognitive impairment in aged rats, and likely work in concert.

203

The reduction in GLT1 in our aged F-344 rats is consistent with both an observed decrease of GLT1 in aged Sprague-Dawley rats that is associated with reduced glutamate uptake, and in the elderly (Potier et al. 2010). GLT1 expression is reduced in early stage AD in the hippocampus but not in the cerebellum, a region that is spared by AD pathology (Jacob et al. 2007). GLT1 increases along with astrogliosis in the regions with the most pathology in AD (Masliah et al. 1996), but is reduced globally in autopsied AD brains compared to non-AD controls (Scott et al., 2011). Increased astrogliosis and GLT1 expression in regions with AD pathology may be interpreted as a compensatory response in order to clear more glutamate from the synapse. GLT1 interacts with phosphorylated tau and is colocalized with neurofibrillary tangles (NFTs) in AD patients (Sasaki et al. 2009; Woltjer et al. 2010). GLT1 is also colocalized with Aß in AD tissue and is oxidized by Aß to a detergent insoluble form that is associated with progression of AD cognitive impairment (Li et al. 1997; Lauderback et al. 2001b; Woltjer et al. 2010). Glutamate transport in patients with AD pathology tested in fibroblasts was decreased by 60% compared to controls, and transport was halved again by the introduction of oxidative stress, a feature that was not seen in fibroblasts from controls and is blocked by the antioxidant glutathione (Begni et al. 2004). Furthermore, splice variants that decrease glutamate transport are increased with regional variability in autopsied AD tissue (Scott et al., 2011). Neuroinflammation associated with age and pathology may be one factor in the neurodegenerative brain that promotes the expression of various splice variants, the overall reduction of GLT1 expression and reduced glutamate clearance.

204

Figure 62. Proposed relationship between GLT1, xCT and cognitive impairment GLT1 is correlated with improved memory performance, and xCT with negative performance. Hypothetically, if neuroinflammation of aging reduces glutamate clearance by GLT1 and increases glutamate transport by xCT, then glutamate will accumulate in the extracellular space. Excessive glutamate may trigger increased activity at NMDARs and disrupt LTP as well as lead to excitotoxicity.

6.5.4 Glutamate regulation as a target for inflammation- and age-associated cognitive impairment Cef and Ril had moderate effects on spatial learning and memory in the Morris water maze. In young aCSF-infused rats, Cef treatment did not change water maze performance, but Ril treatment reduced latency, total distance and thigmotaxis. LPS- infused rats were impaired in the water maze, and Ril treatment was not beneficial in this group, but instead increased latency to find the platform and total distance swam. Ril and Cef treatment in LPS-infused rats both increased velocity, but unlike Ril treatment, Cef treatment in LPS-infused rats lead to reduced latency and distance swam to find the hidden platform. While an increase in velocity could explain reduced latency to find the hidden platform, LPS infused rats treated with Cef maintained a similar velocity across the final 3 days of testing, the days in which latency and distance improved the most. LPS-infused Cef treated rats also spent less percentage of total swim time along the edge of the pool on successive days; suggesting that they learned to choose a better swim strategy and that this contributed to better cognitive performance with Cef treatment. Another interpretation of 205 reduced thigmotaxis may be reduced anxiety and willingness to travel into the interior of the pool, but my previous study demonstrated that LPS-infused rats did not show increased anxiety as measured by time spent in the perimeter of the open field (Section 3.4.1). Oddly, Ril and Cef treatments in young rats did not increase GLT1 expression; therefore, we do not know what properties of these drugs are responsible for these effects on behavior in young rats. Ril and Cef were both associated with a modest improvement in cognitive function in aged rats in which they only improved in latency to find the hidden platform across days if treated with Ril or Cef, but performance was not significantly different from aged controls overall. The reduced latency to find the platform in drug treated aged rats was not associated with increased velocity or decreased distance swam, but was associated with significantly less time spent in the pool perimeter (see Figure 48-52). Reduced thigmotaxis may mean that Ril and Cef reduce anxiety in aged rats and make them more willing to venture into the area in which the platform can be found, or that these drugs are associated with choosing a better swim strategy. Ril and Cef treatment in aged rats were associated with increased GLT1, but only Ril treatment reached significance. Additionally, both Ril and Cef treatment were associated with reduced IL-1ß and IL-10 in aged rats, but only Cef reached significance. The positive effects of Cef and Ril on cognitive performance in Aged and of Cef in young rats with neuroinflammation, albeit positive, were modest, and the expected elevation in GLT1 (see Appendix E and Appendix F) only occurred with Cef treatment in aged rats. However, the correlations between GLT1 and xCT expression with spatial learning and memory suggest that the pharmacological enhancement of GLT1 and modulation of the glutamatergic system in general are worthwhile investigations. Currently, there are no disease-modifying treatments for Alzheimer’s and the best course for prevention involves use of anti-inflammatories and antioxidants, but these have failed clinical trials in Alzheimer’s patients. Glutamate dysregulation is a mechanism that may directly lead to memory impairment and neuronal cell death (see Section 1.4.1) and is one of the best targets for disease-modifying therapy. Already, drugs that modify glutamatergic function including Memantine, a non-competitive NMDAR antagonist, and the anti- epileptic Levetiracetam (Keppra) are currently being used and tested for use, respectively,

206 in AD (Danysz & Parsons 2012; Wenk et al. 2006; Palop & Mucke 2010; Sanchez et al. 2012; ClinicalTrials.gov). Cef, Ril and other drugs that might enhance glutamate clearance are currently in clinical trials for use in AD, amyotrophic lateral sclerosis (ALS) and multiple sclerosis (MS); highlighting the therapeutic potential of enhancing glutamate transport in diseases characterized by neuroinflammation and neurodegeneration. Enhancing glutamate uptake within the tripartite synapse is likely to counteract the effects of neuroinflammation as well as improve cognition in AD, and will be a beneficial part of a multifactorial treatment approach to AD.

207

Chapter 7: General Conclusions

Chronic neuroinflammation and glutamatergic dysregulation interact to drive the clinical symptoms in Alzheimer’s (AD) and Parkinson’s (PD) diseases. Understanding this relationship was the primary goal of my work. I designed experiments to investigate chronic neuroinflammation and glutamatergic regulation in young rats with experimentally-induced neuroinflammation as well as aged rats, and measured various biochemical markers and behavioral outcomes related to AD and PD. These studies provide evidence suggesting that chronic neuroinflammation produces changes in aminergic neurotransmitter production in the basal ganglia (decreased tyrosine hydroxylase [TH] and dopamine-ß-hydroxylase [DBH]), effects the regulation of glutamate in the hippocampus (changes in GLT1 and poor performance in a hippocampal- sensitive behavioral task), and that the brains of young rats compensated against the continued presence of neuroinflammation over time. In order to understand the involvement of glutamatergic signalling in neuroinflammation-induced impairment, I tested drugs that regulate glutamate neurotransmission at three different positions in the tripartite synapse: pre-synaptic, post-synaptic and astrocytic (Figure 63). Attenuation of glutamatergic neurotransmission with these drugs reduced microglia activation and benefited cognitive function under conditions of inflammation or aging. Time-dependent biochemical and behavioral characteristics of AD and PD in the temporal lobe (Chapters 2, 4-6), midbrain and brainstem (Chapter 3) were reproduced by neuroinflammation created in young rats by intracerebroventricular (i.c.v.) lipopolysaccharide (LPS) infusion over 2, 4 or 8 weeks. These results support the hypothesis that neuroinflammation contributes to the onset and progression of AD and PD in a time-dependent manner. The studies described in this thesis demonstrated that continuous LPS infusion drove robust microglia activation (major histocompatibility complex II [MHCII+] expression) and prolonged production of pro-inflammatory cytokines (interleukin 1α and 1ß [IL-1α/ß]) in vulnerable temporal lobe (hippocampus), midbrain 208

(substantia nigra [SN]) and brainstem (locus coeruleus [LC] and raphe nuclei) regions. Chronic neuroinflammation temporarily reduced the capacity to produce aminergic neurotransmitters (decreased number of neurons immunopositive for tyrosine hydroxylase [TH+], phosphorylated TH [pTH+] and dopamine-ß-hydroxylase [DBH+]). Furthermore, prolonged neuroinflammation was associated with AD-like memory impairment (poor performance in the hippocampal-dependent water maze task), and a PD-like motor impairment (reduced rearing). Interestingly, the brains of young rats recovered over time from the consequences of low-level chronic neuroinflammation. For example, when the period of neuroinflammation is extended, the enzymes necessary to produce the neurotransmitters dopamine and norepinephrine were initially reduced but later surpassed basal levels (Chapter 3). Similarly, performance in a hippocampal- and glutamate-sensitive cognitive task were impaired after 4 weeks but normal after 8 weeks of chronic neuroinflammation (Chapters 4 & 5). These results suggest that there are natural mechanisms by which the brain adjusts to the presence of neuroinflammation and, more importantly, that we may be able to pharmacologically reproduce these changes and interrupt disease progression. My next series of investigations were designed to uncover the mechanism by which the young brain recovers cognitive function in the presence of a continuous pro- inflammatory environment. I narrowed the focus of these experiments to factors involved in glutamatergic neurotransmission and the consequences of chronic neuroinflammation indicative of AD-like pathology. In order to identify which changes in glutamatergic signalling attenuate the outcomes of chronic neuroinflammation, I tested drugs that manipulate pre-synaptic glutamate release, post-synaptic glutamate receptor function, and glutamate clearance by astrocytes. Caffeine reduces pre-synaptic glutamate release and attenuates microglia activation in the hippocampus of LPS-infused young rats (Chapter 2). Reduction of pre-synaptic glutamate release by endocannabinoids also reduces microglia activation and improves spatial memory performance in young rats treated with continuous LPS and in aged rats (Marchalant et al. 2008; Marchalant et al. 2007; Marchalant et al. 2009). Together, these results suggest that neuroinflammation and age are associated with excessive glutamatergic activity and that restoring glutamatergic signalling toward homeostasis is beneficial.

209

Next, I investigated blockade of excessive glutamatergic signalling at the post- synaptic N-methyl-D-aspartate receptor (NMDAR) with the drug memantine (Chapter 4). I choose doses that prevent microglia activation and memory impairment when given during the chronic infusion of LPS (Rosi et al. 2006; Rosi et al. 2009). In order to determine whether memantine reduces the effects of chronic neuroinflammation or prevents the development of these effects, I collected pilot data in which treatment with memantine was delayed. The infusion period of LPS was also extended to 8 consecutive weeks. Contrary to our predictions, longer exposure to neuroinflammation did not potentiate memory impairment; instead, rats infused with LPS over 8 weeks performed similarly to controls in the Morris water maze. Even more interesting, this cognitive recovery was disrupted by delayed treatment with memantine. Attenuating glutamate signaling by partial antagonism of the NMDAR is sufficient to protect the brain from neuroinflammation-induced cognitive impairment (Rosi et al. 2006; Rosi et al. 2009), but not if treatment is delayed. Therefore I speculated that the compensatory response against neuroinflammation-induced impairment may have been a mechanism that also attenuated glutamatergic function, after which, antagonism of the NMDAR by glutamate produced a hypoglutamatergic state that is not conductive to memory processing. In order to discern whether a glutamatergic mechanism was involved in recovery from the consequences of neuroinflammation, I investigated the expression of pre-synaptic (synaptosomal-associated protein 25 [SNAP25]), post-synaptic (NMDAR subunits and calcium binging proteins) and astrocytic (glutamate transporter 1 [GLT1]) proteins. I found that only the expression of GLT1 was positively correlated with improved memory performance over time in the presence of a neuroinflammatory environment (Chapter 5), and postulated that augmenting GLT1 function would be protective against inflammation- induced deficits. If this is true, it confirms my speculation that the brain recovers from inflammation-induced cognitive impairment by reestablishing the homeostasis of glutamate, and that delayed memantine treatment impaired cognition after prolonged neuroinflammation because partial blockade of NMDARs after increased removal of glutamate prohibited normal glutamatergic signalling. To test whether increased glutamate clearance is a mechanism by which the brain can reestablish homeostatic glutamatergic function after prolonged neuroinflammation, I

210 next treated young rats exposed to LPS over three weeks with two drugs that elevate GLT1 and increase glutamate clearance, Riluzole and ceftriaxone (Chapter 6). The effects of Riluzole and ceftriaxone were modest but positive; they attenuated the age-associated increase in pro-inflammatory cytokine (IL-1ß) expression and allowed aged rats to improve performance across days in the spatial memory task. These findings are significant because few treatments improve cognition in aged rats, and few drugs improve cognition in later stages of disease progression. Elevated glutamate is able to directly drive cognitive impairment and cell death in AD (see Section 1.4.1a). Neuroinflammation is a risk factor strongly associated with AD and predictive of the later degree of pathology and manifestation of symptoms (see Section 1.5.2). Neuroinflammation drives glutamatergic hyperactivity, and excessive glutamate fuels neuroinflammation (see Section 1.4.2). Decreasing neuroinflammation early in the disease process is protective, but later treatment in not effective (see Section 1.5.2c). Therefore, manipulation of glutamatergic regulation is an attractive target for AD pharmacotherapy. The studies in this thesis elucidate the interactions between chronic neuroinflammation and glutamate over time in brain regions susceptible to degeneration, and the way in which they correspond to behavioral outcomes. More importantly, these studies identify pharmacological targets, and safe drugs that manipulate these targets. As we develop more effective drugs, the possibility of a disease-modifying treatment for AD becomes more tangible.

211

Figure 63. Summary of drugs tested and their generalized effects on glutamate Caffeine (Chapter 2) reduces pre-synaptic glutamate release. Memantine (Chapter 4) is a non-competitive antagonist of NMDARs. Riluzole (Ril) and Ceftriaxone (Cef, Chapter 6) increase glutamate clearance through GLT1. All three mechanisms of reducing glutamatergic neurotransmission reduce microglia activation (MHCII expression or IL-1ß production) to some degree.

212

References

Abarca, C. et al., 2000. Neurochemical changes after morphine, dizocilpine or riluzole in the ventral posterolateral thalamic nuclei of rats with hyperalgesia. European journal of pharmacology, 403(1-2), pp.67–74. Available at: http://www.ncbi.nlm.nih.gov/pubmed/10969145 [Accessed January 17, 2012]. Aguiar, L.M.V. et al., 2006. Neuroprotective effects of caffeine in the model of 6- hydroxydopamine lesion in rats. Pharmacology, biochemistry, and behavior, 84(3), pp.415–9. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16844208 [Accessed October 20, 2010]. Akiyama, Haruhiko et al., 2000. Inflammation and Alzheimer’s disease. Neurobiology of aging, 21(3), pp.383–421. Available at: http://www.ncbi.nlm.nih.gov/pubmed/10858586 [Accessed October 20, 2010]. Alloisio, S. et al., 2004. Differential modulation of ATP-induced calcium signalling by A1 and A2 adenosine receptors in cultured cortical astrocytes. British journal of pharmacology, 141(6), pp.935–42. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=1574277&tool=pmcentr ez&rendertype=abstract [Accessed October 20, 2010]. Ankarcrona, M. et al., 1995. Glutamate-induced neuronal death: a succession of necrosis or apoptosis depending on mitochondrial function. Neuron, 15(4), pp.961–73. Available at: http://www.ncbi.nlm.nih.gov/pubmed/7576644 [Accessed January 30, 2011]. Anon, 1998. Astrocytes regulate microglial phagocytosis of senile plaque cores of Alzheimer’s disease. EXPERIMENTAL NEUROLOGY, 149(2), pp.329–340. Available at: http://apps.isiknowledge.com.proxy.lib.ohio- state.edu/full_record.do?product=WOS&search_mode=Refine&qid=3&SID=2ClGk OlgpD@Glm@3Nii&page=1&doc=3 [Accessed March 7, 2011]. Antuono, P.G. et al., 2001. Decreased glutamate + glutamine in Alzheimer’s disease detected in vivo with (1)H-MRS at 0.5 T. Neurology, 56(6), pp.737–42. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11274307 [Accessed March 10, 2011]. Arendash, G W et al., 2006. Caffeine protects Alzheimer’s mice against cognitive impairment and reduces brain beta-amyloid production. Neuroscience, 142(4), pp.941–52. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16938404 [Accessed August 24, 2010]. Arendash, Gary W et al., 2009. Caffeine reverses cognitive impairment and decreases brain amyloid-beta levels in aged Alzheimer’s disease mice. Journal of Alzheimer’s disease : JAD, 17(3), pp.661–80. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19581722 [Accessed September 21, 2010]. Ascherio, A et al., 2001. Prospective study of caffeine consumption and risk of Parkinson’s disease in men and women. Annals of neurology, 50(1), pp.56–63.

213

Available at: http://www.ncbi.nlm.nih.gov/pubmed/11456310 [Accessed September 24, 2010]. Ascherio, Alberto et al., 2004. Coffee consumption, gender, and Parkinson’s disease mortality in the cancer prevention study II cohort: the modifying effects of estrogen. American journal of epidemiology, 160(10), pp.977–84. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15522854 [Accessed September 28, 2010]. Avital, A. et al., 2003. Impaired interleukin-1 signaling is associated with deficits in hippocampal memory processes and neural plasticity. Hippocampus, 13(7), pp.826– 34. Available at: http://www.ncbi.nlm.nih.gov/pubmed/14620878 [Accessed November 22, 2012]. Azbill, R.D., Mu, X. & Springer, J.E., 2000. Riluzole increases high-affinity glutamate uptake in rat spinal cord synaptosomes. Brain research, 871(2), pp.175–80. Available at: http://www.ncbi.nlm.nih.gov/pubmed/10899284 [Accessed August 31, 2011]. Bakker, A. et al., 2012. Reduction of hippocampal hyperactivity improves cognition in amnestic mild cognitive impairment. Neuron, 74(3), pp.467–74. Available at: http://dx.doi.org/10.1016/j.neuron.2012.03.023 [Accessed October 22, 2012]. Ballard, C., Kahn, Z. & Corbett, A., 2011. Treatment of dementia with Lewy bodies and Parkinson’s disease dementia. Drugs & aging, 28(10), pp.769–77. Available at: http://www.ncbi.nlm.nih.gov/pubmed/21970305 [Accessed December 15, 2012]. Bao, Xiaodong et al., 2009. Transgenic expression of Glud1 (glutamate dehydrogenase 1) in neurons: in vivo model of enhanced glutamate release, altered synaptic plasticity, and selective neuronal vulnerability. The Journal of neuroscience : the official journal of the Society for Neuroscience, 29(44), pp.13929–44. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2811066&tool=pmcentr ez&rendertype=abstract [Accessed August 24, 2010]. Barger, S W & Basile, A.S., 2001. Activation of microglia by secreted amyloid precursor protein evokes release of glutamate by cystine exchange and attenuates synaptic function. Journal of neurochemistry, 76(3), pp.846–54. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11158256 [Accessed July 27, 2011]. Barger, Steven W et al., 2007. Glutamate release from activated microglia requires the oxidative burst and lipid peroxidation. Journal of neurochemistry, 101(5), pp.1205– 13. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=1949347&tool=pmcentr ez&rendertype=abstract [Accessed December 1, 2011]. Barrientos, R.M. et al., 2010. Memory impairments in healthy aging: Role of aging- induced microglial sensitization. Aging and disease, 1(3), pp.212–231. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2995216&tool=pmcentr ez&rendertype=abstract [Accessed January 3, 2011]. Barrientos, R.M. et al., 2009. Time course of hippocampal IL-1 beta and memory consolidation impairments in aging rats following peripheral infection. Brain, behavior, and immunity, 23(1), pp.46–54. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2630971&tool=pmcentr ez&rendertype=abstract [Accessed August 24, 2010]. Beart, P.M. & O’Shea, R.D., 2007. Transporters for L-glutamate: an update on their molecular pharmacology and pathological involvement. British journal of

214

pharmacology, 150(1), pp.5–17. Available at: http://www.ncbi.nlm.nih.gov/pubmed/17088867 [Accessed January 7, 2011]. Beckstrøm, H. et al., 1999. Interindividual differences in the levels of the glutamate transporters GLAST and GLT, but no clear correlation with Alzheimer’s disease. Journal of Neuroscience Research, 55(2), pp.218–229. Available at: http://doi.wiley.com/10.1002/(SICI)1097-4547(19990115)55:2<218::AID- JNR9>3.3.CO;2-C. Begni, B. et al., 2004. Oxidative stress impairs glutamate uptake in fibroblasts from patients with Alzheimer’s disease. Free radical biology & medicine, 37(6), pp.892– 901. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15304259 [Accessed October 30, 2012]. Belarbi, K. et al., 2012. TNF-alpha protein synthesis inhibitor restores neuronal function and reverses cognitive deficits induced by chronic neuroinflammation. Journal of Neuroinflammation, 9(1), p.23. Available at: http://www.jneuroinflammation.com/content/9/1/23 [Accessed January 26, 2012]. Belcastro, V. et al., 2007. Levetiracetam monotherapy in Alzheimer patients with late- onset seizures: a prospective observational study. European journal of neurology : the official journal of the European Federation of Neurological Societies, 14(10), pp.1176–8. Available at: http://www.ncbi.nlm.nih.gov/pubmed/17880574 [Accessed November 16, 2012]. Bertrand, E et al., 1997. Qualitative and quantitative analysis of locus coeruleus neurons in Parkinson’s disease. Folia neuropathologica / Association of Polish Neuropathologists and Medical Research Centre, Polish Academy of Sciences, 35(2), pp.80–6. Available at: http://www.ncbi.nlm.nih.gov/pubmed/9377080 [Accessed October 20, 2010]. Bezzi, P, Domercq, M., Brambilla, L, et al., 2001. CXCR4-activated astrocyte glutamate release via TNFalpha: amplification by microglia triggers neurotoxicity. Nature neuroscience, 4(7), pp.702–10. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11426226 [Accessed June 28, 2012]. Bezzi, P, Domercq, M., Vesce, S, et al., 2001. Neuron-astrocyte cross-talk during synaptic transmission: physiological and neuropathological implications. Progress in brain research, 132, pp.255–65. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11544994 [Accessed June 28, 2012]. Bezzi, P et al., 1998. Prostaglandins stimulate calcium-dependent glutamate release in astrocytes. Nature, 391(6664), pp.281–5. Available at: http://www.ncbi.nlm.nih.gov/pubmed/9440691 [Accessed June 28, 2012]. Bezzi, Paola et al., 2004. Astrocytes contain a vesicular compartment that is competent for regulated exocytosis of glutamate. Nature neuroscience, 7(6), pp.613–20. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15156145 [Accessed March 10, 2012]. Bibbiani, F. et al., 2003. A2A antagonist prevents dopamine agonist-induced motor complications in animal models of Parkinson’s disease. Experimental neurology, 184(1), pp.285–94. Available at: http://www.ncbi.nlm.nih.gov/pubmed/14637099 [Accessed August 11, 2010]. Boje, K.M. & Arora, P.K., 1992. Microglial-produced nitric oxide and reactive nitrogen oxides mediate neuronal cell death. Brain Research, 587(2), pp.250–256. Available

215

at: http://apps.webofknowledge.com/full_record.do?product=WOS&search_mode=Gen eralSearch&qid=23&SID=3BNBmgOLelG2dpbHFi2&page=1&doc=1 [Accessed December 9, 2012]. Botsakis, K. et al., 2010. Blockade of adenosine A2A receptors downregulates DARPP- 32 but increases ERK1/2 activity in striatum of dopamine deficient “weaver” mouse. Neurochemistry international, 56(2), pp.245–9. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19852993 [Accessed August 24, 2010]. Breitner, J.C. et al., 2011. Extended results of the Alzheimer’s disease anti-inflammatory prevention trial. Alzheimer’s & dementia : the journal of the Alzheimer's Association, 7(4), pp.402–11. Available at: http://www.ncbi.nlm.nih.gov/pubmed/21784351 [Accessed June 29, 2012]. Brothers, Holly M, Marchalant, Yannick & Wenk, Gary L, 2010a. Caffeine attenuates lipopolysaccharide-induced neuroinflammation. Neuroscience letters, (2008). Available at: http://www.ncbi.nlm.nih.gov/pubmed/20541589 [Accessed July 5, 2010]. Brothers, Holly M, Marchalant, Yannick & Wenk, Gary L, 2010b. Caffeine attenuates lipopolysaccharide-induced neuroinflammation. Neuroscience letters, 480(2), pp.97–100. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2905864&tool=pmcentr ez&rendertype=abstract [Accessed July 5, 2010]. Brothers, Holly M, Marchalant, Yannick & Wenk, Gary L, 2010c. Caffeine attenuates lipopolysaccharide-induced neuroinflammation. Author’s personal copy Neuroscience Letters. Neuroscience Letters. Brown, G.C. & Bal-Price, A., 2003. Inflammatory neurodegeneration mediated by nitric oxide, glutamate, and mitochondria. Molecular neurobiology, 27(3), pp.325–55. Available at: http://apps.webofknowledge.com/full_record.do?product=WOS&search_mode=Cita tionReport&qid=18&SID=2FE8IIAddNG@3jAk3CG&page=6&doc=51&cacheurlF romRightClick=no [Accessed December 15, 2012]. Brown, G.C. & Neher, J.J., 2010. Inflammatory neurodegeneration and mechanisms of microglial killing of neurons. Molecular neurobiology, 41(2-3), pp.242–7. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20195798 [Accessed September 2, 2010]. Burn, D.J., 2010. The treatment of cognitive impairment associated with Parkinson’s disease. Brain pathology (Zurich, Switzerland), 20(3), pp.672–8. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20522093 [Accessed October 20, 2010]. Burnstock, G. & Rev, P., 2010. Cross Talk Between A 1 and A 2A Adenosine Receptors in the Hippocampus and Cortex of Young Adult and Old Rats. Society, pp.3196– 3203. Butcher, S.P. & Hamberger, A., 1987. In vivo studies on the extracellular, and veratrine- releasable, pools of endogenous amino acids in the rat striatum: effects of corticostriatal deafferentation and kainic acid lesion. Journal of neurochemistry, 48(3), pp.713–21. Available at: http://www.ncbi.nlm.nih.gov/pubmed/2879888 [Accessed January 24, 2011]. Cagnin, a et al., 2006. In vivo evidence for microglial activation in neurodegenerative dementia. Acta neurologica Scandinavica. Supplementum, 185, pp.107–14.

216

Available at: http://www.ncbi.nlm.nih.gov/pubmed/16866919 [Accessed October 20, 2010]. Cagnin, A et al., 2001. In-vivo measurement of activated microglia in dementia. The Lancet, 358(9280), pp.461–467. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11513911 [Accessed October 20, 2010]. Cagnin, Annachiara et al., 2004. In vivo detection of microglial activation in frontotemporal dementia. Annals of neurology, 56(6), pp.894–7. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15562429 [Accessed September 27, 2010]. Cagnin, Annachiara et al., 2001. In vivo measurement of activated microglia in dementia. The Lancet, 358(table 1), pp.461–467. Camacho, a, Montiel, T. & Massieu, L., 2007. Sustained metabolic inhibition induces an increase in the content and phosphorylation of the NR2B subunit of N-methyl-D- aspartate receptors and a decrease in glutamate transport in the rat hippocampus in vivo. Neuroscience, 145(3), pp.873–86. Available at: http://www.ncbi.nlm.nih.gov/pubmed/17331654 [Accessed October 8, 2010]. Cameron, B. & Landreth, G.E., 2010. Inflammation, microglia, and Alzheimer’s disease. Neurobiology of disease, 37(3), pp.503–9. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2823849&tool=pmcentr ez&rendertype=abstract [Accessed December 8, 2010]. Canas, Paula M et al., 2009. Adenosine A2A receptor blockade prevents synaptotoxicity and memory dysfunction caused by beta-amyloid peptides via p38 mitogen- activated protein kinase pathway. The Journal of neuroscience : the official journal of the Society for Neuroscience, 29(47), pp.14741–51. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19940169. Cao, C. et al., 2009. Caffeine suppresses amyloid-beta levels in plasma and brain of Alzheimer’s disease transgenic mice. Journal of Alzheimer’s disease : JAD, 17(3), pp.681–97. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19581723 [Accessed September 2, 2010]. Cerbai, F. et al., 2012. The neuron-astrocyte-microglia triad in normal brain ageing and in a model of neuroinflammation in the rat hippocampus. PloS one, 7(9), p.e45250. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3445467&tool=pmcentr ez&rendertype=abstract [Accessed November 17, 2012]. Chandler, D. et al., 2010. Effects of plant-derived polyphenols on TNF-alpha and nitric oxide production induced by advanced glycation endproducts. Molecular nutrition & food research, 54 Suppl 2, pp.S141–50. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20540146 [Accessed January 16, 2011]. Chao, C.C. et al., 1995. Interleukin-1 and tumor necrosis factor-alpha synergistically mediate neurotoxicity: involvement of nitric oxide and of N-methyl-D-aspartate receptors. Brain, behavior, and immunity, 9(4), pp.355–65. Available at: http://www.ncbi.nlm.nih.gov/pubmed/8903852 [Accessed March 28, 2012]. Chapman, T.R. et al., 2010. Synaptic correlates of increased cognitive vulnerability with aging: peripheral immune challenge and aging interact to disrupt theta-burst late- phase long-term potentiation in hippocampal area CA1. The Journal of neuroscience : the official journal of the Society for Neuroscience, 30(22), pp.7598– 603. Available at:

217

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2891807&tool=pmcentr ez&rendertype=abstract [Accessed October 26, 2012]. Chaudhry, F. a et al., 1995. Glutamate transporters in glial plasma membranes: highly differentiated localizations revealed by quantitative ultrastructural immunocytochemistry. Neuron, 15(3), pp.711–20. Available at: http://www.ncbi.nlm.nih.gov/pubmed/7546749. Chen, Jiang-fan et al., 2001. Neuroprotection by Caffeine and A 2A Adenosine Receptor Inactivation in a Model of Parkinson ’ s Disease. , 21, pp.2–7. Chen, Jing et al., 2008. Neuroinflammation and disruption in working memory in aged mice after acute stimulation of the peripheral innate immune system. Brain, behavior, and immunity, 22(3), pp.301–11. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2374919&tool=pmcentr ez&rendertype=abstract [Accessed July 5, 2010]. Chen, W. et al., 2004. The glutamate transporter GLT1a is expressed in excitatory axon terminals of mature hippocampal neurons. The Journal of neuroscience : the official journal of the Society for Neuroscience, 24(5), pp.1136–48. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2849838&tool=pmcentr ez&rendertype=abstract [Accessed January 7, 2011]. Cheng, X. et al., 2010. Differential activation of tumor necrosis factor receptors distinguishes between brains from Alzheimer’s disease and non-demented patients. Journal of Alzheimer’s disease : JAD, 19(2), pp.621–30. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20110607 [Accessed December 15, 2012]. Ciruela, F et al., 2006. Heterodimeric adenosine receptors: a device to regulate neurotransmitter release. Cellular and molecular life sciences : CMLS, 63(21), pp.2427–31. Available at: http://www.ncbi.nlm.nih.gov/pubmed/17058035 [Accessed August 24, 2010]. Ciruela, Francisco et al., 2006. Presynaptic control of striatal glutamatergic neurotransmission by adenosine A1-A2A receptor heteromers. The Journal of neuroscience : the official journal of the Society for Neuroscience, 26(7), pp.2080–7. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16481441. Clements, J.D. et al., 1992. The time course of glutamate in the synaptic cleft. Science (New York, N.Y.), 258(5087), pp.1498–501. Available at: http://www.ncbi.nlm.nih.gov/pubmed/1359647 [Accessed February 5, 2011]. Colton, C. a et al., 2006. Expression profiles for macrophage alternative activation genes in AD and in mouse models of AD. Journal of neuroinflammation, 3, p.27. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=1609108&tool=pmcentr ez&rendertype=abstract [Accessed July 25, 2012]. Colton, C.A., 2009. Heterogeneity of microglial activation in the innate immune response in the brain. Journal of neuroimmune pharmacology : the official journal of the Society on NeuroImmune Pharmacology, 4(4), pp.399–418. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2773116&tool=pmcentr ez&rendertype=abstract [Accessed June 16, 2011]. Colton, C.A. & Wilcock, D.M., 2010a. Assessing Activation States in Microglia. CNS & neurological disorders drug targets, 9(2), pp.174–191. Available at: http://apps.isiknowledge.com.proxy.lib.ohio-

218

state.edu/full_record.do?product=WOS&search_mode=GeneralSearch&qid=27&SI D=3D@IDameIEl@cCF3gIJ&page=1&doc=4&cacheurlFromRightClick=no [Accessed October 20, 2010]. Colton, C.A. & Wilcock, D.M., 2010b. Assessing activation states in microglia. CNS & neurological disorders drug targets, 9(2), pp.174–91. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20205642 [Accessed August 10, 2011]. Cooper, N. et al., 2000. Key issues in Alzheimer’s disease inflammation. Neurobiology of aging, 21(3), pp.451–453. Available at: http://www.ncbi.nlm.nih.gov/pubmed/10858594 [Accessed October 20, 2010]. Costenla, Ana R et al., 2011. Enhanced role of adenosine A(2A) receptors in the modulation of LTP in the rat hippocampus upon ageing. The European journal of neuroscience. Available at: http://www.ncbi.nlm.nih.gov/pubmed/21615561 [Accessed June 2, 2011]. Costenla, Ana Rita, Cunha, Rodrigo A & De Mendonça, Alexandre, 2010. Caffeine, adenosine receptors, and synaptic plasticity. Journal of Alzheimer’s disease : JAD, 20 Suppl 1, pp.S25–34. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20182030 [Accessed August 16, 2010]. Cox, F.F. et al., 2011. CD200 fusion protein decreases microglial activation in the hippocampus of aged rats. Brain, behavior, and immunity. Available at: http://www.ncbi.nlm.nih.gov/pubmed/22041297 [Accessed November 9, 2011]. Cox, J.A. et al., 1989. Excitatory amino acid neurotoxicity at the N-methyl-d-aspartate receptor in cultured neurons: role of the voltage-dependent magnesium block, Available at: http://journals.ohiolink.edu/ejc/article.cgi?issn=00068993&issue=v499i0002&articl e=267_eaanatrotvmb [Accessed December 9, 2012]. Cumbo, E. & Ligori, L.D., 2010. Levetiracetam, lamotrigine, and phenobarbital in patients with epileptic seizures and Alzheimer’s disease. Epilepsy & behavior : E&B, 17(4), pp.461–6. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20188634 [Accessed December 1, 2012]. Cunha, R A et al., 1996. Evidence for high-affinity binding sites for the adenosine A2A receptor agonist [3H] CGS 21680 in the rat hippocampus and cerebral cortex that are different from striatal A2A receptors. Naunyn-Schmiedeberg’s archives of pharmacology, 353(3), pp.261–71. Available at: http://www.ncbi.nlm.nih.gov/pubmed/8692280 [Accessed October 20, 2010]. Dabir, D. V et al., 2006. Impaired glutamate transport in a mouse model of tau pathology in astrocytes. The Journal of neuroscience : the official journal of the Society for Neuroscience, 26(2), pp.644–54. Available at: http://apps.isiknowledge.com/full_record.do?product=WOS&search_mode=General Search&qid=13&SID=3C3NnE5do9Nk77LCkPa&page=13&doc=129&cacheurlFro mRightClick=no [Accessed July 21, 2010]. Dai, S.-S. et al., 2010. Local glutamate level dictates adenosine A2A receptor regulation of neuroinflammation and traumatic brain injury. The Journal of neuroscience : the official journal of the Society for Neuroscience, 30(16), pp.5802–10. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20410132 [Accessed July 22, 2010]. Dai, S.-S. & Zhou, Y.-G., 2011. Adenosine 2A receptor: a crucial neuromodulator with bidirectional effect in neuroinflammation and brain injury. Reviews in the

219

neurosciences, 22(2), pp.231–9. Available at: http://www.ncbi.nlm.nih.gov/pubmed/21476942 [Accessed December 10, 2012]. Dall’Igna, O.P. et al., 2007. Caffeine and adenosine A(2a) receptor antagonists prevent beta-amyloid (25-35)-induced cognitive deficits in mice. Experimental neurology, 203(1), pp.241–5. Available at: http://www.ncbi.nlm.nih.gov/pubmed/17007839 [Accessed August 24, 2010]. Danbolt, N C, 2001. Glutamate uptake. Progress in neurobiology, 65(1), pp.1–105. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11369436 [Accessed December 9, 2012]. Dani, J.W., Chernjavsky, A. & Smith, S.J., 1992. Neuronal activity triggers calcium waves in hippocampal astrocyte networks. Neuron, 8(3), pp.429–440. Available at: http://apps.webofknowledge.com/full_record.do?product=WOS&search_mode=Gen eralSearch&qid=35&SID=3BNBmgOLelG2dpbHFi2&page=1&doc=1&cacheurlFr omRightClick=no [Accessed December 9, 2012]. Danysz, W. & Parsons, C.G., 2012. Alzheimer’s disease, β-amyloid, glutamate, NMDA receptors and memantine--searching for the connections. British journal of pharmacology, 167(2), pp.324–52. Available at: http://apps.webofknowledge.com/full_record.do?product=WOS&search_mode=Cita tionReport&qid=39&SID=2FE8IIAddNG@3jAk3CG&page=10&doc=98&cacheurl FromRightClick=no [Accessed October 26, 2012]. Daré, E. et al., 2007. Modulation of glial cell functions by adenosine receptors. Physiology & Behavior, pp.1–6. Deuchars, S.A., Brooke, R.E. & Deuchars, J., 2001. Adenosine A1 receptors reduce release from excitatory but not inhibitory synaptic inputs onto lateral horn neurons. The Journal of neuroscience : the official journal of the Society for Neuroscience, 21(16), pp.6308–20. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11487654 [Accessed October 20, 2010]. Duarte, J.M.N. et al., 2006. Modification of adenosine A1 and A2A receptor density in the hippocampus of streptozotocin-induced diabetic rats. Neurochemistry international, 48(2), pp.144–50. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16256246 [Accessed August 24, 2010]. Dunlop, J. et al., 2003. Impaired spinal cord glutamate transport capacity and reduced sensitivity to riluzole in a transgenic superoxide dismutase mutant rat model of amyotrophic lateral sclerosis. The Journal of neuroscience : the official journal of the Society for Neuroscience, 23(5), pp.1688–96. Available at: http://www.ncbi.nlm.nih.gov/pubmed/12629173 [Accessed January 7, 2011]. Durafourt, B. a et al., 2012. Comparison of polarization properties of human adult microglia and blood-derived macrophages. Glia, 727(January), pp.717–727. Available at: http://www.ncbi.nlm.nih.gov/pubmed/22290798 [Accessed March 12, 2012]. D’Amato, R., Wenk, GL & Snyder, S., 1987. Aminergic systems in Alzheimer’s disease and Parkinson's disease. Ann Neurol, 22, pp.229–236. D’Andrea, M.R., Cole, Gregory M & Ard, M.D., 2004. The microglial phagocytic role with specific plaque types in the Alzheimer disease brain. Neurobiology of aging, 25(5), pp.675–83. Available at: http://apps.webofknowledge.com.proxy.lib.ohio- state.edu/full_record.do?product=WOS&search_mode=CitationReport&qid=3&SID

220

=1A263P6PcLNnKcD6kL@&page=3&doc=30&cacheurlFromRightClick=no [Accessed December 19, 2012]. Edison, Paul et al., 2008a. Microglia, amyloid, and cognition in Alzheimer’s disease: An [11C](R)PK11195-PET and [11C]PIB-PET study. Neurobiology of disease, 32(3), pp.412–9. Available at: http://www.ncbi.nlm.nih.gov/pubmed/18786637 [Accessed August 16, 2010]. Edison, Paul et al., 2008b. Microglia, amyloid, and cognition in Alzheimer’s disease: An [11C](R)PK11195-PET and [11C]PIB-PET study. Neurobiology of disease, 32(3), pp.412–9. Available at: http://www.ncbi.nlm.nih.gov/pubmed/18786637 [Accessed August 16, 2010]. Emre, M. et al., 2010. Memantine for patients with Parkinson’s disease dementia or dementia with Lewy bodies: a randomised, double-blind, placebo-controlled trial. The Lancet Neurology. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20729148 [Accessed September 1, 2010]. Eskelinen, M.H. et al., 2009. Midlife coffee and tea drinking and the risk of late-life dementia: a population-based CAIDE study. Journal of Alzheimer’s disease : JAD, 16(1), pp.85–91. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19158424. Fayed, N. et al., 2011. Brain Glutamate Levels Are Decreased in Alzheimer’s Disease: A Magnetic Resonance Spectroscopy Study. American journal of Alzheimer’s disease and other dementias. Available at: http://www.ncbi.nlm.nih.gov/pubmed/21921084 [Accessed September 21, 2011]. Featherstone, D.E. & Shippy, S.A., 2008. Regulation of synaptic transmission by ambient extracellular glutamate. The Neuroscientist : a review journal bringing neurobiology, neurology and psychiatry, 14(2), pp.171–81. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2322853&tool=pmcentr ez&rendertype=abstract [Accessed December 11, 2012]. Fenn, A.M. et al., 2011. Lipopolysaccharide-induced interleukin (IL)-4 receptor-α expression and corresponding sensitivity to the M2 promoting effects of IL-4 are impaired in microglia of aged mice. Brain, behavior, and immunity. Available at: http://www.ncbi.nlm.nih.gov/pubmed/22024136 [Accessed December 21, 2011]. Ferré, S et al., 2001. Adenosine/dopamine interaction: implications for the treatment of Parkinson’s disease. Parkinsonism & related disorders, 7(3), pp.235–241. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11331192 [Accessed September 28, 2010]. Ferré, Sergi, 2008. An update on the mechanisms of the psychostimulant effects of caffeine. Journal of neurochemistry, 105(4), pp.1067–79. Available at: http://www.ncbi.nlm.nih.gov/pubmed/18088379 [Accessed January 22, 2011]. Fiebich, B L et al., 1996. Cyclooxygenase-2 expression in rat microglia is induced by adenosine A2a-receptors. Glia, 18(2), pp.152–60. Available at: http://www.ncbi.nlm.nih.gov/pubmed/8913778 [Accessed September 24, 2010]. Fiebich, B L et al., 2000. Effects of caffeine and paracetamol alone or in combination with acetylsalicylic acid on prostaglandin E(2) synthesis in rat microglial cells. Neuropharmacology, 39(11), pp.2205–13. Available at: http://www.ncbi.nlm.nih.gov/pubmed/10963764. Flanary, B.E. et al., 2007. Evidence that aging and amyloid promote microglial cell senescence. Rejuvenation research, 10(1), pp.61–74. Available at: http://www.ncbi.nlm.nih.gov/pubmed/17378753 [Accessed July 25, 2012].

221

Floden, A.M., Li, Shanshan & Combs, C.K., 2005. Beta-amyloid-stimulated microglia induce neuron death via synergistic stimulation of tumor necrosis factor alpha and NMDA receptors. The Journal of neuroscience : the official journal of the Society for Neuroscience, 25(10), pp.2566–75. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15758166 [Accessed November 17, 2012]. Fogal, B. et al., 2007. System x(c)- activity and astrocytes are necessary for interleukin-1 beta-mediated hypoxic neuronal injury. The Journal of neuroscience : the official journal of the Society for Neuroscience, 27(38), pp.10094–105. Available at: http://www.ncbi.nlm.nih.gov/pubmed/17881516 [Accessed December 18, 2012]. Fontinha, B.M. et al., 2009. Adenosine A(2A) receptor modulation of hippocampal CA3- CA1 synapse plasticity during associative learning in behaving mice. Neuropsychopharmacology : official publication of the American College of Neuropsychopharmacology, 34(7), pp.1865–74. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19212319 [Accessed September 1, 2010]. Frank, M.G., Barrientos, R.M., Watkins, L.R., et al., 2010. Aging sensitizes rapidly isolated hippocampal microglia to LPS ex vivo. Journal of neuroimmunology, 226(1-2), pp.181–4. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2937085&tool=pmcentr ez&rendertype=abstract [Accessed October 13, 2010]. Frank, M.G., Barrientos, R.M., Hein, A.M., et al., 2010. IL-1RA blocks E. coli-induced suppression of Arc and long-term memory in aged F344xBN F1 rats. Brain, behavior, and immunity, 24(2), pp.254–62. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2818379&tool=pmcentr ez&rendertype=abstract [Accessed October 26, 2012]. Frank, M.G., Barrientos, R.M., et al., 2006. mRNA up-regulation of MHC II and pivotal pro-inflammatory genes in normal brain aging. Neurobiology of aging, 27(5), pp.717–22. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15890435 [Accessed August 24, 2010]. Frank, M.G., Wieseler-Frank, J.L., et al., 2006. Rapid isolation of highly enriched and quiescent microglia from adult rat hippocampus: immunophenotypic and functional characteristics. Journal of neuroscience methods, 151(2), pp.121–30. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16125247 [Accessed July 25, 2012]. Fredholm, Bertil B et al., 1999. Actions of Caffeine in the Brain with Special Reference to Factors That Contribute to Its Widespread Use. Pharmacology, 51(1). Fredholm, Bertil B, 2004. Connection between caffeine, adenosine receptors and dopamine. Coffee reduces the risk of Parkinson disease. Läkartidningen, 101(34), pp.2552–5. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15446691 [Accessed October 20, 2010]. Fredholm, Bertil B & Svenningsson, Per, 2003. Adenosine-dopamine interactions: development of a concept and some comments on therapeutic possibilities. Neurology, 61(11 Suppl 6), pp.S5–9. Available at: http://www.ncbi.nlm.nih.gov/pubmed/14663001 [Accessed September 28, 2010]. Frizzo, M.E. dos S. et al., 2004. Riluzole enhances glutamate uptake in rat astrocyte cultures. Cellular and molecular neurobiology, 24(1), pp.123–8. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15049516 [Accessed January 8, 2012].

222

Fukumoto, H., Asami-Odaka, A., Suzuki, N., Shimada, H., et al., 1996. Amyloid beta protein deposition in normal aging has the same characteristics as that in Alzheimer’s disease. Predominance of A beta 42(43) and association of A beta 40 with cored plaques. The American journal of pathology, 148(1), pp.259–65. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=1861616&tool=pmcentr ez&rendertype=abstract [Accessed November 25, 2012]. Fukumoto, H., Asami-Odaka, A., Suzuki, N. & Iwatsubo, T., 1996. Association of A beta 40-positive senile plaques with microglial cells in the brains of patients with Alzheimer’s disease and in non-demented aged individuals. Neurodegeneration : a journal for neurodegenerative disorders, neuroprotection, and neuroregeneration, 5(1), pp.13–7. Available at: http://www.ncbi.nlm.nih.gov/pubmed/8731377 [Accessed December 18, 2012]. Gao, H.-M. et al., 2011. Neuroinflammation and α-Synuclein Dysfunction Potentiate Each Other Driving Chronic Progression of Neurodegeneration in a Mouse Model of Parkinson’s Disease. Environmental health perspectives. Available at: http://www.ncbi.nlm.nih.gov/pubmed/21245015 [Accessed January 21, 2011]. Gebicke-haerter, 1996. Both Adenosine A1- and A2-receptors are required to stimulate microglial proliferation.pdf. Van Gelder, B.M. et al., 2007. Coffee consumption is inversely associated with cognitive decline in elderly European men: the FINE Study. European journal of clinical nutrition, 61(2), pp.226–32. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16929246 [Accessed January 22, 2011]. Ginhoux, F. et al., 2010. Fate Mapping Analysis Reveals That Adult Microglia Derive from Primitive Macrophages. Science (New York, N.Y.), 841(2010). Available at: http://www.ncbi.nlm.nih.gov/pubmed/20966214 [Accessed October 24, 2010]. Godbout, J P et al., 2005. Exaggerated neuroinflammation and sickness behavior in aged mice after activation of the peripheral innate immune system. The FASEB Journal, 24(1), pp.1–24. Godbout, Jonathan P & Johnson, Rodney W, 2009. Age and neuroinflammation: a lifetime of psychoneuroimmune consequences. Immunology and allergy clinics of North America, 29(2), pp.321–37. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19389585 [Accessed August 24, 2010]. Golembiowska, K., 2004. Effect of the adenosine A2A receptor antagonist 8-(3- chlorostyryl)caffeine on ?-DOPA biotransformation in rat striatum. Brain Research, 998(2), pp.208–217. Available at: http://linkinghub.elsevier.com/retrieve/pii/S0006899303040678 [Accessed August 24, 2010]. Gordon, S., 2003. Alternative activation of macrophages. Nature reviews. Immunology, 3(1), pp.23–35. Available at: http://www.ncbi.nlm.nih.gov/pubmed/12511873 [Accessed February 29, 2012]. Goshen, I. et al., A dual role for interleukin-1 in hippocampal-dependent memory processes. Psychoneuroendocrinology, 32(8-10), pp.1106–15. Available at: http://www.ncbi.nlm.nih.gov/pubmed/17976923 [Accessed October 9, 2012].

223

Graeber, M. B., 2010. Changing Face of Microglia. Science, 330(6005), pp.783–788. Available at: http://www.sciencemag.org/cgi/doi/10.1126/science.1190929 [Accessed November 4, 2010]. Graeber, Manuel B & Streit, Wolfgang J, 2010. Microglia: biology and pathology. Acta neuropathologica, 119(1), pp.89–105. Available at: http://apps.isiknowledge.com.proxy.lib.ohio- state.edu/full_record.do?product=WOS&search_mode=GeneralSearch&qid=14&SI D=3D@IDameIEl@cCF3gIJ&page=1&doc=1&cacheurlFromRightClick=no [Accessed August 16, 2010]. Griffin, W.S. et al., 1998. Glial-neuronal interactions in Alzheimer’s disease: the potential role of a “cytokine cycle” in disease progression. Brain pathology (Zurich, Switzerland), 8(1), pp.65–72. Available at: http://www.ncbi.nlm.nih.gov/pubmed/9458167 [Accessed December 18, 2012]. De Haas, A.H., Boddeke, H.W.G.M. & Biber, Knut, 2008. Region-specific expression of immunoregulatory proteins on microglia in the healthy CNS. Glia, 56(8), pp.888– 94. Available at: http://www.ncbi.nlm.nih.gov/pubmed/18338796 [Accessed March 21, 2012]. Halldner, L. et al., 2004. The adenosine A1 receptor contributes to the stimulatory, but not the inhibitory effect of caffeine on locomotion: a study in mice lacking adenosine A1 and/or A2A receptors. Neuropharmacology, 46(7), pp.1008–17. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15081797 [Accessed September 1, 2010]. Hammarberg, C., Schulte, G. & Fredholm, Bertil B, 2003. Evidence for functional adenosine A3 receptors in microglia cells. Journal of neurochemistry, 86(4), pp.1051–4. Available at: http://www.ncbi.nlm.nih.gov/pubmed/12887702 [Accessed September 24, 2010]. Hardingham, G.E. & Bading, H., 2010. Synaptic versus extrasynaptic NMDA receptor signalling: implications for neurodegenerative disorders. Nature reviews. Neuroscience. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20842175 [Accessed September 16, 2010]. Hargus, N.J., Bertram, E.H. & Patel, M.K., 2009. Adenosine A1 receptors presynaptically modulate excitatory synaptic input onto subiculum neurons. Brain research, 1280, pp.60–8. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2731234&tool=pmcentr ez&rendertype=abstract [Accessed October 20, 2010]. Harris, A.Z. & Pettit, D.L., 2007. Extrasynaptic and synaptic NMDA receptors form stable and uniform pools in rat hippocampal slices. The Journal of physiology, 584(Pt 2), pp.509–19. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2277145&tool=pmcentr ez&rendertype=abstract [Accessed October 26, 2012]. Haskó, G. et al., 2005. Adenosine receptor signaling in the brain immune system. Trends in pharmacological sciences, 26(10), pp.511–6. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2228262&tool=pmcentr ez&rendertype=abstract [Accessed December 10, 2012]. Hassinger, T.D. et al., 1995. Evidence for glutamate-mediated activation of hippocampal neurons by glial calcium waves. Journal of neurobiology, 28(2), pp.159–70.

224

Available at: http://www.ncbi.nlm.nih.gov/pubmed/8537822 [Accessed December 9, 2012]. Hauss-Wegrzyniak, B. et al., 1998. Chronic neuroinflammation in rats reproduces components of the neurobiology of Alzheimer ’ s disease. , pp.294–303. Hauss-Wegrzyniak, B.H. et al., 2002. Chronic Brain Inflammation Results in Cell Loss in the Entorhinal Cortex and Impaired LTP in Perforant Path-Granule Cell Synapses. Experimental Neurology, 341, pp.336 –341. Henry, Christopher J et al., 2009. Peripheral lipopolysaccharide (LPS) challenge promotes microglial hyperactivity in aged mice that is associated with exaggerated induction of both pro-inflammatory IL-1beta and anti-inflammatory IL-10 cytokines. Brain, behavior, and immunity, 23(3), pp.309–17. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2692986&tool=pmcentr ez&rendertype=abstract [Accessed July 15, 2010]. Heo, S. et al., 2011. Hippocampal glutamate transporter 1 (GLT-1) complex levels are paralleling memory training in the Multiple T-Maze in C57BL/6J mice. Brain structure & function. Available at: http://www.ncbi.nlm.nih.gov/pubmed/22113856 [Accessed November 30, 2011]. Hermans, E., 2007. Review Neuroinflammation and Regulation of Glial Glutamate Uptake in Neurological Disorders. Journal of Neuroscience Research, 2070, pp.2059–2070. Hernán, Miguel A et al., 2002. A meta-analysis of coffee drinking, cigarette smoking, and the risk of Parkinson’s disease. Annals of neurology, 52(3), pp.276–84. Available at: http://www.ncbi.nlm.nih.gov/pubmed/12205639 [Accessed December 10, 2012]. Hickman, S.E., Allison, E.K. & El Khoury, J., 2008. Microglial dysfunction and defective beta-amyloid clearance pathways in aging Alzheimer’s disease mice. The Journal of neuroscience : the official journal of the Society for Neuroscience, 28(33), pp.8354–60. Available at: http://www.ncbi.nlm.nih.gov/pubmed/18701698. Hoarau, J.-J. et al., 2010. Activation and Control of CNS Innate Immune Responses in Health and Diseases: A Balancing Act Finely Tuned by Neuroimmune Regulators (NIReg). CNS & neurological disorders drug targets. Available at: http://www.ncbi.nlm.nih.gov/pubmed/21143144 [Accessed December 20, 2010]. Hoozemans, J.J.M. et al., 2006. Neuroinflammation and regeneration in the early stages of Alzheimer’s disease pathology. International journal of developmental neuroscience : the official journal of the International Society for Developmental Neuroscience, 24(2-3), pp.157–65. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16384684 [Accessed August 23, 2010]. Hu, G. et al., 2007. Coffee and tea consumption and the risk of Parkinson’s disease. Movement disorders : official journal of the Movement Disorder Society, 22(15), pp.2242–8. Available at: http://www.ncbi.nlm.nih.gov/pubmed/17712848 [Accessed August 24, 2010]. Hu, S. et al., 2000. Cytokine effects on glutamate uptake by human astrocytes. Neuroimmunomodulation, 7(3), pp.153–9. Available at: http://www.ncbi.nlm.nih.gov/pubmed/10754403 [Accessed March 30, 2012].

225

Hu, W. et al., 2007. Development of a novel therapeutic suppressor of brain proinflammatory cytokine up-regulation that attenuates synaptic dysfunction and behavioral deficits. , 17, pp.414–418. Huang, Y et al., 2008. Exaggerated sickness behavior and brain proinflammatory cytokine expression in aged mice in response to intracerebroventricular lipopolysaccharide. Neurobiology of aging, 29(11), pp.1744–53. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2647751&tool=pmcentr ez&rendertype=abstract [Accessed August 24, 2010]. Huang, Yadong & Mucke, L., 2012. Alzheimer mechanisms and therapeutic strategies. Cell, 148(6), pp.1204–22. Available at: http://www.ncbi.nlm.nih.gov/pubmed/22424230 [Accessed October 10, 2012]. Hynd, M.R., Scott, H.L. & Dodd, P.R., 2004a. Differential expression of N-methyl-D- aspartate receptor NR2 isoforms in Alzheimer’s disease. Journal of neurochemistry, 90(4), pp.913–9. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15287897 [Accessed October 30, 2012]. Hynd, M.R., Scott, H.L. & Dodd, P.R., 2004b. Glutamate-mediated excitotoxicity and neurodegeneration in Alzheimer’s disease. Neurochemistry international, 45(5), pp.583–95. Available at: http://apps.webofknowledge.com/full_record.do?product=WOS&search_mode=Cita tionReport&qid=18&SID=2FE8IIAddNG@3jAk3CG&page=3&doc=29&cacheurlF romRightClick=no [Accessed October 30, 2012]. Hynd, M.R., Scott, H.L. & Dodd, P.R., 2004c. Glutamate-mediated excitotoxicity and neurodegeneration in Alzheimer’s disease. Neurochemistry international, 45(5), pp.583–95. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15234100 [Accessed October 30, 2012]. Hynd, M.R., Scott, H.L. & Dodd, P.R., 2004d. Selective loss of NMDA receptor NR1 subunit isoforms in Alzheimer’s disease. Journal of neurochemistry, 89(1), pp.240– 7. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15030408 [Accessed October 30, 2012]. Imamura, K. et al., 2003. Distribution of major histocompatibility complex class II- positive microglia and cytokine profile of Parkinson’s disease brains. Acta neuropathologica, 106(6), pp.518–26. Available at: http://www.ncbi.nlm.nih.gov/pubmed/14513261 [Accessed December 7, 2012]. Izumi, Y. et al., 2002. Glutamate transporters and retinal excitotoxicity. Glia, 39(1), pp.58–68. Available at: http://www.ncbi.nlm.nih.gov/pubmed/12112376 [Accessed May 22, 2012]. Jacob, C.P. et al., 2007. Alterations in expression of glutamatergic transporters and receptors in sporadic Alzheimer’s disease. Journal of Alzheimer’s disease : JAD, 11(1), pp.97–116. Available at: http://www.ncbi.nlm.nih.gov/pubmed/17361039 [Accessed January 7, 2011]. Kalda, A. et al., 2006. Novel neuroprotection by caffeine and adenosine A(2A) receptor antagonists in animal models of Parkinson’s disease. Journal of the neurological sciences, 248(1-2), pp.9–15. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16806272. Kambe, Y. et al., 2010. Induced tolerance to glutamate neurotoxicity through downregulation of NR2 subunits of N-methyl-D-aspartate receptors in cultured rat striatal

226

neurons. Journal of neuroscience research, 88(10), pp.2177–87. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20336776 [Accessed October 27, 2010]. Kamikawa, H. et al., 1998. IL-1beta increases norepinephrine level in rat frontal cortex: involvement of prostanoids, NO, and glutamate. The American journal of physiology, 275(3 Pt 2), pp.R803–10. Available at: http://www.ncbi.nlm.nih.gov/pubmed/9728078 [Accessed December 18, 2012]. Kanda, T. et al., 1998. Adenosine A2A antagonist: a novel antiparkinsonian agent that does not provoke dyskinesia in parkinsonian monkeys. Annals of neurology, 43(4), pp.507–13. Available at: http://www.ncbi.nlm.nih.gov/pubmed/9546333 [Accessed September 24, 2010]. Kandinov, B., Giladi, N. & Korczyn, A.D., 2007. The effect of cigarette smoking, tea, and coffee consumption on the progression of Parkinson’s disease. Parkinsonism & related disorders, 13(4), pp.243–5. Available at: http://www.ncbi.nlm.nih.gov/pubmed/17275394 [Accessed August 24, 2010]. Kaufer, D.I. & Tröster, A.I., 2008. Neuropsychology and Behavioral Neurology, Elsevier. Available at: http://www.ncbi.nlm.nih.gov/pubmed/18631712 [Accessed September 24, 2010]. Kelly-Welch, A.E. et al., 2003. Interleukin-4 and interleukin-13 signaling connections maps. Science (New York, N.Y.), 300(5625), pp.1527–8. Available at: http://apps.webofknowledge.com.proxy.lib.ohio- state.edu/full_record.do?product=WOS&search_mode=CitationReport&qid=20&SI D=3AAOpbeEIk9j5mBeH53&page=1&doc=3&cacheurlFromRightClick=no [Accessed November 27, 2012]. Kettenmann, H. & Ransom, B.R., 2005. Neuroglia, Oxford University Press. Available at: http://books.google.com/books?id=gIP0FVc7j_cC&pgis=1 [Accessed January 30, 2011]. Kitagawa, M., Houzen, H. & Tashiro, K., 2007. Effects of caffeine on the freezing of gait in Parkinson’s disease. Movement disorders : official journal of the Movement Disorder Society, 22(5), pp.710–2. Available at: http://www.ncbi.nlm.nih.gov/pubmed/17373724 [Accessed August 24, 2010]. Klausnitzer, J. & Manahan-Vaughan, D., 2008. Frequency facilitation at mossy fiber- CA3 synapses of freely behaving rats is regulated by adenosine A1 receptors. The Journal of neuroscience : the official journal of the Society for Neuroscience, 28(18), pp.4836–40. Available at: http://www.ncbi.nlm.nih.gov/pubmed/18448660 [Accessed September 2, 2010]. Koning, N. et al., 2009. Distribution of the immune inhibitory molecules CD200 and CD200R in the normal central nervous system and multiple sclerosis lesions suggests neuron-glia and glia-glia interactions. Journal of neuropathology and experimental neurology, 68(2), pp.159–67. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19151626. Koscsó, B. et al., 2012. Adenosine augments IL-10 production by microglial cells through an A2B adenosine receptor-mediated process. Journal of immunology (Baltimore, Md. : 1950), 188(1), pp.445–53. Available at: http://www.ncbi.nlm.nih.gov/pubmed/22116830 [Accessed December 10, 2012]. Lai, A.Y. et al., 2006. Interleukin-1 beta modulates AMPA receptor expression and phosphorylation in hippocampal neurons. Journal of neuroimmunology, 175(1-2),

227

pp.97–106. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16626814 [Accessed November 7, 2012]. Lauderback, C.M. et al., 2001a. The glial glutamate transporter, GLT-1, is oxidatively modified by 4-hydroxy-2-nonenal in the Alzheimer’s disease brain: the role of Aβ1- 42. Journal of Neurochemistry, 78(2), pp.413–416. Available at: http://doi.wiley.com/10.1046/j.1471-4159.2001.00451.x. Lauderback, C.M. et al., 2001b. The glial glutamate transporter, GLT-1, is oxidatively modified by 4-hydroxy-2-nonenal in the Alzheimer’s disease brain: the role of Aβ1- 42. Journal of Neurochemistry, 78(2), pp.413–416. Available at: http://doi.wiley.com/10.1046/j.1471-4159.2001.00451.x. Lawson, 1990. Heterogeneity in the distribution and morphology of microglia in the normal adult mouse brain. Lee, C.Y.D. & Landreth, G.E., 2010. The role of microglia in amyloid clearance from the AD brain. Journal of neural transmission (Vienna, Austria : 1996), 117(8), pp.949– 60. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20552234 [Accessed July 29, 2010]. Lee, E.-J. et al., 2010. alpha-Synuclein activates microglia by inducing the expressions of matrix metalloproteinases and the subsequent activation of protease-activated receptor-1. Journal of immunology (Baltimore, Md. : 1950), 185(1), pp.615–23. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20511551 [Accessed September 1, 2010]. LEE, S. et al., CYTOKINE PRODUCTION BY HUMAN FETAL MICROGLIA AND ASTROCYTES - DIFFERENTIAL INDUCTION BY LIPOPOLYSACCHARIDE AND IL-1-BETA . Available at: http://apps.webofknowledge.com/full_record.do?product=WOS&search_mode=Cita tionReport&qid=4&SID=3BNBmgOLelG2dpbHFi2&page=1&doc=1&cacheurlFro mRightClick=no [Accessed December 9, 2012]. Lehre, K P et al., 1995. Differential expression of two glial glutamate transporters in the rat brain: quantitative and immunocytochemical observations. The Journal of neuroscience : the official journal of the Society for Neuroscience, 15(3 Pt 1), pp.1835–53. Available at: http://www.ncbi.nlm.nih.gov/pubmed/7891138 [Accessed January 7, 2011]. Leoutsakos, J.-M.S. et al., 2012. Effects of non-steroidal anti-inflammatory drug treatments on cognitive decline vary by phase of pre-clinical Alzheimer disease: findings from the randomized controlled Alzheimer’s Disease Anti-inflammatory Prevention Trial. International journal of geriatric psychiatry, 27(4), pp.364–74. Available at: http://www.ncbi.nlm.nih.gov/pubmed/21560159 [Accessed December 15, 2012]. Leshem-Lev, D. et al., 2010. Adenosine A(1) and A (3) receptor agonists reduce hypoxic injury through the involvement of P38 MAPK. Molecular and cellular biochemistry, 345(1-2), pp.153–60. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20730620 [Accessed December 10, 2012]. Lewandowska, Eliza et al., 2004. Ultrastructural evaluation of activated forms of microglia in human brain in selected neurological diseases (SSPE, Wilson’s disease and Alzheimer's disease). Folia neuropathologica / Association of Polish Neuropathologists and Medical Research Centre, Polish Academy of Sciences,

228

42(2), pp.81–91. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15266782 [Accessed December 19, 2012]. Lewerenz, J. et al., 2012. The Cystine/Glutamate Antiporter System x(c)(-) in Health and Disease: From Molecular Mechanisms to Novel Therapeutic Opportunities. Antioxidants & redox signaling. Available at: http://www.ncbi.nlm.nih.gov/pubmed/22667998 [Accessed December 11, 2012]. Lewerenz, J., Klein, M. & Methner, A., 2006. Cooperative action of glutamate transporters and cystine/glutamate antiporter system Xc- protects from oxidative glutamate toxicity. Journal of neurochemistry, 98(3), pp.916–25. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16771835 [Accessed January 30, 2011]. Li, R. et al., 2004. Tumor necrosis factor death receptor signaling cascade is required for amyloid-beta protein-induced neuron death. The Journal of neuroscience : the official journal of the Society for Neuroscience, 24(7), pp.1760–71. Available at: http://www.ncbi.nlm.nih.gov/pubmed/14973251 [Accessed December 15, 2012]. Li, S et al., 1997. Glutamate transporter alterations in Alzheimer disease are possibly associated with abnormal APP expression. Journal of neuropathology and experimental neurology, 56(8), pp.901–11. Available at: http://www.ncbi.nlm.nih.gov/pubmed/9258260 [Accessed January 7, 2011]. Lin, A.P. et al., 2003. Reduced glutamate neurotransmission in patients with Alzheimer’s disease -- an in vivo (13)C magnetic resonance spectroscopy study. Magma (New York, N.Y.), 16(1), pp.29–42. Available at: http://www.ncbi.nlm.nih.gov/pubmed/12695884 [Accessed March 10, 2011]. Lindsay, J. et al., 2002. Risk factors for Alzheimer’s disease: a prospective analysis from the Canadian Study of Health and Aging. American journal of epidemiology, 156(5), pp.445–53. Available at: http://www.ncbi.nlm.nih.gov/pubmed/12196314 [Accessed September 24, 2010]. Lippa, C.F. et al., 2010. Levetiracetam: a practical option for seizure management in elderly patients with cognitive impairment. American journal of Alzheimer’s disease and other dementias, 25(2), pp.149–54. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19001351 [Accessed December 1, 2012]. Liu, Ling et al., 2011. Apolipoprotein E expression is elevated by interleukin 1 and other interleukin 1-induced factors. Journal of neuroinflammation, 8, p.175. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3286434&tool=pmcentr ez&rendertype=abstract [Accessed December 5, 2012]. Liu, Yang et al., 2005. LPS receptor (CD14): a receptor for phagocytosis of Alzheimer’s amyloid peptide. Brain : a journal of neurology, 128(Pt 8), pp.1778–89. Available at: http://apps.webofknowledge.com.proxy.lib.ohio- state.edu/full_record.do?product=WOS&search_mode=CitationReport&qid=3&SID =1A263P6PcLNnKcD6kL@&page=2&doc=17&cacheurlFromRightClick=no [Accessed November 17, 2012]. Liu, Yitao et al., 2007. NMDA receptor subunits have differential roles in mediating excitotoxic neuronal death both in vitro and in vivo. The Journal of neuroscience : the official journal of the Society for Neuroscience, 27(11), pp.2846–57. Available at: http://www.ncbi.nlm.nih.gov/pubmed/17360906 [Accessed November 12, 2012]. LOUGHLIN, A., WOODROOFE, M. & CUZNER, M., 1993. MODULATION OF INTERFERON-GAMMA-INDUCED MAJOR HISTOCOMPATIBILITY

229

COMPLEX CLASS-II AND FC-RECEPTOR EXPRESSION ON ISOLATED MICROGLIA BY TRANSFORMING GROWTH-FACTOR-BETA-1, INTERLEUKIN-4, NORADRENALINE AND GLUCOCORTICOIDS. Available at: http://apps.webofknowledge.com/full_record.do?product=WOS&search_mode=Cita tionReport&qid=10&SID=3BNBmgOLelG2dpbHFi2&page=1&doc=7&cacheurlFr omRightClick=no [Accessed December 9, 2012]. Lucin, K.M. & Wyss-Coray, Tony, 2009a. Immune activation in brain aging and neurodegeneration: too much or too little? Neuron, 64(1), pp.110–22. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2834890&tool=pmcentr ez&rendertype=abstract [Accessed October 20, 2010]. Lucin, K.M. & Wyss-Coray, Tony, 2009b. Immune activation in brain aging and neurodegeneration: too much or too little? Neuron, 64(1), pp.110–22. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19840553 [Accessed October 20, 2010]. Lue, 1996. Inflammation, A beta deposition, and neurofibrillary tangle formation as correlates of Alzheimer’s disease neurodegeneration.pdf. Lue, L F et al., 2001. Involvement of microglial receptor for advanced glycation endproducts (RAGE) in Alzheimer’s disease: identification of a cellular activation mechanism. Experimental neurology, 171(1), pp.29–45. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11520119 [Accessed December 13, 2012]. Lue, Lih-fen et al., 2001. Inflammatory Repertoire of Alzheimer ’ s Disease and Nondemented Elderly Microglia In Vitro. American Journal Of Pathology, 79(June), pp.72–79. Lynch, M. a, 2009. The multifaceted profile of activated microglia. Molecular neurobiology, 40(2), pp.139–56. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19629762 [Accessed August 13, 2010]. Lyons, A. et al., 2007. CD200 ligand receptor interaction modulates microglial activation in vivo and in vitro: a role for IL-4. The Journal of neuroscience : the official journal of the Society for Neuroscience, 27(31), pp.8309–13. Available at: http://www.ncbi.nlm.nih.gov/pubmed/17670977 [Accessed September 29, 2011]. Lyons, A. et al., 2009. Decreased neuronal CD200 expression in IL-4-deficient mice results in increased neuroinflammation in response to lipopolysaccharide. Brain, behavior, and immunity, 23(7), pp.1020–7. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19501645 [Accessed August 24, 2010]. Maher, Pamela, 2005. The effects of stress and aging on glutathione metabolism. Ageing research reviews, 4(2), pp.288–314. Available at: http://apps.webofknowledge.com/full_record.do?product=WOS&search_mode=Cita tionReport&qid=29&SID=2FE8IIAddNG@3jAk3CG&page=5&doc=47&cacheurlF romRightClick=no [Accessed November 22, 2012]. Maia, L. & De Mendonça, A, 2002. Does caffeine intake protect from Alzheimer’s disease? European journal of neurology : the official journal of the European Federation of Neurological Societies, 9(4), pp.377–82. Available at: http://www.ncbi.nlm.nih.gov/pubmed/12099922 [Accessed November 8, 2012]. Malarkey, E.B. & Parpura, Vladimir, 2008. Mechanisms of glutamate release from astrocytes. Neurochemistry international, 52(1-2), pp.142–54. Available at: http://www.ncbi.nlm.nih.gov/pubmed/17669556 [Accessed February 6, 2011].

230

Malgouris, C. et al., 1989. Riluzole, a novel antiglutamate, prevents memory loss and hippocampal neuronal damage in ischemic gerbils. The Journal of neuroscience : the official journal of the Society for Neuroscience, 9(11), pp.3720–7. Available at: http://www.ncbi.nlm.nih.gov/pubmed/2585051 [Accessed January 8, 2012]. Malva, J.O., Silva, A.P. & Cunha, Rodrigo A, 2003. Presynaptic Modulation Controlling Neuronal Excitability and Epileptogenesis : Role of Kainate , Adenosine and Neuropeptide Y Receptors *. October, 28(10), pp.1501–1515. Mandrekar-Colucci, S. & Landreth, G.E., 2010. Microglia and inflammation in Alzheimer’s disease. CNS & neurological disorders drug targets, 9(2), pp.156–67. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20205644 [Accessed August 17, 2010]. Maragakis, N.J., Dykes-Hoberg, M. & Rothstein, J.D., 2004. Altered expression of the glutamate transporter EAAT2b in neurological disease. Annals of neurology, 55(4), pp.469–77. Available at: http://apps.isiknowledge.com/full_record.do?product=WOS&search_mode=General Search&qid=13&SID=3C3NnE5do9Nk77LCkPa&page=18&doc=171&cacheurlFro mRightClick=no [Accessed December 2, 2010]. Maragakis, N.J. & Rothstein, J.D., 2004. Glutamate transporters: animal models to neurologic disease. Neurobiology of disease, 15(3), pp.461–73. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15056453 [Accessed July 24, 2010]. Marchalant, Y, Brothers, H M & Wenk, G L, 2009a. Cannabinoid agonist WIN-55,212-2 partially restores neurogenesis in the aged rat brain. Molecular psychiatry, 14(12), pp.1068–9. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19920834 [Accessed August 24, 2010]. Marchalant, Y, Brothers, H M & Wenk, G L, 2009b. New neuron production can be increased in the hippocampus of aged rats following cannabinoid treatment. Molecular psychiatry, 14(12), pp.1067, 1068–9. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19920833 [Accessed August 24, 2010]. Marchalant, Y, Rosi, S & Wenk, G L, 2007. Anti-inflammatory property of the cannabinoid agonist WIN-55212-2 in a rodent model of chronic brain inflammation. Neuroscience, 144(4), pp.1516–22. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=1852513&tool=pmcentr ez&rendertype=abstract [Accessed August 24, 2010]. Marchalant, Yannick et al., 2008. Cannabinoid receptor stimulation is anti-inflammatory and improves memory in old rats. Neurobiology of aging, 29(12), pp.1894–901. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2586121&tool=pmcentr ez&rendertype=abstract [Accessed August 24, 2010]. Marchalant, Yannick et al., 2009. Cannabinoids attenuate the effects of aging upon neuroinflammation and neurogenesis☆. Neurobiology of Disease, 34(2), pp.300– 307. Available at: http://linkinghub.elsevier.com/retrieve/pii/S0969996109000254 [Accessed August 24, 2010]. Marchalant, Yannick et al., 2009. Neurobiology of Disease Cannabinoids attenuate the effects of aging upon neuroinflammation and neurogenesis ☆. Neurobiology of Disease, 34(2), pp.300–307. Available at: http://dx.doi.org/10.1016/j.nbd.2009.01.014.

231

Marcyniuk, B. et al., 1988. Topography of nerve cell loss from the locus caeruleus in middle aged persons with Down’s syndrome. Journal of the neurological sciences, 83(1), pp.15–24. Available at: http://www.ncbi.nlm.nih.gov/pubmed/2964513 [Accessed October 20, 2010]. Marcyniuk, B., Mann, D.M. & Yates, P.O., 1986a. Loss of nerve cells from locus coeruleus in Alzheimer’s disease is topographically arranged. Neuroscience letters, 64(3), pp.247–52. Available at: http://www.ncbi.nlm.nih.gov/pubmed/3960404 [Accessed October 20, 2010]. Marcyniuk, B., Mann, D.M. & Yates, P.O., 1986b. The topography of cell loss from locus caeruleus in Alzheimer’s disease. Journal of the neurological sciences, 76(2- 3), pp.335–45. Available at: http://www.ncbi.nlm.nih.gov/pubmed/3794754 [Accessed October 20, 2010]. Martin, B.K. et al., 2008. Cognitive function over time in the Alzheimer’s Disease Anti- inflammatory Prevention Trial (ADAPT): results of a randomized, controlled trial of naproxen and celecoxib. Archives of neurology, 65(7), pp.896–905. Available at: http://www.ncbi.nlm.nih.gov/pubmed/18474729. Martín, E.D. & Buño, W., 2003. Caffeine-mediated presynaptic long-term potentiation in hippocampal CA1 pyramidal neurons. Journal of neurophysiology, 89(6), pp.3029– 38. Available at: http://www.ncbi.nlm.nih.gov/pubmed/12783948 [Accessed October 20, 2010]. Mascarucci, P. et al., 1998. Glutamate release in the nucleus tractus solitarius induced by peripheral lipopolysaccharide and interleukin-1 beta. Neuroscience, 86(4), pp.1285– 90. Available at: http://www.ncbi.nlm.nih.gov/pubmed/9697133 [Accessed December 18, 2012]. Masliah, E. et al., 1996. Deficient glutamate transport is associated with neurodegeneration in Alzheimer’s disease. Annals of neurology, 40(5), pp.759–66. Available at: http://www.ncbi.nlm.nih.gov/pubmed/8957017 [Accessed January 13, 2012]. Massie, A. et al., 2010. Time-dependent changes in GLT-1 functioning in striatum of hemi-Parkinson rats. Neurochemistry international. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20643175 [Accessed July 28, 2010]. Mastroeni, D. et al., 2009. Microglial responses to dopamine in a cell culture model of Parkinson’s disease. Neurobiology of aging, 30(11), pp.1805–17. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2762863&tool=pmcentr ez&rendertype=abstract [Accessed August 24, 2010]. Matos, M. et al., 2008. Amyloid-beta peptide decreases glutamate uptake in cultured astrocytes: involvement of oxidative stress and mitogen-activated protein kinase cascades. Neuroscience, 156(4), pp.898–910. Available at: http://www.ncbi.nlm.nih.gov/pubmed/18790019. Matsuya, T. et al., 2007. Synergistic effects of adenosine A2A antagonist and L-DOPA on rotational behaviors in 6-hydroxydopamine-induced hemi-Parkinsonian mouse model. Journal of pharmacological sciences, 103(3), pp.329–32. Available at: http://www.ncbi.nlm.nih.gov/pubmed/17341841 [Accessed September 17, 2010]. Matyszak, M.K., 1998. Inflammation in the CNS: balance between immunological privilege and immune responses. Progress in neurobiology, 56(1), pp.19–35.

232

Available at: http://www.ncbi.nlm.nih.gov/pubmed/9723129 [Accessed December 8, 2012]. McGeer, P L et al., 1988. Reactive microglia are positive for HLA-DR in the substantia nigra of Parkinson’s and Alzheimer's disease brains. Neurology, 38(8), pp.1285–91. Available at: http://www.ncbi.nlm.nih.gov/pubmed/3399080 [Accessed October 20, 2010]. McGeer, P L & McGeer, E G, 1998. Glial cell reactions in neurodegenerative diseases: pathophysiology and therapeutic interventions. Alzheimer disease and associated disorders, 12 Suppl 2, pp.S1–6. Available at: http://www.ncbi.nlm.nih.gov/pubmed/9769023 [Accessed February 1, 2012]. McGeer, P L & McGeer, E G, 1999. Inflammation of the brain in Alzheimer’s disease: implications for therapy. Journal of leukocyte biology, 65(4), pp.409–15. Available at: http://www.ncbi.nlm.nih.gov/pubmed/10204568 [Accessed October 20, 2010]. McGeer, Patrick L & McGeer, Edith G, 2002. Local neuroinflammation and the progression of Alzheimer’s disease. Journal of neurovirology, 8(6), pp.529–38. Available at: http://www.ncbi.nlm.nih.gov/pubmed/12476347 [Accessed October 20, 2010]. Melani, A. et al., 2006. The selective A2A receptor antagonist SCH 58261 protects from neurological deficit, brain damage and activation of p38 MAPK in rat focal cerebral ischemia. Brain research, 1073-1074, pp.470–80. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16443200 [Accessed September 24, 2010]. Melief, J. et al., 2012. Phenotyping primary human microglia: Tight regulation of LPS responsiveness. Glia, 000, pp.1–12. Available at: http://www.ncbi.nlm.nih.gov/pubmed/22740309 [Accessed July 18, 2012]. De Mendonça, A & Ribeiro, J A, 1993. Adenosine inhibits the NMDA receptor-mediated excitatory postsynaptic potential in the hippocampus. Brain research, 606(2), pp.351–6. Available at: http://www.ncbi.nlm.nih.gov/pubmed/8098255 [Accessed December 10, 2012]. De Mendonça, A & Ribeiro, J A, 2000. Long-term potentiation observed upon blockade of adenosine A1 receptors in rat hippocampus is N-methyl-D-aspartate receptor- dependent. Neuroscience letters, 291(2), pp.81–4. Available at: http://www.ncbi.nlm.nih.gov/pubmed/10978579 [Accessed October 20, 2010]. Mennerick, S., Benz, A. & Zorumski, C F, 1996. Components of glial responses to exogenous and synaptic glutamate in rat hippocampal microcultures. The Journal of neuroscience : the official journal of the Society for Neuroscience, 16(1), pp.55–64. Available at: http://www.ncbi.nlm.nih.gov/pubmed/8613809 [Accessed December 9, 2012]. Michaelis, E K et al., 2011. Neuronal Glud1 (Glutamate Dehydrogenase 1) Over- Expressing Mice: Increased Glutamate Formation and Synaptic Release, Loss of Synaptic Activity, and Adaptive Changes in Genomic Expression. Neurochemistry international, 1. Available at: http://www.ncbi.nlm.nih.gov/pubmed/21397652 [Accessed March 16, 2011]. Mihara, T., Noda, A., et al., 2008. Brain adenosine A2A receptor occupancy by a novel A1/A2A receptor antagonist, ASP5854, in rhesus monkeys: relationship to anticataleptic effect. Journal of nuclear medicine : official publication, Society of

233

Nuclear Medicine, 49(7), pp.1183–8. Available at: http://www.ncbi.nlm.nih.gov/pubmed/18552135 [Accessed December 10, 2012]. Mihara, T., Iwashita, A. & Matsuoka, N., 2008. A novel adenosine A(1) and A(2A) receptor antagonist ASP5854 ameliorates motor impairment in MPTP-treated marmosets: comparison with existing anti-Parkinson’s disease drugs. Behavioural brain research, 194(2), pp.152–61. Available at: http://www.ncbi.nlm.nih.gov/pubmed/18657577 [Accessed September 24, 2010]. Miller, K.R. & Streit, Wolfgang J, 2007. The effects of aging, injury and disease on microglial function: a case for cellular senescence. Neuron glia biology, 3(3), pp.245–53. Available at: http://www.ncbi.nlm.nih.gov/pubmed/18634615 [Accessed July 25, 2012]. Milton, I.D. et al., 1997. Expression of the glial glutamate transporter EAAT2 in the human CNS: an immunohistochemical study. Brain research. Molecular brain research, 52(1), pp.17–31. Available at: http://www.ncbi.nlm.nih.gov/pubmed/9450673. Minkeviciene, R. et al., 2008. Age-related decrease in stimulated glutamate release and vesicular glutamate transporters in APP/PS1 transgenic and wild-type mice. Journal of neurochemistry, 105(3), pp.584–94. Available at: http://www.ncbi.nlm.nih.gov/pubmed/18042177 [Accessed October 18, 2010]. Mishina, M., Ishii, K. & Ishiwata, K., 2007. Adenosine receptor--relation to dopaminergic system. Rinshō shinkeigaku = Clinical neurology, 47(11), pp.835–7. Available at: http://www.ncbi.nlm.nih.gov/pubmed/18210811 [Accessed October 20, 2010]. Mittelbronn, M. et al., 2001. Local distribution of microglia in the normal adult human central nervous system differs by up to one order of magnitude. Acta neuropathologica, 101(3), pp.249–55. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11307625 [Accessed June 28, 2012]. Miyamoto, M. et al., 1989. Antioxidants protect against glutamate-induced cytotoxicity in a neuronal cell line. The Journal of pharmacology and experimental therapeutics, 250(3), pp.1132–40. Available at: http://www.ncbi.nlm.nih.gov/pubmed/2778712 [Accessed December 12, 2012]. Montori, S. et al., 2010. Age-dependent modifications in the mRNA levels of the rat excitatory amino acid transporters (EAATs) at 48 hour reperfusion following global ischemia. Brain research. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20709031 [Accessed August 24, 2010]. Mookherjee, P. et al., 2011. GLT-1 Loss Accelerates Cognitive Deficit Onset in an Alzheimer’s Disease Animal Model. Journal of Alzheimer’s disease : JAD. Available at: http://www.ncbi.nlm.nih.gov/pubmed/21677376 [Accessed June 20, 2011]. Moran, M. et al., 2003. Cystine/Glutamate antiporter regulation of vesicular glutamate release. , 0, pp.8–10. Mosser, D.M. & Edwards, J.P., 2008. Exploring the full spectrum of macrophage activation. Nature reviews. Immunology, 8(12), pp.958–69. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2724991&tool=pmcentr ez&rendertype=abstract [Accessed March 9, 2012].

234

Mu, X., Azbill, R.D. & Springer, J.E., 2000a. Riluzole and methylprednisolone combined treatment improves functional recovery in traumatic spinal cord injury. Journal of neurotrauma, 17(9), pp.773–80. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11011817 [Accessed January 17, 2012]. Mu, X., Azbill, R.D. & Springer, J.E., 2000b. Riluzole improves measures of oxidative stress following traumatic spinal cord injury. Brain research, 870(1-2), pp.66–72. Available at: http://www.ncbi.nlm.nih.gov/pubmed/10869502 [Accessed January 8, 2012]. Murphy, T.H. et al., 1989. Glutamate toxicity in a neuronal cell line involves inhibition of cystine transport leading to oxidative stress. Neuron, 2(6), pp.1547–58. Available at: http://www.ncbi.nlm.nih.gov/pubmed/2576375 [Accessed December 12, 2012]. Nedergaard, M, 1994. Direct signaling from astrocytes to neurons in cultures of mammalian brain cells. Science, 263(5154), pp.1768–1771. Available at: http://apps.webofknowledge.com/full_record.do?product=WOS&search_mode=Gen eralSearch&qid=39&SID=3BNBmgOLelG2dpbHFi2&page=1&doc=1 [Accessed December 9, 2012]. Nedergaard, Maiken, Takano, T. & Hansen, A.J., 2002. Beyond the role of glutamate as a neurotransmitter. Nature reviews. Neuroscience, 3(9), pp.748–55. Available at: http://www.ncbi.nlm.nih.gov/pubmed/12209123. Nimmerjahn, A., Kirchhoff, F. & Helmchen, F., 2005. Resting microglial cells are highly dynamic surveillants of brain parenchyma in vivo. Science (New York, N.Y.), 308(5726), pp.1314–8. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15831717 [Accessed November 1, 2012]. Niswender, C.M. & Conn, P.J., 2010. Metabotropic glutamate receptors: physiology, pharmacology, and disease. Annual review of pharmacology and toxicology, 50, pp.295–322. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2904507&tool=pmcentr ez&rendertype=abstract [Accessed November 2, 2012]. Nobre, H.V. et al., 2010. Caffeine and CSC, adenosine A2A antagonists, offer neuroprotection against 6-OHDA-induced neurotoxicity in rat mesencephalic cells. Neurochemistry international, 56(1), pp.51–8. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19782116 [Accessed August 24, 2010]. Ochiishi, T. et al., 1999. High level of adenosine A1 receptor-like immunoreactivity in the CA2/CA3a region of the adult rat hippocampus. Neuroscience, 93(3), pp.955– 67. Available at: http://www.ncbi.nlm.nih.gov/pubmed/10473260 [Accessed October 20, 2010]. Okello, A. et al., 2009. Microglial activation and amyloid deposition in mild cognitive impairment: a PET study. Neurology, 72(1), pp.56–62. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2817573&tool=pmcentr ez&rendertype=abstract. Olivares, D. et al., 2011. N-Methyl D-Aspartate (NMDA) Receptor Antagonists and Memantine Treatment for Alzheimer’s Disease, Vascular Dementia and Parkinson's Disease. Current Alzheimer research. Available at: http://www.ncbi.nlm.nih.gov/pubmed/21875407 [Accessed January 3, 2012]. Orr, A.G. et al., 2009. Adenosine A(2A) receptor mediates microglial process retraction. Nature neuroscience, 12(7), pp.872–8. Available at:

235

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2712729&tool=pmcentr ez&rendertype=abstract [Accessed November 21, 2012]. Overmyer, M. et al., 1999. Reactive microglia in aging and dementia: an immunohistochemical study of postmortem human brain tissue. Acta neuropathologica, 97(4), pp.383–92. Available at: http://www.ncbi.nlm.nih.gov/pubmed/10208278 [Accessed December 9, 2012]. O’Kane, E.M. & Stone, T.W., 1998. Interaction between adenosine A1 and A2 receptor- mediated responses in the rat hippocampus in vitro. European journal of pharmacology, 362(1), pp.17–25. Available at: http://www.ncbi.nlm.nih.gov/pubmed/9865525 [Accessed December 10, 2012]. O’Keefe, G., Nguyen, V. & Benveniste, E., Class II transactivator and class II MHC gene expression in microglia: modulation by the cytokines TGF-beta, IL-4, IL-13 and IL- 10. Available at: http://apps.webofknowledge.com.proxy.lib.ohio- state.edu/full_record.do?product=WOS&search_mode=AdvancedSearch&qid=14& SID=3AAOpbeEIk9j5mBeH53&page=1&doc=4&cacheurlFromRightClick=no [Accessed December 9, 2012]. Pabon, M. et al., 2011. CX3CL1 reduces neurotoxicity and microglial activation in a rat model of Parkinson’s disease. Journal of Neuroinflammation, 8(1), p.9. Available at: http://www.jneuroinflammation.com/content/8/1/9 [Accessed January 26, 2011]. Pais, T.F. et al., 2008. Necrotic neurons enhance microglial neurotoxicity through induction of glutaminase by a MyD88-dependent pathway. Journal of neuroinflammation, 5, p.43. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2572162&tool=pmcentr ez&rendertype=abstract [Accessed October 31, 2012]. Palop, J.J., Chin, J. & Mucke, L., 2006. A network dysfunction perspective on neurodegenerative diseases. Nature, 443(7113), pp.768–73. Available at: http://www.ncbi.nlm.nih.gov/pubmed/17051202 [Accessed July 24, 2010]. Palop, J.J. & Mucke, L., 2010. Synaptic depression and aberrant excitatory network activity in Alzheimer’s disease: two faces of the same coin? Neuromolecular medicine, 12(1), pp.48–55. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19838821 [Accessed July 12, 2011]. Parachikova, a et al., 2007. Inflammatory changes parallel the early stages of Alzheimer disease. Neurobiology of aging, 28(12), pp.1821–33. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2198930&tool=pmcentr ez&rendertype=abstract [Accessed August 16, 2010]. Parpura, V et al., 1994. Glutamate-mediated astrocyte-neuron signalling. Nature, 369(6483), pp.744–7. Available at: http://www.ncbi.nlm.nih.gov/pubmed/7911978 [Accessed October 26, 2012]. Parsons, C.G., Stöffler, A. & Danysz, W., 2007a. Memantine: a NMDA receptor antagonist that improves memory by restoration of homeostasis in the glutamatergic system--too little activation is bad, too much is even worse. Neuropharmacology, 53(6), pp.699–723. Available at: http://apps.webofknowledge.com/full_record.do?product=WOS&search_mode=Cita tionReport&qid=29&SID=2FE8IIAddNG@3jAk3CG&page=3&doc=26&cacheurlF romRightClick=no [Accessed October 26, 2012].

236

Parsons, C.G., Stöffler, A. & Danysz, W., 2007b. Memantine: a NMDA receptor antagonist that improves memory by restoration of homeostasis in the glutamatergic system--too little activation is bad, too much is even worse. Neuropharmacology, 53(6), pp.699–723. Available at: http://www.ncbi.nlm.nih.gov/pubmed/17904591 [Accessed August 24, 2010]. Pasti, Lucia et al., 1997. Intracellular Calcium Oscillations in Astrocytes : A Highly Plastic , Astrocytes In Situ. , 17(20), pp.7817–7830. Perrin, R.J., Fagan, A.M. & Holtzman, D.M., 2009. Multimodal techniques for diagnosis and prognosis of Alzheimer’s disease. Nature, 461(7266), pp.916–22. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2810658&tool=pmcentr ez&rendertype=abstract [Accessed July 17, 2010]. Persson, M. et al., 2005. Lipopolysaccharide increases microglial GLT-1 expression and glutamate uptake capacity in vitro by a mechanism dependent on TNF-alpha. Glia, 51(2), pp.111–20. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15789431 [Accessed July 25, 2010]. Phillips, L.M., Simon, P.J. & Lampson, L. a, 1999. Site-specific immune regulation in the brain: differential modulation of major histocompatibility complex (MHC) proteins in brainstem vs. hippocampus. The Journal of comparative neurology, 405(3), pp.322–33. Available at: http://www.ncbi.nlm.nih.gov/pubmed/10076928 [Accessed June 28, 2012]. Piani, D. & Fontana, A., 1994. Involvement of the cystine transport system xc- in the macrophage-induced glutamate-dependent cytotoxicity to neurons. Journal of immunology (Baltimore, Md. : 1950), 152(7), pp.3578–85. Available at: http://www.ncbi.nlm.nih.gov/pubmed/8144936 [Accessed February 3, 2012]. Pierri, M. et al., 2005. KW-6002 protects from MPTP induced dopaminergic toxicity in the mouse. Neuropharmacology, 48(4), pp.517–24. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15755479 [Accessed August 24, 2010]. Popoli, P, Reggio, R. & Pèzzola, A., 2000. Effects of SCH 58261, an adenosine A(2A) receptor antagonist, on quinpirole-induced turning in 6-hydroxydopamine-lesioned rats. Lack of tolerance after chronic caffeine intake. Neuropsychopharmacology : official publication of the American College of Neuropsychopharmacology, 22(5), pp.522–9. Available at: http://www.ncbi.nlm.nih.gov/pubmed/10731627 [Accessed December 10, 2012]. Porter, J.T. & McCarthy, K.D., 1996. Hippocampal astrocytes in situ respond to glutamate released from synaptic terminals. The Journal of neuroscience : the official journal of the Society for Neuroscience, 16(16), pp.5073–81. Available at: http://www.ncbi.nlm.nih.gov/pubmed/8756437 [Accessed December 9, 2012]. Potier, B. et al., 2010. Reduction in glutamate uptake is associated with extrasynaptic NMDA and metabotropic glutamate receptor activation at the hippocampal CA1 synapse of aged rats. Aging cell, 33(May), pp.722–735. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20569241 [Accessed August 24, 2010]. Potter, P.E., 2010. Investigational medications for treatment of patients with Alzheimer disease. The Journal of the American Osteopathic Association, 110(9 Suppl 8), pp.S27–36. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20926740 [Accessed October 15, 2010].

237

Pow, D. V & Cook, D.G., 2009. Neuronal expression of splice variants of “glial” glutamate transporters in brains afflicted by Alzheimer’s disease: unmasking an intrinsic neuronal property. Neurochemical research, 34(10), pp.1748–57. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19319679 [Accessed October 18, 2010]. Powers, K.M. et al., 2008. Combined effects of smoking, coffee, and NSAIDs on Parkinson’s disease risk. Movement disorders : official journal of the Movement Disorder Society, 23(1), pp.88–95. Available at: http://www.ncbi.nlm.nih.gov/pubmed/17987647 [Accessed August 24, 2010]. Prow, N.A. & Irani, D.N., 2008. The inflammatory cytokine, interleukin-1 beta, mediates loss of astroglial glutamate transport and drives excitotoxic motor neuron injury in the spinal cord during acute viral encephalomyelitis. Journal of neurochemistry, 105(4), pp.1276–86. Available at: http://apps.isiknowledge.com/full_record.do?product=WOS&search_mode=General Search&qid=13&SID=3C3NnE5do9Nk77LCkPa&page=7&doc=65&cacheurlFrom RightClick=no [Accessed November 16, 2010]. Quarta, D. et al., 2004. Opposite modulatory roles for adenosine A1 and A2A receptors on glutamate and dopamine release in the shell of the nucleus accumbens. Effects of chronic caffeine exposure. Journal of Neurochemistry, 88(5), pp.1151–1158. Available at: http://doi.wiley.com/10.1046/j.1471-4159.2003.02245.x [Accessed August 24, 2010]. Ransohoff, R.M. & Perry, V.H., 2009. Microglial physiology: unique stimuli, specialized responses. Annual review of immunology, 27, pp.119–45. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19302036 [Accessed July 12, 2010]. Rebola, N et al., 2005. Different synaptic and subsynaptic localization of adenosine A2A receptors in the hippocampus and striatum of the rat. Neuroscience, 132(4), pp.893– 903. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15857695 [Accessed September 1, 2010]. Rebola, Nelson et al., 2011. Adenosine a(2a) Receptors Control Neuroinflammation and Consequent Hippocampal Neuronal Dysfunction. Journal of neurochemistry, pp.100–111. Available at: http://www.ncbi.nlm.nih.gov/pubmed/21235574 [Accessed January 20, 2011]. Rebola, Nelson et al., 2008. Adenosine A2A receptors are essential for long-term potentiation of NMDA-EPSCs at hippocampal mossy fiber synapses. Neuron, 57(1), pp.121–34. Available at: http://www.ncbi.nlm.nih.gov/pubmed/18184569. Rebola, Nelson, Coelho, J.E., et al., 2003. Decrease of adenosine A1 receptor density and of adenosine neuromodulation in the hippocampus of kindled rats. The European journal of neuroscience, 18(4), pp.820–8. Available at: http://www.ncbi.nlm.nih.gov/pubmed/12925008 [Accessed December 10, 2012]. Rebola, Nelson, Sebastião, Ana M, et al., 2003. Enhanced adenosine A2A receptor facilitation of synaptic transmission in the hippocampus of aged rats. Journal of neurophysiology, 90(2), pp.1295–303. Available at: http://www.ncbi.nlm.nih.gov/pubmed/12904509 [Accessed September 1, 2010]. Rodgers, K.M. et al., 2009. The cortical innate immune response increases local neuronal excitability leading to seizures. Brain : a journal of neurology, 132(Pt 9), pp.2478– 86. Available at:

238

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2732268&tool=pmcentr ez&rendertype=abstract [Accessed July 25, 2012]. Rodriguez-Kern, a, 2003. Beta-amyloid and brain-derived neurotrophic factor, BDNF, up-regulate the expression of glutamate transporter GLT-1/EAAT2 via different signaling pathways utilizing transcription factor NF-κB. Neurochemistry International, 43(4-5), pp.363–370. Available at: http://linkinghub.elsevier.com/retrieve/pii/S0197018603000238 [Accessed September 18, 2010]. Rogawski, M.A. & Wenk, Gary L, 2003. The neuropharmacological basis for the use of memantine in the treatment of Alzheimer’s disease. CNS drug reviews, 9(3), pp.275–308. Available at: http://www.ncbi.nlm.nih.gov/pubmed/14530799 [Accessed February 4, 2011]. Rogers, 2002. Elucidating molecular mechanisms of Alzheimer’s disease in microglial cultures. Rojo, A.I. et al., 2010. Nrf2 regulates microglial dynamics and neuroinflammation in experimental Parkinson’s disease. Glia, 58(5), pp.588–98. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19908287 [Accessed August 24, 2010]. Rose, S., Ramsay Croft, N. & Jenner, Peter, 2007. The novel adenosine A2a antagonist ST1535 potentiates the effects of a threshold dose of l-dopa in unilaterally 6-OHDA- lesioned rats. Brain research, 1133(1), pp.110–4. Available at: http://www.ncbi.nlm.nih.gov/pubmed/17196564 [Accessed August 24, 2010]. Rosenzweig, E.S. & Barnes, Carol A, 2003. Impact of aging on hippocampal function: plasticity, network dynamics, and cognition. Progress in Neurobiology, 69(3), pp.143–179. Available at: http://apps.webofknowledge.com/full_record.do?product=WOS&search_mode=Cita tionReport&qid=18&SID=2FE8IIAddNG@3jAk3CG&page=3&doc=21&cacheurlF romRightClick=no [Accessed November 16, 2012]. Rosi, S et al., 2009. Accuracy of hippocampal network activity is disrupted by neuroinflammation: rescue by memantine. Brain : a journal of neurology, 132(Pt 9), pp.2464–77. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19531533 [Accessed September 24, 2010]. Rosi, S et al., 2006. Memantine protects against LPS-induced neuroinflammation, restores behaviorally-induced gene expression and spatial learning in the rat. Neuroscience, 142(4), pp.1303–15. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16989956 [Accessed August 24, 2010]. Rosi, S. et al., 2005. Neuroinflammation alters the hippocampal pattern of behaviorally induced Arc expression. The Journal of neuroscience : the official journal of the Society for Neuroscience, 25(3), pp.723–31. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15659610 [Accessed August 24, 2010]. Rosi, Susanna et al., 2004. Chronic brain inflammation leads to a decline in hippocampal NMDA-R1 receptors. Journal of neuroinflammation, 1(1), p.12. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=500869&tool=pmcentre z&rendertype=abstract [Accessed August 24, 2010]. Rosi, Susanna et al., 2003. The influence of brain inflammation upon neuronal adenosine A 2B receptors. Society, pp.1–8.

239

Ross, G.W. et al., 2000. Association of coffee and caffeine intake with the risk of Parkinson disease. JAMA : the journal of the American Medical Association, 283(20), pp.2674–9. Available at: http://www.ncbi.nlm.nih.gov/pubmed/10819950 [Accessed September 24, 2010]. Rossi, Daniela et al., 2005. Defective tumor necrosis factor-alpha-dependent control of astrocyte glutamate release in a transgenic mouse model of Alzheimer disease. The Journal of biological chemistry, 280(51), pp.42088–96. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16253995 [Accessed October 17, 2011]. Rothstein, J.D. et al., 1994. Localization of neuronal and glial glutamate transporters. Neuron, 13(3), pp.713–725. Available at: http://linkinghub.elsevier.com/retrieve/pii/0896627394900388. Rupsingh, R. et al., 2009. Reduced hippocampal glutamate in Alzheimer disease. Neurobiology of aging. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19501936 [Accessed July 21, 2010]. Sama, M.A. et al., 2008. Interleukin-1beta-dependent signaling between astrocytes and neurons depends critically on astrocytic calcineurin/NFAT activity. The Journal of biological chemistry, 283(32), pp.21953–64. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2494911&tool=pmcentr ez&rendertype=abstract [Accessed November 29, 2012]. Sanchez, P.E. et al., 2012. Levetiracetam suppresses neuronal network dysfunction and reverses synaptic and cognitive deficits in an Alzheimer’s disease model. Proceedings of the National Academy of Sciences of the United States of America, 109(42), pp.E2895–903. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3479491&tool=pmcentr ez&rendertype=abstract [Accessed October 26, 2012]. Santello, M., Bezzi, Paola & Volterra, Andrea, 2011. TNFα controls glutamatergic gliotransmission in the hippocampal dentate gyrus. Neuron, 69(5), pp.988–1001. Available at: http://www.ncbi.nlm.nih.gov/pubmed/21382557 [Accessed March 19, 2012]. Sasaki, K. et al., 2009. Excitatory amino acid transporter 2 associates with phosphorylated tau and is localized in neurofibrillary tangles of tauopathic brains. FEBS letters, 583(13), pp.2194–200. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19527721 [Accessed January 7, 2011]. Saura, C.A. et al., 1998. Adenosine deaminase and A1 adenosine receptors internalize together following agonist-induced receptor desensitization. The Journal of biological chemistry, 273(28), pp.17610–7. Available at: http://www.ncbi.nlm.nih.gov/pubmed/9651356 [Accessed December 10, 2012]. Saura, J. et al., 2005. Adenosine A2A receptor stimulation potentiates nitric oxide release by activated microglia. Journal of neurochemistry, 95(4), pp.919–29. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16092928 [Accessed September 24, 2010]. Scott, H. a et al., 2011. Glutamate transporter variants reduce glutamate uptake in Alzheimer’s disease. Neurobiology of aging, 32(3), pp.553.e1–11. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20416976 [Accessed December 1, 2012]. Selkirk, J. V, Stiefel, T.H., et al., 2005. Over-expression of the human EAAT2 glutamate transporter within neurons of mouse organotypic hippocampal slice cultures leads to increased vulnerability of CA1 pyramidal cells. The European journal of

240

neuroscience, 21(8), pp.2291–6. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15869527 [Accessed January 7, 2011]. Selkirk, J. V, Nottebaum, L.M., et al., 2005. Role of the GLT-1 subtype of glutamate transporter in glutamate homeostasis: the GLT-1-preferring inhibitor WAY-855 produces marginal neurotoxicity in the rat hippocampus. The European journal of neuroscience, 21(12), pp.3217–28. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16026460 [Accessed January 7, 2011]. Sheffield, L.G. & Berman, N.E., 1998. Microglial expression of MHC class II increases in normal aging of nonhuman primates. Neurobiology of aging, 19(1), pp.47–55. Available at: http://www.ncbi.nlm.nih.gov/pubmed/9562503 [Accessed June 28, 2012]. Sheffield, L.G., Marquis, J.G. & Berman, N.E., 2000. Regional distribution of cortical microglia parallels that of neurofibrillary tangles in Alzheimer’s disease. Neuroscience letters, 285(3), pp.165–8. Available at: http://www.ncbi.nlm.nih.gov/pubmed/10806312 [Accessed December 18, 2012]. Sheldon, A.L. & Robinson, M.B., 2007. The role of glutamate transporters in neurodegenerative diseases and potential opportunities for intervention. Neurochemistry international, 51(6-7), pp.333–55. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2075474&tool=pmcentr ez&rendertype=abstract [Accessed December 18, 2010]. Shen, Y et al., 2001. Complement activation by neurofibrillary tangles in Alzheimer’s disease. Neuroscience letters, 305(3), pp.165–8. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11403931 [Accessed December 18, 2012]. Sheng, J.G., Griffin, W.S., et al., 1998. Distribution of interleukin-1-immunoreactive microglia in cerebral cortical layers: implications for neuritic plaque formation in Alzheimer’s disease. Neuropathology and applied neurobiology, 24(4), pp.278–83. Available at: http://www.ncbi.nlm.nih.gov/pubmed/9775393 [Accessed December 18, 2012]. Sheng, J.G., Mrak, R E & Griffin, W.S., 1998. Enlarged and phagocytic, but not primed, interleukin-1 alpha-immunoreactive microglia increase with age in normal human brain. Acta neuropathologica, 95(3), pp.229–34. Available at: http://www.ncbi.nlm.nih.gov/pubmed/9542587 [Accessed December 18, 2012]. Sheng, J.G., Mrak, R E & Griffin, W.S., 1997. Neuritic plaque evolution in Alzheimer’s disease is accompanied by transition of activated microglia from primed to enlarged to phagocytic forms. Acta neuropathologica, 94(1), pp.1–5. Available at: http://www.ncbi.nlm.nih.gov/pubmed/9224523 [Accessed December 19, 2012]. Shijie, J. et al., 2009. Blockade of glutamate release from microglia attenuates experimental autoimmune encephalomyelitis in mice. The Tohoku journal of experimental medicine, 217(2), pp.87–92. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19212100. Shiozaki, S. et al., 1999. Actions of adenosine A2A receptor antagonist KW-6002 on drug-induced catalepsy and hypokinesia caused by reserpine or MPTP. Psychopharmacology, 147(1), pp.90–5. Available at: http://www.ncbi.nlm.nih.gov/pubmed/10591873 [Accessed September 24, 2010]. Simard, A.R. et al., 2006. Bone marrow-derived microglia play a critical role in restricting senile plaque formation in Alzheimer’s disease. Neuron, 49(4), pp.489–

241

502. Available at: http://apps.webofknowledge.com.proxy.lib.ohio- state.edu/full_record.do?product=WOS&search_mode=CitationReport&qid=11&SI D=1A263P6PcLNnKcD6kL@&page=1&doc=1&cacheurlFromRightClick=no [Accessed November 6, 2012]. Simic, G. et al., 2009. Does Alzheimer’s disease begin in the brainstem? Neuropathology and applied neurobiology, 35(6), pp.532–54. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19682326. Simões, A.P. et al., 2012. Blockade of adenosine A(2A) receptors prevents interleukin- 1β-induced exacerbation of neuronal toxicity through a p38 mitogen-activated protein kinase pathway. Journal of neuroinflammation, 9, p.204. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3439355&tool=pmcentr ez&rendertype=abstract [Accessed November 12, 2012]. Snell, B.J. et al., 2000. Visualisation of AMPA binding sites in the brain of mice lacking the adenosine A 2a receptor. Neuroscience Letters, 291, pp.97–100. Sondag, C.M., Dhawan, G. & Combs, C.K., 2009. Beta amyloid oligomers and fibrils stimulate differential activation of primary microglia. Journal of neuroinflammation, 6, p.1. Available at: http://apps.isiknowledge.com.proxy.lib.ohio- state.edu/full_record.do?product=WOS&search_mode=GeneralSearch&qid=31&SI D=3D@IDameIEl@cCF3gIJ&page=2&doc=11&cacheurlFromRightClick=no [Accessed January 23, 2011]. Stanika, R.I. et al., 2009. Coupling diverse routes of calcium entry to mitochondrial dysfunction and glutamate excitotoxicity. Proceedings of the National Academy of Sciences of the United States of America, 106(24), pp.9854–9. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2701040&tool=pmcentr ez&rendertype=abstract [Accessed December 10, 2012]. Stein, J.L. et al., 2010. Genome-wide analysis reveals novel genes influencing temporal lobe structure with relevance to neurodegeneration in Alzheimer’s disease. NeuroImage, 51(2), pp.542–54. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2856746&tool=pmcentr ez&rendertype=abstract [Accessed November 19, 2012]. Stephens, M.L. et al., 2009a. Age-related changes in glutamate release in the CA3 and dentate gyrus of the rat hippocampus. Neurobiology of aging. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19535175 [Accessed August 24, 2010]. Stephens, M.L. et al., 2009b. Age-related changes in glutamate release in the CA3 and dentate gyrus of the rat hippocampus. Neurobiology of aging. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19535175 [Accessed August 24, 2010]. Streit, W J & Graeber, M B, 1996. Microglia: a pictorial. Progress in histochemistry and cytochemistry, 31(1), pp.1–89. Available at: http://www.ncbi.nlm.nih.gov/pubmed/8893306 [Accessed December 17, 2012]. Streit, Wolfgang J et al., 2009. Dystrophic (senescent) rather than activated microglial cells are associated with tau pathology and likely precede neurodegeneration in Alzheimer’s disease. Acta neuropathologica, 118(4), pp.475–85. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2737117&tool=pmcentr ez&rendertype=abstract. Streit, Wolfgang J et al., 2004. Dystrophic microglia in the aging human brain. Glia, 45(2), pp.208–12. Available at: http://apps.isiknowledge.com.proxy.lib.ohio-

242

state.edu/full_record.do?product=WOS&search_mode=GeneralSearch&qid=3&SID =3D@IDameIEl@cCF3gIJ&page=1&doc=7&cacheurlFromRightClick=no [Accessed February 12, 2011]. Streit, Wolfgang J, 2006. Microglial senescence: does the brain’s immune system have an expiration date? Trends in neurosciences, 29(9), pp.506–10. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16859761 [Accessed August 3, 2010]. Streit, Wolfgang J & Xue, Q., 2010. The Brain ’ s Aging Immune System. , 1(3), pp.254– 261. Strohmeyer, Ron & Rogers, Joseph, 2001. Molecular and cellular mediators of Alzheimer’s disease inflammation. Journal of Alzheimer’s disease : JAD, 3(1), pp.131–157. Available at: http://www.ncbi.nlm.nih.gov/pubmed/12214082 [Accessed December 19, 2012]. Su, X. et al., 2008. Synuclein activates microglia in a model of Parkinson’s disease. Neurobiology of aging, 29(11), pp.1690–701. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2621109&tool=pmcentr ez&rendertype=abstract [Accessed August 2, 2010]. Svenningsson, P et al., 1999. Distribution, biochemistry and function of striatal adenosine A2A receptors. Progress in neurobiology, 59(4), pp.355–96. Available at: http://www.ncbi.nlm.nih.gov/pubmed/10501634 [Accessed October 20, 2010]. Szczepanik, A., 2001. IL-4, IL-10 and IL-13 modulate Aβ(1–42)-induced cytokine and chemokine production in primary murine microglia and a human monocyte cell line. Journal of Neuroimmunology, 113(1), pp.49–62. Available at: http://apps.webofknowledge.com.proxy.lib.ohio- state.edu/full_record.do?product=WOS&search_mode=AdvancedSearch&qid=14& SID=3AAOpbeEIk9j5mBeH53&page=1&doc=2&cacheurlFromRightClick=no [Accessed December 9, 2012]. Szpak, G M et al., 2001. Neurones and microglia in central nervous system immune response to degenerative processes. Part 1: Alzheimer’s disease and Lewy body variant of Alzheimer's disease. Quantitative study. Folia neuropathologica / Association of Polish Neuropathologists and Medical Research Centre, Polish Academy of Sciences, 39(3), pp.181–92. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11770129 [Accessed December 18, 2012]. Takeuchi, H. et al., 2008. Blockade of microglial glutamate release protects against ischemic brain injury. Experimental neurology, 214(1), pp.144–6. Available at: http://www.ncbi.nlm.nih.gov/pubmed/18775425 [Accessed November 29, 2012]. Takeuchi, H. et al., 2006a. Tumor necrosis factor-alpha induces neurotoxicity via glutamate release from hemichannels of activated microglia in an autocrine manner. The Journal of biological chemistry, 281(30), pp.21362–8. Available at: http://apps.webofknowledge.com/full_record.do?product=WOS&search_mode=Cita tionReport&qid=16&SID=3BNBmgOLelG2dpbHFi2&page=1&doc=9&cacheurlFr omRightClick=no [Accessed November 29, 2012]. Takeuchi, H. et al., 2006b. Tumor necrosis factor-alpha induces neurotoxicity via glutamate release from hemichannels of activated microglia in an autocrine manner. The Journal of biological chemistry, 281(30), pp.21362–8. Available at: http://apps.isiknowledge.com/full_record.do?product=WOS&search_mode=Refine

243

&qid=19&SID=1C6g@M88o86nDnFg@f1&page=1&doc=4&cacheurlFromRightC lick=no [Accessed July 24, 2010]. Tan, S., Schubert, D. & Maher, P, 2001. Oxytosis: A novel form of programmed cell death. Current topics in medicinal chemistry, 1(6), pp.497–506. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11895126 [Accessed December 11, 2012]. Taylor, D.L. et al., 2005. Stimulation of microglial metabotropic glutamate receptor mGlu2 triggers tumor necrosis factor alpha-induced neurotoxicity in concert with microglial-derived Fas ligand. The Journal of neuroscience : the official journal of the Society for Neuroscience, 25(11), pp.2952–64. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15772355 [Accessed October 31, 2012]. Thai, D.R., 2002. Excitatory amino acid transporter EAAT-2 in tangle-bearing neurons in Alzheimer’s disease. Brain pathology (Zurich, Switzerland), 12(4), pp.405–11. Available at: http://www.ncbi.nlm.nih.gov/pubmed/12408226 [Accessed January 7, 2011]. Tobinick, E., 2009. Tumour necrosis factor modulation for treatment of Alzheimer’s disease: rationale and current evidence. CNS drugs, 23(9), pp.713–25. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19689163 [Accessed December 15, 2012]. Tooyama, I. et al., 1990. Reactive microglia express class I and class II major histocompatibility complex antigens in Alzheimer’s disease. Brain research, 523(2), pp.273–80. Available at: http://www.ncbi.nlm.nih.gov/pubmed/2400911 [Accessed October 20, 2010]. Town, T., Nikolic, V. & Tan, Jun, 2005. The microglial “activation” continuum: from innate to adaptive responses. Journal of neuroinflammation, 2, p.24. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=1298325&tool=pmcentr ez&rendertype=abstract [Accessed August 15, 2011]. Trevitt, J. et al., 2009. Adenosine antagonists reverse the cataleptic effects of haloperidol: implications for the treatment of Parkinson’s disease. Pharmacology, biochemistry, and behavior, 92(3), pp.521–7. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19463269 [Accessed September 28, 2010]. Trotti, D., Danbolt, N C & Volterra, a, 1998. Glutamate transporters are oxidant- vulnerable: a molecular link between oxidative and excitotoxic neurodegeneration? Trends in pharmacological sciences, 19(8), pp.328–34. Available at: http://www.ncbi.nlm.nih.gov/pubmed/9745361. Tröster, A.I., 2008. Neuropsychological characteristics of dementia with Lewy bodies and Parkinson’s disease with dementia: differentiation, early detection, and implications for “mild cognitive impairment” and biomarkers. Neuropsychology review, 18(1), pp.103–19. Available at: http://www.ncbi.nlm.nih.gov/pubmed/18322801. Tsutsui, S. et al., 2004. A1 adenosine receptor upregulation and activation attenuates neuroinflammation and demyelination in a model of multiple sclerosis. The Journal of neuroscience : the official journal of the Society for Neuroscience, 24(6), pp.1521–9. Available at: http://www.ncbi.nlm.nih.gov/pubmed/14960625 [Accessed August 24, 2010]. Tweedie, D. et al., 2012. Tumor necrosis factor-α synthesis inhibitor 3,6’- dithiothalidomide attenuates markers of inflammation, Alzheimer pathology and behavioral deficits in animal models of neuroinflammation and Alzheimer's disease.

244

Journal of neuroinflammation, 9, p.106. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=3405480&tool=pmcentr ez&rendertype=abstract [Accessed December 15, 2012]. Vanguilder, H.D. et al., 2011. Concurrent hippocampal induction of MHC II pathway components and glial activation with advanced aging is not correlated with cognitive impairment. Journal of neuroinflammation, 8(1), p.138. Available at: http://www.ncbi.nlm.nih.gov/pubmed/21989322 [Accessed October 24, 2011]. Varnum, M.M. & Ikezu, T., 2012. The classification of microglial activation phenotypes on neurodegeneration and regeneration in Alzheimer’s disease brain. Archivum immunologiae et therapiae experimentalis, 60(4), pp.251–66. Available at: http://www.ncbi.nlm.nih.gov/pubmed/22710659 [Accessed July 25, 2012]. Vazquez, J.F. et al., 2008. Local stimulation of the adenosine A2B receptors induces an increased release of IL-6 in mouse striatum: an in vivo microdialysis study. Journal of neurochemistry, 105(3), pp.904–9. Available at: http://www.ncbi.nlm.nih.gov/pubmed/18088370 [Accessed August 24, 2010]. Vesce, Sabino et al., 2007. Glutamate release from astrocytes in physiological conditions and in neurodegenerative disorders characterized by neuroinflammation. International review of neurobiology, 82, pp.57–71. Available at: http://www.ncbi.nlm.nih.gov/pubmed/17678955 [Accessed June 28, 2012]. Vidwans, A.S. & Hewett, S.J., 2004. Enhanced release of synaptic glutamate underlies the potentiation of oxygen-glucose deprivation-induced neuronal injury after induction of NOS-2. Experimental neurology, 190(1), pp.91–101. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15473983 [Accessed June 15, 2011]. Viviani, B. et al., 2003. Interleukin-1beta enhances NMDA receptor-mediated intracellular calcium increase through activation of the Src family of kinases. The Journal of neuroscience : the official journal of the Society for Neuroscience, 23(25), pp.8692–700. Available at: http://www.ncbi.nlm.nih.gov/pubmed/14507968 [Accessed December 10, 2012]. Volterra, a et al., 1994. Reactive oxygen species inhibit high-affinity glutamate uptake: molecular mechanism and neuropathological implications. Annals of the New York Academy of Sciences, 738, pp.153–62. Available at: http://www.ncbi.nlm.nih.gov/pubmed/7832426 [Accessed June 28, 2012]. Volterra, Andrea & Steinhäuser, C., 2004. Glial modulation of synaptic transmission in the hippocampus. Glia, 47(3), pp.249–57. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15252814 [Accessed March 3, 2012]. Vorwerk, C.K. et al., 2004. Effects of axonal injury on ganglion cell survival and glutamate homeostasis. Brain research bulletin, 62(6), pp.485–90. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15036562 [Accessed January 17, 2012]. Walker, Douglas G et al., 2009. Decreased expression of CD200 and CD200 receptor in Alzheimer’s disease: a potential mechanism leading to chronic inflammation. Experimental neurology, 215(1), pp.5–19. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2765462&tool=pmcentr ez&rendertype=abstract [Accessed August 24, 2010]. Wang, Xinkun et al., 2010. Transcriptomic responses in mouse brain exposed to chronic excess of the neurotransmitter glutamate. BMC genomics, 11, p.360. Available at:

245

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2896956&tool=pmcentr ez&rendertype=abstract. Wang, Xinkun & Michaelis, Elias K, 2010. Selective neuronal vulnerability to oxidative stress in the brain. Frontiers in aging neuroscience, 2, p.12. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2874397&tool=pmcentr ez&rendertype=abstract [Accessed September 16, 2010]. Wang, Y et al., 1999. Effect of interleukin-1 beta on the elevation of cytoplasmic free calcium of the cultured hippocampal neurons induced by L-glutamate. Journal of Tongji Medical University = Tong ji yi ke da xue xue bao, 19(2), pp.120–3. Available at: http://www.ncbi.nlm.nih.gov/pubmed/12840854 [Accessed December 18, 2012]. Wardas, J., Konieczny, J. & Lorenc-Koci, E., 2001. SCH 58261, an A(2A) adenosine receptor antagonist, counteracts parkinsonian-like muscle rigidity in rats. Synapse (New York, N.Y.), 41(2), pp.160–71. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11400182 [Accessed September 24, 2010]. Wei, R. & Jonakait, G.M., 1999. Neurotrophins and the anti-inflammatory agents interleukin-4 (IL-4), IL-10, IL-11 and transforming growth factor-β1 (TGF-β1) down-regulate T cell costimulatory molecules B7 and CD40 on cultured rat microglia. Journal of Neuroimmunology, 95(1-2), pp.8–18. Available at: http://apps.webofknowledge.com.proxy.lib.ohio- state.edu/full_record.do?product=WOS&search_mode=CitationReport&qid=4&SID =3AAOpbeEIk9j5mBeH53&page=1&doc=3&cacheurlFromRightClick=no [Accessed December 9, 2012]. Wenk, Gary L, Parsons, C.G. & Danysz, W., 2006. Potential role of N-methyl-D- aspartate receptors as executors of neurodegeneration resulting from diverse insults: focus on memantine. Behavioural pharmacology, 17(5-6), pp.411–24. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16940762. Wenk, Gary L., Parsons, C.G. & Danysz, W., 2006. Potential role of N-methyl-D- aspartate receptors as executors of neurodegeneration resulting from diverse insults: focus on memantine. BEHAVIOURAL PHARMACOLOGY, 17(5-6), pp.411–424. Available at: http://apps.isiknowledge.com/full_record.do?product=WOS&search_mode=Refine &qid=19&SID=1C6g@M88o86nDnFg@f1&page=4&doc=38&cacheurlFromRight Click=no [Accessed February 10, 2011]. Woltjer, R.L. et al., 2010. Aberrant detergent-insoluble excitatory amino acid transporter 2 accumulates in Alzheimer disease. Journal of neuropathology and experimental neurology, 69(7), pp.667–76. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20535038 [Accessed January 13, 2012]. Wynne, A.M. et al., 2010a. Protracted downregulation of CX(3)CR1 on microglia of aged mice after lipopolysaccharide challenge. Brain, behavior, and immunity, 24(7), pp.1190–201. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2939290&tool=pmcentr ez&rendertype=abstract [Accessed August 26, 2010]. Wynne, A.M. et al., 2010b. Protracted downregulation of CX3CR1 on microglia of aged mice after lipopolysaccharide challenge. Brain, behavior, and immunity, 24(7), pp.1190–201. Available at:

246

http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2939290&tool=pmcentr ez&rendertype=abstract [Accessed July 25, 2012]. Wynne, A.M., Henry, Christopher J & Godbout, Jonathan P, 2009. Immune and behavioral consequences of microglial reactivity in the aged brain. Integrative and comparative biology, 49(3), pp.254–66. Available at: http://www.ncbi.nlm.nih.gov/pubmed/21665818 [Accessed April 18, 2012]. Wyss-Coray, Tony & Mucke, L., 2002. Inflammation in neurodegenerative disease--a double-edged sword. Neuron, 35(3), pp.419–32. Available at: http://www.ncbi.nlm.nih.gov/pubmed/12165466 [Accessed August 10, 2011]. Xia, P. et al., 2010. Memantine Preferentially Blocks Extrasynaptic over Synaptic NMDA Receptor Currents in Hippocampal Autapses. Journal of Neuroscience, 30(33), pp.11246–11250. Available at: http://www.jneurosci.org/cgi/doi/10.1523/JNEUROSCI.2488-10.2010 [Accessed August 19, 2010]. Xiang, Z. et al., 2006. Microglia activation in the brain as inflammatory biomarker of Alzheimer’s disease neuropathology and clinical dementia. Disease markers, 22(1- 2), pp.95–102. Available at: http://www.ncbi.nlm.nih.gov/pubmed/16410654 [Accessed December 18, 2012]. Yadav, S. et al., 2012. Role of secondary mediators in caffeine-mediated neuroprotection in maneb- and paraquat-induced Parkinson’s disease phenotype in the mouse. Neurochemical research, 37(4), pp.875–84. Available at: http://www.ncbi.nlm.nih.gov/pubmed/22201039 [Accessed December 10, 2012]. Ye, Z.C. & Sontheimer, H., 1996. Cytokine modulation of glial glutamate uptake: a possible involvement of nitric oxide. Neuroreport, 7(13), pp.2181–5. Available at: http://www.ncbi.nlm.nih.gov/pubmed/8930985 [Accessed December 18, 2012]. Zerangue, N. & Kavanaugh, M.P., 1996a. Flux coupling in a neuronal glutamate transporter. Nature, 383(6601), pp.634–7. Available at: http://www.ncbi.nlm.nih.gov/pubmed/8857541 [Accessed January 30, 2011]. Zerangue, N. & Kavanaugh, M.P., 1996b. Interaction of L-cysteine with a human excitatory amino acid transporter. The Journal of physiology, 493 ( Pt 2, pp.419–23. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=1158927&tool=pmcentr ez&rendertype=abstract [Accessed January 30, 2011]. Zhan, S.S. et al., 1995. Distribution of beta amyloid associated proteins in plaques in Alzheimer’s disease and in the non-demented elderly. Neurodegeneration : a journal for neurodegenerative disorders, neuroprotection, and neuroregeneration, 4(3), pp.291–7. Available at: http://www.ncbi.nlm.nih.gov/pubmed/8581561 [Accessed November 23, 2012]. Zhang, Wei et al., 2005. Aggregated alpha-synuclein activates microglia: a process leading to disease progression in Parkinson’s disease. The FASEB journal : official publication of the Federation of American Societies for Experimental Biology, 19(6), pp.533–42. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15791003 [Accessed August 16, 2010]. Zhang, Wei et al., 2009. Neuromelanin Activates Microglia and Induces Degeneration of Dopaminergic Neurons: Implications for Progression of Parkinson’s Disease.

247

Neurotoxicity research. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19957214 [Accessed October 20, 2010].

248

Appendix A: List of Abbreviations

6-OHDA 6-hydroxydopamine A1R Adenosine receptor 1 A2AR Adenosine receptor 2A A2BR Adenosine receptor 2B A3R Adenosine receptor 3 AchEI Acetylcholinesterase inhibitor aCSF Artificial cerebral spinal fluid AD Alzheimer’s disease ADAPT Alzheimer’s Disease Anti-inflammatory Prevention Trial ADNI Alzheimer’s Disease Neuroimaging Initiative AMPAR α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors ANOVA Analysis of variance AP-1 Activator protein 1 APC Antigen presenting cell ApoE4 Apolipoprotein E isoform 4 APP Amyloid precursor protein Arc Activity-regulated cytoskeletal-associated protein Arg1 Arginase 1 Arg2 Arginase 2 Aß Amyloid ß ATP Adenosine triphosphate BBB Blood-brain barrier BDNF Brain derived neurotrophic factor BSA Bovine serum albumin CA1 Hippocampal Cornu Ammonis area 1 CA3 Hippocampal Cornu Ammonis area 3 Caf Caffeine cAMP Cyclic adenosine monophosphate CAT2 Cationic amino acid transporter 2 CB2 Cannabinoid receptor 2 CC Corpus collosum CCL- Chemokine (C-C motif) ligand CD- Cluster of differentiation CEBPA CCAAT enhancer binding protein alpha Cef Ceftriaxone (Rocephin) CHA N6-cyclohexyl adenosine 249

Chi3L Chitinase-3-like protein CNS Central nervous system COX Cyclooxygenase CSF Cerebral spinal fluid CSF1R Colony stimulating 1 receptor Ct Threshold cycle CX3CL Fractalkine CX3CR Fractalkine receptor CXCL- Chemokine (C-X-C motif) ligand D1R Dopamine receptor 1 D2R Dopamine receptor 2 D3R Dopamine receptor 3 DAB 3,3-Diaminobenzidine tetrahydrochloride DAMP Damage-associated molecular pattern molecule DAPI Diamidino-2-phenylindole DAT Dopamine active transporter DBH Dopamine-ß-hydroxylase DBH+ DBH immunopositive DC-SIGN Dendritic cell-specific intercellular adhesion molecule-3-grabbing non- integrin DG Dentate gyrus DLB Lewy body dementia DPM Decays per minute EAAT Excitatory amino acid transporter EC Entorhinal cortex ELISA Enzyme-linked immunosorbent assay EPSP Excitatory post-synaptic potential ERK Extracellular signal-regulated kinase FIZZ Found in inflammatory zone; FCεR2 GABA γ-Aminobutyric acid GAPDH Glyceraldehyde 3-phosphate dehydrogenase GDNF Glial cell line-derived neurotrophic factor GFAP Glial fibrillary acidic protein GFAP+ GFAP immuno-reactive GLAST Glutamate asparate transporter Gln Glutamine Gln Syn Glutamine synthase GLT1 Glutamate transporter 1, EAAT2, solute carrier family 1 member 2 Glu Glutamate GLUD1 Glutamate dehydrogenase 1 GM-CSF Granulocyte-macrophage colony-stimulating factor GPe External globus pallidus Gpi Internal globus pallidus

250

GSH Glutathione H2O2 Hydrogen peroxide HBSS Hank’s Balanced Salt Solution HMGB1 High-mobility box group 1 HO-1 Heme-oxygenase 1 HSP Heat shock protein i.c.v Intracerebroventricular i.p. Intraperitoneal ICAM Intercellular adhesion molecule ICAM1 Intercellular adhesion molecule 1 IFIT3 Interferon-induced protein with tetratricopeptide repeats 3 IFNγ Interferon γ IKK IκB kinase IL- Interleukin IL-1Ra Interleukin 1 receptor antagonist IL-4Ra Interleukin 4 receptor alpha iNOS Inducible nitric oxide synthase IRAK Interleukin-1 receptor-associated kinases IRF Interferon regulatory factor ISRE Interferon stimulated response element JNK c-Jun N-terminal kinase LB Lewy body LC Locus coeruleus L-DOPA Levadopa; L-3,4-dihydroxyphenylalanine LPS Lipopolysaccharide LRP1 LDL receptor-related protein 1 LTP Long-term potentiation MΦ Macrophage M1 Classical microglia activation M2a Alternative microglia activation M2c Acquired microglia deactivation Mac-1 Macrophage-1 antigen MAO Monoamine oxidase MAPK Mitogen activated protein kinase MARCO Macrophage receptor M-CSF Macrophage colony stimulating factor MCI Mild cognitive impairment MF Mossy fibers mGluR Metabotropic glutamate receptor MHCI Major histocompatibility complex I MHCII Major histocompatibility complex II MHCII+ MHCII immunopositive MKK Mitogen activated protein kinase 251

MMP- Matrix metalloproteinase MMSE Mini-mental state exam MPO Myeloperoxidase MPTP 1-Methyl-4-phenyl-1,2,3,6-tetrahydropyridine MR Mannose receptor MSR1 Macrophage scavenger receptor 1 MyD88 Myeloid differentiation primary response gene 88 NADPH Nicotinamide adenine dinucleotide phosphate (reduced) NFT Neurofibrillary tangle NFκB nuclear factor-κB NGF Nerve growth factor NMDA N-methyl-D-aspartate NMDAR N-methyl-D-aspartic acid receptor NO Nitric oxide NOS Nitric oxide synthase NSAID Non-steroidal anti-inflammatory P2RX7 Purinergic receptor P2X, ligand-gated ion channel, 7 PAMP Pathogen-associated molecular pattern PBSt PBS with 0.05% tween 20 PD Parkinson’s disease PG Prostaglandin PGE2 Prostaglandin E2 PP Perforant pathway PPARγ Peroxisome proliferator-activated receptor γ PRR Pattern recognition receptor pTH Phosphorylated tyrosine hydroxylase qrtPCR Quantitative reverse-transcription polymerase chain reaction r Pearson correlation coefficient RAGE Receptor for advanced glycation endproducts RANTES Regulated and normal T cell expressed and secreted; Chemokine (C-C motif) ligand 5 RELA transcription factor p65 Ril Riluzole (Rilutek) ROS/RNS Reactive oxygen and nitrogen species s.c. Subcutaneous S1PR1 Sphingosine 1 phosphate receptor 1 Sal or SAL Saline SA-PE Streptavidin-phycoerythrin conjugate Sb Subiculum SC Schaffer collaterals SLC1A2 Solute carrier family 1 member 2; GLT1; EAAT2 SLC7A11 Cysteine/glutamate transporter system xCT SMAD Mothers against decapentaplegic homolog

252

SN Substantia nigra SNAP25 Synaptosomal-associated protein 25 SNARE Sensitive fusion protein attachment protein receptor SNpc Substantia nigra pars compacta SOCS Suppressors of cytokine signaling SOD Superoxide dismutase SphK-1/2 Sphingosine kinase 1/2 SR Scavenger receptor STAT Signal transducers and activators of transcription STN Subthalamic nucleus TAP2 Antigen peptide transporter 2 TBS Tris-buffered saline TBSt TBS with 0.05% tween 20 TCA Citric acid cycle TGFß Transforming growth factor ß TH Tyrosine hydroxylase TH+ TH immunopositive TIMP Tissue inhibitors of metalloproteinases TLR Toll-like receptor TNF Tumor necrosis factor TNFα Tumor necrosis factor α TNFαR TNFα receptor TNFR- TNFα receptor TRADD TNFαR death domain TREM2 Triggering receptor expressed on myeloid cells 2 TrypH Tryptophan hydroxylase TrypH+ TrypH immunopositive VGluT Vesicular glutamate transporter xCT Cystine/glutamate antiporter; SLC7A11 YM1 Chi3L3

253

Appendix B: Microglia activation states and immune factors

This table lists elements and describes how they are related to various microglia activation states, to the best of our current knowledge, according to (Frank, Wieseler- Frank, et al. 2006; Vanguilder et al. 2011; Melief et al. 2012; Town et al. 2005; C. A. Colton 2009; C. A. Colton & Wilcock 2010b; Lucin & Tony Wyss-Coray 2009b; M. a Lynch 2009; Ransohoff & Perry 2009; Cameron & G. E. Landreth 2010; Varnum & Ikezu 2012; Durafourt et al. 2012) and others listed. Details in this table are generalized and not exhaustive, but are meant to be relatively comprehensive and contain factors that are not discussed in the text of this document (see Section 1.2.1). Data are comprised from primary articles as well as reviews, and microglia activation state is occasionally categorized based upon the activating stimulus and/or factors expressed, when the activation state is not declared by the primary authors. Reported changes in gene expression are not differentiated from protein expression. A color-coding scheme is used to designate classical activation (M1, red), alternative activation (M2a, blue), acquired deactivation (M2c, teal) and resting (green), with intermediate colors used to represent factors that are observed in or drive multiple or intermediate states. When available, fold change of gene expression is indicated by arrows (↑ = > 10x, ↑↑ > 100x, ↑↑↑ > 1000x). Abbreviations not included in the table can be found in the preceding Appendix A. This table is intended to accompany Figure 3 and Table 6.

254

Factor Location Function Expression State Ref determinants TLRs TLRs on PRRs activated by PAMPS and DAMPs → Induce M1 M1; Wilcock & Colton, microglia cell transcription factors (NFκB) → pro-inflammatory aging 2010; Darafourt et al., surface cytokines. TLR1,2,6 – lipoproteins, TLR5,11 – 2012; Varnuum & proteins, TLR4 – LPS Ikezu, 2012; VanGuilder et al., 2011 ICAM1 Microglia, MΦ, Intercellular adhesion molecule. Bind to MAC1 Induced by M1; Darafourt et al., 2012; leukocyte on T cells. GM- aging VanGuilder et al., 2011 surface CSF+IFNγ/LP S CD40 Microglia, MΦ, Binds CD40L on CD4+ T cells, interacts with Induced by M1 Wilcock & Colton, astrocyte surface TRAFs. GM- 2010; Darafourt et al., CSF+IFNγ/LP 2012

255 S

CD44 Cell surface Receptor for ligands that include MMPs. Induced by M1 Darafourt et al., 2012 GM- CSF+IFNγ/LP S MARCO Microglia PRR, class A macrophage scavenger receptors, M1 Colton, 2009; Wilcock surface macrophage receptor with collagenous structure. & Colton, 2010 Binds bacteria. MSR1 Microglia PRR, class A macrophage scavenger receptors. Aß M1 Wilcock & Colton, surface Binds Aß. 2010; Hickman et al., 2008 CD36 Microglia PRR, class B scavenger receptor. Binds Aß. → Aß, prion M1, Colton, 2009; Wilcock surface cell adhesion and apoptosis. protein M2a, & Colton, 2010; fragment M2c Kouadir et al., 2012; Hickman et al., 2008

255

Factor Location Function Expression State Ref determinants NFκB1, RELA Intracellular, all NFκB is a transcription mediator, RELA is part Induced by M1 Darafourt et al., 2012 cells. of the NFκB complex. GM- CSF+IFNγ/LP S STAT1, STAT4 Transcription mediators Induced by M1 Colton, 2009 IFNγ Caspases (1, 3, 7, Intracellular Apoptosis. Induced by M1 Wilcock & Colton, 8) LPS, IFNγ; 2010; Burguillos et al., caspase (3/7) 2011 necessary for M1 IFNγ ↑ Released Cytokine Induces M1, M1; Colton et al., 2006; induced by aging Colton, 2009; Wilcock

256 GM- & Colton, 2010; +

CSF IFNγ/LP Darafourt et al., 2012; S Varnuum & Ikezu, 2012; Frank IFIT3↑↑ Cytokine-associated. Interferon-induced protein. Induced by M1 Darafourt et al., 2012 GM- CSF+IFNγ/LP S TNFα ↑ Released Cytokine Induces M1, M1 > Colton et al., 2006; induced by Ø > Colton, 2009; Wilcock IFNγ, LPS, M2 & Colton, 2010; reduced byIL- Darafourt et al., 2012; 4 & IL-13 Varnuum & Ikezu, 2012; Chen et al., 2008; Njie et al., 2012

256

Factor Location Function Expression State Ref determinants TNFR1 Cytokine-associated. Induces M1, M1 Wilcock & Colton, upon 2010 activation TRADD Cytokine-associated. TNFR1-associated death M1 Wilcock & Colton, domain → apoptosis 2010 TNFR6 Cytokine-associated. TNFR → apoptosis. Induced by M1 Darafourt et al., 2012 GM- CSF+IFNγ/LP S IL-1ß, IL-6 Released Cytokines (interleukins) Induce M1, M1 Colton, 2009; Wilcock Induced by & Colton, 2010; GM- Darafourt et al., 2012; CSF+IFNγ/LP Cameron & Landreth, S, LPS, IL-1ß 2010. Wynne et al., 257 ↑ in aging 201; Chen et al., 2008; ;

Njie et al., 2012 IL-8↑↑, IL-12A↑, Released Cytokines (interleukins) Induced by M1 Colton, 2009; Wilcock IL-12B↑↑↑ GM- & Colton, 2010; CSF+IFNγ/LP Darafourt et al., 2012 S Chemokines: Released Cytokines (C-C motif chemokines) Induced by M1 Darafourt et al., 2012; CCL2, CCL3↑, GM- Cameron & Landreth, CCL19↑↑, CSF+IFNγ/LP 2010 CCL3L1↑↑, S CCL4↑, CCL5↑, CCL7↑, CCL8↑,

257

Factor Location Function Expression State Ref determinants Chemokines: Released Cytokines (C-X-C motif chemokines) Induced by M1 Darafourt et al., 2012 CXCL1↑, GM- CXCL2↑↑, CSF+IFNγ/LP CXCL10↑↑↑, S CXCL12 GM-CSF Granulocyte-macrophage colony stimulating factor NO, ROS Released, in Break down proteins, lipids and DNA of cells M1 Colton & Wilcock; phagosomes and pathogens. Cmeron & Landreth, 2010 NADPH oxidase Microglia, MΦ Redox. NADPH oxidase synthesizes ROS/RNS M1 Wilcock & Colton, (NOX1, NOX2, and neutrophils. including superoxide. 2010 P22phox, P40phox, P47phox, P67phox, RAC1, NOXA1, 258 NOXO1)

MPO Neutrophil Redox. Synthesizes ROS. M1 Wilcock & Colton, granulocyte 2010 lysosomes iNOS Redox, inducible nitric oxide synthase. Induced by M1 Colton et al., 2006; Synthesizes NO from L-arginine. IFNγ Colton, 2009; Wilcock & Colton, 2010; Cameron & Landreth, 2010 CAT2 Cationic amino acid transporter 2 supplies M1, Colton 2009 arginine to be used in production of NO. M2a

258

Factor Location Function Expression State Ref determinants GLT1 (EAAT2, Astrocyte, Excitotoxicity. Glutamate transporter M1 Wilcock & Colton, SLC1A2) microglia 1/excitatory amino acid transporter 2/ solute 2010 surface carrier family 1 member 2. In astrocytes, GLT1 ↓ extracellular glutamate, ↑ antioxidants. xCT (slc7a11) Astrocyte, Excitotoxicity. Cystine/glutamate transporter. ↑ M1 Wilcock & Colton, microglia extracellular glutamate. 2010 surface mGluR2 Excitotoxicity. Inhibits cAMP. M1 Wilcock & Colton, 2010 P2RX7 Microglia Excitotoxicity. ATP receptor, ATP-dependent M1 Wilcock & Colton, surface lysis of macrophages. 2010 MHCI Surface of all Antigen presentation of cytosolic proteins (self, Induced by M1 Darafourt et al., 2012 cells viruses) to CD8+ T cells GM- CSF+IFNγ/LP S 259 TAP2↑ Microglia ER Transports antigen peptides from cytosol into ER Induced by M1 Darafourt et al., 2012; to load MHC I. GM- Parcej & Tampe, 2010 CSF+IFNγ/LP S MHCII Microglia, MΦ, Antigen presentation to T cell receptor on CD4+ Induced by M1, Wilcock & Colton, astrocyte surface T cells GM- M2a 2010; Colton 2009; CSF+IFNγ/LP > Darafourt et al., 2012; S M2c, Cameron & Landreth, Ø; 2010; Melief et al., aging 2012; VanGuilder et al., 2011; Frank; Godbout et al., 2005; Frank et al., 2010

259

Factor Location Function Expression State Ref determinants CD80↑ Microglia, MΦ Antigen presentation, MHCII co-factor, binds Induced by M1 > Bechmann et al., 2001; surface CD28 and CTLA4 on T cells. IFNγ, GM- M2 Satoh et al., 1995; CSF Wilcock & Colton, 2010 CD86 Microglia, Antigen presentation, MHCII co-factor, binds Induced by M1 < Bechmann et al., 2001; astrocyte surface CD28 and CTLA4 on T cells. IFNγ, GM- M2, Satoh et al., 1995; CSF Ø; Wilcock & Colton, aging 2010; Frank; Frank CD1A↑↑, CD1B↑ Microglia Antigen presentation, lipids to T cells Induced by M- M2 Darafourt et al., 2012 surface CSF+IL-4/IL- 13 Mac-1 Microglia, T Constitutively expressed PRR (complement Induced by M- M1 < Darafourt et al., 2012; (CD11b/CD18, cell, monocyte receptor) → chemotaxis, phagocytosis, lysis CSF+IL-4/IL- M2 > Lynch 2009 ITGAM/ITGB2, surface 13, LPS, Aß Ø; CR3) aging 260 MR (Mrc1, Microglia, MΦ, PRR, c-type lectin (CLR), mannose receptor, Induced by IL- M2a, Colton et al., 2006;

CD206) DC surface binds glycoproteins and glycolipids → 4, IL-13, M2c Colton, 2009; Cameron complement, phagocytosis dexamethason & Landreth, 2010; e Colton, 2009; M2c but not M2a by Melief et al., 2012 DC-SIGN Microglia, MΦ, PRR, c-type lectin (CLR), dendritic cell-specific Induced by M- M2 Colton, 2009; Darafourt (CD209)↑↑↑ DC surface intracellular adhesion molecule 3-grabbing CSF+IL-4/IL- et al., 2012; Colton, integrin → phagocytosis, dendritic cell adhesion 13 2009 FIZZ (FCεR2)↑↑↑ Microglia, MΦ, PRR, c-type lectin (CLR), found in inflammatory Induced by IL- M2a monocyte, T cell zone 1 4 surface

260

Factor Location Function Expression State Ref determinants Chi3L1, Chi3L2, Chitinase-like lectins Induced by IL- M2a YM1 (Chi3l3) 4 LRP1 LDL receptor-related protein 1 Induced by M- M2 Darafourt et al., 2012 CSF+IL-4/IL- 13 PPARγ Nuclear receptor → fatty acid storage and M2a Wilcock & Colton, glucose metabolism. 2010; Cameron & Landreth, 2010 STAT6 Transcription mediator. Responds to IL-4 → IL-4 M2a Colton, 2009 anti-apoptotic effect. Arg1 Cystolic Competes with iNOS for arginine, ↓ NO Induced by IL- M2a Colton et al., 2006; production. Repair, polyamine synthesis. 4, IL-13 Colton, 2009; Wilcock & Colton, 2010 IL-4, IL-13 Cytokines (interleukins) Induce M2a, M2a Colton et al., 2006; LPS ↓ IL-4 Colton, 2009; Wilcock 261 & Colton, 2010; Varnuum & Ikezu, 2012; Cameron & Landreth, 2010; Wynne et al., 2010 IL-16 Cytokines (interleukins) Induced by M- M2 Darafourt et al., 2012 CSF+IL-4/IL- 13 CCL13↑ Cytokines (C-C motif chemokines) Induced by M- M2 Darafourt et al., 2012 CSF+IL-4/IL- 13 IL-1Ra Cytokines (interleukins). IL-1 receptor M2a, Colton, 2009; Wilcock antagonist. M2c & Colton, 2010

261

Factor Location Function Expression State Ref determinants M-CSF Macrophage colony-stimulating factor M2 CSF1R Colony stimulating 1 receptor Induced by M- M2 Darafourt et al., 2012 CSF+IL-4/IL- 13 CD200, CD200R R on myeloid Neuron↔Glia communication. Deactivates ↓ in aging M2 Wilcock & Colton, cell surface, … microglia. 2010; ; Lynch 200; RFrank9 CB2 Microglia, Neuron↔Glia communication, decrease AD pathology, M2 Maresz et al., 2005; astrocyte surface glutamate IFNγ/GM-CSF Marchalant et al., 2009; Stella, 2010; Wilcock & Colton, 2010 CD45 Low expression used to distinguish microglia M2 Wilcock & Colton, from macrophages. Binds CD22. 2010 CEBPA Induced by M- M2 Darafourt et al., 2012 CSF+IL-4/IL- 262 13 CD163 Hemoglobin scavenger receptor. M2a, Colton, 2009; Melief et M2c al., 2012; colton 2009 HO-1 Heme-oxygenase 1 M2 Wilcock & Colton, 2010 mGluR5 Neuron↔Glia communication, glutamate M2 Wilcock & Colton, receptor. 2010 TREM2 Neuron↔Glia communication. Binds apoptotic M2 Wilcock & Colton, cells. 2010 Apoptotic cells Bind to TLR4/CD14, scavenger receptors and Induces M2c Colton, 2009 TREM2.

262

Factor Location Function Expression State Ref determinants IL-10 (↑ high) Cytokines (interleukins) Induces M2c Colton, 2009; Wilcock & Colton, 2010; Varnuum & Ikezu, 2012; Cameron & Landreth, 2010 IL-4Rα IL-4 receptor α Induced by M2c Colton, 2009; Wilcock LPS (not in & Colton, 2010; Fenn et aged) al., 2012 CCL18 M2c Colton 2009 TGFß Repair. Transforming growth factor controls Induces M2c Colton, 2009; Wilcock proliferation, differentiation and apoptosis & Colton, 2010; (through SMAD). Varnuum & Ikezu, 2012; Cameron & Landreth, 2010 SMAD Microglia Transduce TGFß signal to transcription mediators M2c Colton, 2009 263 cytoplasm in nucleus.

SOCS1, SOCS3 Suppressors of cytokine signaling M2c Colton, 2009; Wilcock & Colton, 2010 STAT3 Transcription mediators M2c Colton, 2009 SphK-1/2 Cytosol Sphingosine kinase 1/2, catalyze sphingosine 1 M2c Colton, 2009; Wilcock (SphK1), phosphate (S1P), ↓ apoptosis in neighboring & Colton, 2010 nucleus (SphK2) cells. S1PR1 Sip receptor. M2c Wilcock & Colton, 2010 COX2 Anti-inflammatory M2c Wilcock & Colton, 2010 PGE2 M2c Wilcock & Colton, 2010

263

Factor Location Function Expression State Ref determinants Glucocorticoids M2c Varnuum & Ikezu, 2012 NGF Repair. Nerve growth factor. M2c Wilcock & Colton, 2010 MMP3, MMP9 Repair. Matrix metallopeptidases M2c Wilcock & Colton, 2010 TIMP Inhibits metaloproteases. M2c Wilcock & Colton, 2010 Fractalkine Released from Neuron↔Glia communication. Deactivates LPS ↓ Ø Lynch 2009; Wynne et (CX3CL), neurons, R on microglia. CXCR1, al., 2010 Fractalkine microglia, MΦ, CXCL1 ↓ in R(CX3CR1) leukocyte, aging

264 astrocyte,

neuron surface

264

Appendix C: Epidemiological evidence and clinical trials of NSAIDs and other anti- inflammatories in AD and PD

Studies are organized by trial type (epidemiological then interventional). Drugs are categorized by their actions (Mechanism) and specified (Drug) in interventional studies and in epidemiological studies when possible. Duration (Dur) refers to the period of time evaluated in retrospective and prospective studies and the duration of drug exposure in clinical trials. Baseline diagnosis refers to the criteria by which subjects were selected for each study and multiple listings indicate that groups were compared. All of the interventional trials began with cohorts that were already diagnosed with AD. Evidence of reduced risk of AD and PD and therapeutic value in AD and PD are highlighted in grey. Refer to previous text for detail

265

Drug Mechanism Type Dur Baseline n Outcome Selected references Retrospective and prospective epidemiological studies NSAIDs NSAID Retrospective AD, controls 793 ↓ AD risk (CSHA, 1994) NSAIDs NSAID Retrospective AD, twins 100 ↓ AD risk (Breitner et al., 1994) NSAIDs NSAID Retrospective AD, family 1664 ↓ AD risk (Yip et al., 2005) NSAIDs NSAID Prospective 5 yrs. Normal 4,615 ↓ AD risk (Lindsay et al., 2002) NSAIDs NSAID Retrospective Elderly 2,708 ↓ AD risk (50% lower) (Landi et al., 2003) NSAIDs NSAID Prospective 15 yrs. Normal 1,686 ↓ AD risk ≥ 2 yrs. use (Stewart et al., 1997) ↓ AD risk in ApoE4 NSAIDs NSAID Prospective 3 yrs. Normal 3,229 (Szekely et al., 2008a) carriers (Andersen et al.,

26 NSAID Prospective 10 yrs. AD, controls 306 ↓ AD risk, ≥ 6 mo. use 1995, Veld et al., 6 NSAIDs 1998) ↓ AD risk, early but not NSAIDs NSAID Retrospective 3 yrs. Elderly 3,227 (Zandi et al., 2002) recent use, ≥ 2 yrs.

NSAIDs NSAID Retrospective 3 yrs. AD, controls No AD protection (Wolfson et al., 2002) NSAIDs NSAID Retrospective PD, controls 392 No PD protection (Bower et al., 2006) > 5 NSAIDs NSAID Retrospective PD, controls 589 No PD protection (Ton et al., 2006) yrs. NSAIDs NSAID Retrospective 6 yrs. Normal No PD protection (Etminan et al., 2008) ~9 (Bornebroek et al., NSAIDs NSAID Prospective Normal 6,512 No PD protection yrs. 2007) NSAIDs NSAID Prospective 12 yrs. Normal 2736 ↑ AD risk (Breitner et al., 2009) NSAIDs + ↓ AD risk in ApoE4 NSAID Prospective 8 yrs. Elderly 3,376 (Fotuhi et al., 2008) vitamins E & C carriers NSAID Prospective 14 yrs. Normal 142,902 ↓ PD risk (by 45%) (Chen et al., 2003) NSAIDs ↓ PD risk in men, ↑ PD (non-aspirin) NSAID Retrospective PD, control 7896 (Hernán et al., 2006) risk in women 266

Drug Mechanism Type Dur Baseline n Outcome Selected references Aspirin NSAID Retrospective Elderly twins 351 ↓ AD risk (Nilsson et al., 2003) COX inhibitor, not Prospective 15 yrs. Normal 1,686 No AD protection (Stewart et al., 1997) Acetaminophen anti-inflammatory Prospective 3 yrs. Normal 3,229 No AD protection (Szekely et al., 2008a) ~9 Prospective Normal 6,992 ↓ AD risk (Haag et al., 2009) Lipid lowering, yrs. Statins Anti-inflammatory Prospective - ↓ AD risk (Sparks et al., 2008) Retrospective PD, control 7,274 No PD Protection (Becker et al., 2008) Interventional studies No AD protection, ↑ Aspirin NSAID Clinical trial 3 yrs. AD 310 (Bentham et al., 2008) bleeds (Scharf et al., 1999,

Mild-mod Aisen et al., 2003, NSAID Clinical trial 25 wk. 41 No AD protection AD Reines et al., 2004,

Diclofenac Thal et al., 2005)

267 Mild-mod ↓ AD decline in ApoE4 (Pasqualetti et al., Ibuprofen NSAID Clinical trial 1 yr. 132

AD carriers 2009) Mild-mod Clinical trial 6 mo. ↓ AD decline (Rogers et al., 1993) AD Indomethacin NSAID 12 Mild-mod Clinical trial 51 No AD protection (de Jong et al., 2008) mo. AD (Scharf et al., 1999, Aisen et al., 2003, Mild-mod Reines et al., 2004, Naproxen NSAID Clinical trial 1 yr. 351 No AD protection AD Thal et al., 2005, Gómez-Isla et al., 2008)

267

266 Drug Mechanism Type Dur Baseline n Outcome Selected references No AD protection, ↑ (Martin et al. 2008; cardiovascular risk; ≥70 + AD Leoutsakos et al. Clinical trial - 2528 protection depending family history 2012; Breitner et al. on early disease state 2011) and NSAID duration (Scharf et al., 1999, ≤ 4 Mild-mod Aisen et al., 2003, Clinical trial 1457 No AD protection yrs. AD Reines et al., 2004, Thal et al., 2005) Refocoxib NSAID Mild-mod Clinical trial 1 yr. 692 No AD protection (Reines et al., 2004) AD Mild-mod Clinical trial 1 yr. 351 No AD protection (Aisen et al., 2003) AD

268

No AD protection, ↑

(Martin et al. 2008; cardiovascular risk; NSAID, COX2 ≥70 + AD Leoutsakos et al. Celecoxib Clinical trial - 2528 protection depending inhibitor family history 2012; Breitner et al. on disease state and 2011) NSAID duration 16 Prednisone Corticosteroid Clinical trial AD 138 No AD protection (Aisen et al., 2000) mo. Hydroxy- Anti-malarial 18 Mild-mod (Van Gool et al., Clinical trial 168 No AD protection chloroquine Anti-inflammatory mo. AD 2001) (de Jong et al., 2008, 13 Mild-mod Triflusal Anti-inflammatory Clinical trial 257 No AD protection Gómez-Isla et al., mo. AD 2008)

268

Appendix D: Summary of outcomes from i.c.v. LPS infusion

Summary of results of i.c.v. LPS infusion (highlighted in grey) or single injection. Summary includes data about mRNA levels which are not distinguished from increases in protein expression. Rats are typically 3-4 months of age, rats labeled as 'adult' are 9 months of age and rats labeled 'aged' are 24 months of age.

269

Animal Location Dose Duration Measurement Results Reference Microglia and astrocytes Rats IVth 0.25µg/hr 6d 6d ↑ activated microglia in ventricles (Rosenberger et al., 2004) and pia Rats IVth 0.25µg/hr 2w 2w ↑ microglia activation (MHCII+) (Wenk et al., 2004, Rosi et al., 2005a) Rats IVth 0.25µg/hr 3w 3w ↑ microglia activation (MHCII+), (Marchalant et al., 2007); Bardou et hippocampus and substantia nigra al., 2013; Rats IVth 0.25µg/hr 4w 4w ↑ microglia activation (MHCII+), (Hauss-Wegrzyniak et al., 1998a, hippocampus and substantia nigra Hauss-Wegrzyniak et al., 1999b, Wenk et al., 2004, Rosi et al., 2005b, Rosi et al., 2006, Rosi et al.,

27 2009); Chapters 2-5

0 th + Rats IV 0.25µg/hr 5w 5w ↑ microglia activation (MHCII ) (Hauss-Wegrzyniak et al., 1999a, Hauss-Wegrzyniak et al., 2000b, Hauss-Wegrzyniak et al., 2000c) Rats IVth 0.25µg/hr 5w 10w ↑ microglia activation (MHCII+) (Hauss-Wegrzyniak et al., 2000c) Rats IVth 0.25µg/hr 10w 10w ↑ microglia activation (MHCII+) (Hauss-Wegrzyniak et al., 2000c) Rats, adult IVth 0.25µg/hr 5w 5w ↑ microglia activation (MHCII+) (Hauss-Wegrzyniak et al., 1999a) Rats, adult IVth 0.25µg/hr 5w 5w Did not ↑ microglia activation past (Hauss-Wegrzyniak et al., 1999a) basal increase Rats Lateral 0.25µg/hr 4w 4w ↑ microglia activation (MHCII+) (Richardson et al., 2005) Mice, aged Lateral 0.01µg Injection 2h ↑ microglia activation (MHCII+) (Huang et al., 2008) Rats Lateral 0.005 µg Injection 2-24h ↑ microglia activation (MHCII+) (Sugaya et al., 1998) (after 6h)

270

Animal Location Dose Duration Measurement Results Reference Astrocytes Rats IVth 0.25µg/hr 6d 6d ↑ astrogliosis (Rosenberger et al., 2004) Rats IVth 0.25µg/hr 2w 2w → astrocyte hypertrophy (Rosi et al., 2005a) Rats IVth 0.25µg/hr 4w 4w ↑ Astrogliosis (Hauss-Wegrzyniak et al., 1998a) Rats Lateral 0.25µg/hr 4w 4w ↑ Astrogliosis (Richardson et al., 2005) Rats Lateral 0.005 µg Injection 2-24h ↑ astrogliosis (after 6h) (Sugaya et al., 1998) Inflammatory products Rats IVth 0.25µg/hr 8w 8w ↑ IL-1ß brainstem and Bardou et al., 2013 hippocampus; ↑ IL-1α, TNFα CX3CR1 and TGFß in brainstem Rats IVth 0.25µg/hr 3w 3w ↑ IL-1ß & TNFα in hippocampus Bardou et al., 2013; Chapter 6 27 and brainstem; ↑ IL-1α, TNFα

CX3CR1 and TGFß in brainstem Rats IVth 0.25µg/hr 4w 4w ↑ IL-1ß, TNFα (Hauss-Wegrzyniak et al., 1998a) Rats IVth 0.25µg/hr 6d 6h-6d ↑ prostaglandins, arachidonic acid, (Rosenberger et al., 2004) No change COX Rats Lateral 7.5µg Injection 15min – 3h ↑ TNFα (ventricular) (Sanna et al., 1995) Rats Lateral 2.5µg Injection 1- 4h ↑ TNFα and IL-6 (peak 2h) (Terrazzino et al., 1997) Rats, aged Lateral 2.5µg Injection 1-4h ↑ TNFα and ↑ IL-6 above basal increase of IL-6 (peak 2h), Rats Lateral 2.5µg Injection 1-16h ↑ IL-1ß and IL-6 (peak 4-8) (De Simoni et al., 1995) Mice Lateral 1µg Injection 1.5h ↑ TNF and IL-6 (Lund et al., 2005) Rats Lateral 5-10µg Injection 2-20h ↑ TNFα (peak 2h) (Zujovic et al., 2001) Mice, aged Lateral 0.01µg Injection 2h ↑ IL-1β, IL-6, and TNFα (Huang et al., 2008) Rats Lateral 87-112µg Injection 3- 6h ↑ TNFα, IL-1ß and COX2 (Zhang and Rivest, 2001) 271

Animal Location Dose Duration Measurement Results Reference Rats IIIrd 0.5µg Injection 5h ↑ IL-1ß, Il-1 receptor antagonist, (Ilyin et al., 1998) Il-1 receptor and TNFα; no change TGF-ß, NPY Mice Lateral 2.5µg Injection 6h ↑ IL-1α, IL-1ß, IL-6, MCP-1 (Szczepanik and Ringheim, 2003) Mice Lateral Injection 24h ↑ TNFα and I-1ß (Tyagi et al., 2008) Mice Lateral 2.5µg Injection 24h ↑ MCP-1 (Szczepanik and Ringheim, 2003) Lateral 2.5µg Injection 48h No change IL-1α, IL-1ß, IL-6, MCP-1 Rats Lateral 0.5µg Injection 5-7d ↑ IL-1ß, TNFα, prostaglandin and (Xia et al., 2006) COX2 Rats Lateral 0.5µg Injection 5-7d ↑IL-1ß, TNFα (Plata-Salaman et al., 1998) CD14/TLR4 cascade 27 Rats IVth 0.25µg/hr 8w 8w ↑ TLR4 in brainstem Bardou et al., 2013

2 Rats IVth 0.25µg/hr 4w 4w ↑ TLR4 Chapter 6 Rats IVth 0.25µg/hr 4w 4w ↑ p-Erk Chapter 2 Rats IVth 0.25µg/hr 3w 3w ↑ TLR4 in brainstem Bardou et al., 2013 Rats IVth 0.25µg/hr 2w 2w ↑ NFκB (Rosi et al., 2005a) Mice IVth 2.5µg Injection 1-3h ↑ NFκB and IκBα (Ichiyama et al., 1999) Mice, aged Lateral 10ng Injection 2h ↑ TLR2 and TLR4 (Huang et al., 2008) Rats Lateral 87-112µg Injection 3-6h ↑IκBα (Zhang and Rivest, 2001) Rats Lateral 0.5µg Injection 5-7d ↑CD14 mRNA, TLR4 expression (Xia et al., 2006) and ↑ IκBα Oxidative stress Mice Lateral 1µg Injection 10-72h ↑ oxidative damage to neuronal (Montine et al., 2002) membranes (peak 24) 272

Animal Location Dose Duration Measurement Results Reference Rats Lateral 5-10µg Injection 2-20h ↑ oxidative stress (peak 4h) (Zujovic et al., 2001) Rats Lateral 0.005µg Injection 2-24h ↑ iNOS (glia), SOD (neurons), (Sugaya et al., 1998) HSP70 Cell loss Rats IVth 0.25µg/hr 4w 4w ↓ pyramidal cells EC II/III (Hauss-Wegrzyniak et al., 2002) ↓hippocampus size & ↑ lateral Rats IVth 0.25µg/hr 4w 4w (Hauss-Wegrzyniak et al., 2000a) ventricle Memory, cellular level Rats IVth 0.25µg/hr 4w 4w ↓ LTP (Hauss-Wegrzyniak et al., 2002) Rats IVth 0.35µg/hr 4w 4w CA3 (not CA1) shows disrupted (Rosi et al., 2005b, Rosi et al., Arc signaling 2006, Rosi et al., 2009) Rats IVth 0.35µg/hr 4w 4w ↓ LTP CA1 (Min et al., 2009)

3 Memory performance

Rats IVth 0.25µg/hr 3w During w3 ↓ spatial memory (Marchalant et al., 2007); Bardou et al., 2013; Chapter 6 Rats IVth 0.25µg/hr 4w During w4 ↓ spatial memory (Hauss-Wegrzyniak et al., 1998a) Rats IVth 0.35µg/hr 4w During w4 ↓ spatial memory (Min et al., 2009) Rats IVth 0.25µg/hr 5w During w5 ↓ spatial memory (Hauss-Wegrzyniak et al., 1999a, Hauss-Wegrzyniak et al., 2000b, Hauss-Wegrzyniak et al., 2000c) Rats IVth 0.25µg/hr 8w During w8 restored spatial memory Chapter 5 & 6 Rats IVth 0.25µg/hr 8w During w8 ↓ spatial memory Bardou et al., 2013

273

Animal Location Dose Duration Measurement Results Reference Rats IVth 0.25µg/hr 5w During w10 ↓ spatial memory Hauss-Wegrzyniak et al., 1999a, Hauss-Wegrzyniak et al., 2000b, Hauss-Wegrzyniak et al., 2000c) Rats IVth 0.25µg/hr 10w During w10 ↓ spatial memory (Hauss-Wegrzyniak et al., 2000c) Rats Lateral 0.25µg/hr 4w During w4 Behavioral deficit? (Richardson et al., 2005) Rats IVth 0.25µg/hr 5w During w5 Does not ↓ spatial memory past (Hauss-Wegrzyniak et al., 1999a) basal impairment

274

Appendix E: Experiments using Ceftriaxone with evidence of increased glutamate transport and/or cognitive effects.

Beta lactam Dose/delivery Time Model Outcome Citation antibiotic period Ceftriaxone 200 mg/kg, ip On day of Sprague- ↑ glt1 Hota et al., and 7 or 14 dawley 3 expression, ↓ 2008 days after mo, excitotoxicity insult and hypoxia and ↑ memory recovery after hypoxia Ceftriaxone 200 mg/kg ip Daily for 5 7 week HD ↑ glt1 striatum Sari et al., days mice when at 13 2010 weeks when normally decreased Ceftriaxone 200 mg/kg ip Daily 7 Rats, ↑ GLT1 (by Knackstedt days cocaine WB) in NAC et al., 2010 seeking and xCT levels ↓ drug seeking – ceftriaxone did not ↑ glt1 in drug naïve rats Ceftriaxone 200 mg/kg ip 1 week mice ↓ visceral Lin et al., nocicemption 2009 Ceftriaxone 50, 100, 200 Daily 5 Rats, 50 didn’t wor, Sari et al., mg/kg ip days cocaine 100 and 200 2009 seekeing did. ↑ GLT1 expression in PFC and NAC Ceftriaxone 200 mg/kg/day Mouse MS ↓ MS Melzer et ip EAE symptoms al., 2008 through T cell regulation, but not by modulation of glutamate homeostasis, bc ceftriaxone still worked when used with

275

EAAT2 inhibitor dihydrokainate (10 mg/kg/day) Ceftriaxone 200 mg/kg/day 5 days HD mice ↑ GLT1 and ↓ Miller et al., ip HD 2008 Ceftriaxone 50, 100 or 200 5 days Spraque ↓ Glu 40-50% Ramussen et mg/kg ip dawley (microdialysis) al., 2010 up to 20 days (in 200 dose) following discontinuation of ceftriaxone. ceftriaxone effect was prevented by the GLT-1 transporter inhibitor dihydrokainate (1 lM, intra- accumbal) Ceftriaxone 100 mg/kg/day 5 days Ischemia protection Verma et al., 2010

276

Appendix F: Experiments using Riluzole with evidence of increased glutamate transport and/or cognitive effects.

Riluzole: In vivo studies with central or behavioral effects Reference Dose Model Effect 1989 4 mg/kg, in 0.9% Gerbil, ↓ memory loss, neuronal damage Malgouris NaCL and 4% HCl ischemia 0.1N (pH 3.5), ip, day of and 2x daily 14 days after ischemia 1991 4-10 mg/kg, deionized Mice Blocked acquisition of learned fear Sanger water containing 2 drops of Tween 80 in 10 ml water. Injection volume was 20 ml/kg ip 1991 10 nmol icv in DMSO Rat, drug- ↓ MCD and DTX but not 4-AP Stutzmann or 4 mg/kg oral prior induced seizures to convulsant seizure 1993 Wahl 4, 8 mg/kg after Rat, ↓ Infarct, no effect on neurological or occlusion ischemia memory tests, ↑ striatum necrosis. 1994 8 mg/kg in HCl 0.1N Rat, Riluzole, and CNQX, an Heurteaux and 0.9% saline ip 30 forebrain AMPA/kainate antagonist, reduced min after ischemia ischemia ischemia-induced expression of NMDA receptors in hippocampus while D-AP5, a selective NMDA antagonist, had no effect 1995 4 mg/kg once prior or Rhesus, ↓ bradykinesia and rigidity Benazzouz daily MPTP 1995 2x 10 mg/kg po Mice, No effect on ↓ DA, unlike MK-801 Boireau methamphe tamine- induced ↓ DA 1995 Mary 4, 8 mg/kg/ os, Rat, ↓ lesion dissolved in HCl 0.1 quinolinate N after quinolinate HD 1996 Mice, ALS ↑ survival Gurney

277

Riluzole: In vivo studies with central or behavioral effects Reference Dose Model Effect 1998 2, 4 mg/kg. Dissolved Rats, Riluzole influenced neither Kretschmer in cremophor 10%, various stereotyped sniffing behavior nor single ip before glutamater locomotion but impaired motor behavioral testing gic drugs coordination and attenuated rigidity induced by blockade of dopamine D1 and D2 receptor antagonists when given alone. At higher doses spontaneous behavioral activity decreased and motor coordination was more impaired. Augmentation of the riluzole effects were observed when NMDA, but not GYKI 52466, was coadministered. The glycine site agonist DCS increased the anticataleptic properties of riluzole. 1998 Kwon 2, 8 mg/kg Rabbits, No effect ischemia-induced transient glutamate accumulation global ischemia 1998 4 mg/kg, dissolved in Rats, Induced conditioned place preference Tzschentke 10% cremophor, Morphine, alone, but prevented drug-induced injectior prior to amphetami ccp behavior ne induced conditioned place preference 1998 Zhang 8 mg/kg iv 15 min, 6hr Rat TBI ↓ lesion, not hippo neuronal loss and 24hr after injury. (fluid vehicle is 0.9% saline percussion) containing 1.5% poloxamer 188 1999 4-12 mg/kg ip in 0.01 Rat, ↓ seizure, neurodegeneration in Kanthasamy N hydrochloric acid, cardiac hippo and cerebellum i.p.. right after cardiac arrest arrest, 24hr and 48hr induced (right before testing) seizure 2000 2,4,8 mg/kg. 3.5hr or Rat ↓ hyperalgesia and Riluzole Abarca 48hr before testing. hyperalgesi decreased the concentrations of dissolved in 10% a glutamate and aspartate and those of Tween 80. and citrulline and arginine in the administered s.c. in a presence or absence of painful volume of 1 ml/kg stimulation in the ventral posterolateral thalamic nuclei

278

Riluzole: In vivo studies with central or behavioral effects Reference Dose Model Effect 2000 5, 10, 20 mg/kg ip, Mice, Did not prevent convulsions of Brackett pre-treatment. cocaine- lethality dissolved in 50% induced DMSO and 50% convulsion distilled water. s 2000 Itzhak 2.5-20 mg/kg pre- Mice, Did not effect expression nor treatment, dissolved in cocaine and induction of sensitization to cocaine, 10% cremophor meth- (20 mg/kg) blocked induced the acute response to METH on day behavioral 1 and the expression of the sensitized sensitizatio response on day 5 but not the n induction of sensitization to METH 2000 Mu 8 mg/kg pre-treatment, Rat TSCI ↑ mitochondrial function and dissolved in a small enhance glutamate and glucose volume of HCl (0.1 uptake [like Chowdhury 2008] N), diluted in saline at a concentration of 2.5 mg/ ml, and the pH adjusted to 7.4 with NaOH. 2000 Pena 8 mg/kg ip Rat, seizure and riluzole instead markedly by K+ increased the intensity and duration channel of the disharges. blocker Moreover, at the highest dose tested (8 mg/kg, i.p.), riluzole caused a 75% mortality of the rats 2001 Araki 3, 10 mg/kg ip, in Mice, ↓ loss of DA DOPAC and HVA, a,b 0.5% carboxymethyl- MPTP protect SN neurons M PBS, the sections were incubated with biotinylated cellulose?? 30 min prior and 90 min after MPTP 2001 4-8mg/kg ip 30 min Rat, seizure ↓ kindling and seizure in amygdala Yoshida prior to stimulation over 14 days, dissolved in 50% polyethylene glycol 300 in distilled water 2002 8 mg/kg oral 2x/day, 2 Mice, ↑ density of striatal dopaminergic Douhou months, suspended in weaver PD nerve terminals. 0.5% methylcellulose

279

Riluzole: In vivo studies with central or behavioral effects Reference Dose Model Effect 2002 Mao 5-20µg intrathecal Rat, ↓ tolerance and hyperalgesia (spinal cord) acute or morphine- (blocking glutamate transport ↑ 2x daily for 7 days. induced these). hyperalgesi a and ↓ glutamate transporters 2002 4 nmol/rat, Rat, blocked Palazzo microinjection capsaicin in periaquedu ctal grey 2002 10 mg/kg in 0.1 ml Mouse, HD ↑ survival time, ↓ nuclear inclusion Schiefer saline, oral, 3 weeks formation until spontaneous death 2003 10 mg/kg 2/day, i.p., Mouse, ↓ inflammation, T-cells to CNS, Gilgun- before or after clinical MS, EAE demylination, axonal disruption Sherki symptoms. Sigma. Diluted in DMSO and diluten in saline to conc, up to 15 days 2003 10, 30 mg/kg ip, 30 Rats, PD, ↓ drug-induced rotations, unlike Gilgun- min prior to drug- 6OHDA NMDAR antagonist amantadine or Sherki induced rotations synthetic cannabinoid HU-211 or GABA agonist 2003 3, 10 mg/kg pre- MK-801- Moderately ↓ hyperlocomotion (10 Lourenco treatment induced mg/kg) Da Silva hyperloco motion 2003 Sung 10µg intrathecal. Rat, ↓ pain Post-op 2x/day days 1- neuropathic 4 or 5-8. Sigma. 10% pain, spinal DMSO diluted in cord saline. 2004 10 mg/kg in sterile Mice, drug ↑ antiseizure activity of anti- Borowicz saline, 30 min before induced epileptics, motor deficits with some testing seizure combinations, no long-term memory impairments

280

Riluzole: In vivo studies with central or behavioral effects Reference Dose Model Effect 2004 2-8 mg/kg/day ip Rat, post- ↑ ambulation, motor performance Medico starting on injury for 3 traumatic and coordination days, behavior peripheral assessed 7 days after neuropathy end of treatment, dissolved in 2- hydroxypropyl- b- cyclodextrin 2004 Rat, ??? Sepulveda nucleus acumbens, cessation of chronic morphine 2004 8 mg/kg/day, 7 days Rat, optic ↓ rise in glutamate like memantine Vorwerk prior to and after (14 nerve crush and nimodipine days total) crush. 2006 Jin 2-8 mg/kg. Sigma. Rat, ↓ conditioned place aversion dissolved in 0.1 N morphine/n hydrochloric acid. aloxone- injected sc 1 ml/kg, induced day of testing conditioned place aversion 2006 2-8 mg/kg ip 30 min Rat, 2 mg/kg preserved neuronal density Risterucci before and after lesion. excitotoxic in lesion and ↓ motor deficits (like dissolved with NaCl PFC mGluR5 antagonist) (0.9%) in a minimum lesions quantity of 0.1 M HCl 2007 6, 9, 12 mg/kg prior to Rat, sciatic ↓ glutamate, asparate (basal and Coderre and ever 12h for 4 nerve formalin-induced) in the spinal cord days following injury injury dorsal horn ↓ nociception (12 (testing 4, 8, 12 days). mg/kg) Dissolved 10% cyclodextrin in 0.9% saline

281

Riluzole: In vivo studies with central or behavioral effects Reference Dose Model Effect 2008 21 days, 4 mg/kg per Rat, PFC ↑ glucose metabolism, glutamate Chowdhury day, i.p. and Hippo cycling (not release) (NPG) “Weight riluzole in a scintillation tube or any glass container. For 20mg of riluzole, add 500ul of tween20 to the powder and vortex gently. Then Add 2.5 ml of saline and vortex again. Complete with 2ml saline. I usually put a stir bar in and live it stirring for a few minutes.” 2009 Riluzole hydrochloride Rat, ↑ p38, BDNF (through N-type Katoh- was i.p. injected at 1.5weeks voltage gated Ca+ channels) Semba doses of 19 mg/kg. old (M/F) single injection? (Tocris Cookson (Bristol, UK)) 2009 8-10 mg/kg ip, befor Rat, SCI ↓ spasticity (some but not all) Kitzman assessment, dissolved in 0.05% Etoh). 2009 Single ip same day as Rat, ↓ MK-801-induced hyperalgesia and Schmidt testing MK801 CSF EEAs induced hyperalgesi a 2009 Toklu 6 mg/kg, s.c., 30 min Rat sepsis ↓ sepsis-induced weight loss, body after the surgical temperature, brain edema, BBB procedure, and every permeability, oxidative damage, and 12 h as continuing brain injury. increasing the survival treatment (6h or 48h) rate, improved neurological examination scores 2010 21 days, 4 mg/kg per Rat, PFC, ↓ depressive symptoms, restore C- Banasr day, i.p. depression, acetatate and GFAP, ↑ GLT-1 (NPG) unpredicted stress

282

Riluzole: In vivo studies with central or behavioral effects Reference Dose Model Effect 2010 (21 days between day Rat, ↓ glutamate (microdialysis) Dzahini 3 and day 23) Riluzole 6OHDA (Sanofi Avantis, Antony, France, 5 mg/day, i.p) “last injectable dilution is done only in saline (NaCl 0.9%) a few minutes before treatment. Of course concentrated 10X riluzole is pre dissolved in 5% DMSO and stored at 4°C.” 2010 dissolved in saline Rat, spinal ↓ hypersensitivity, ↑ pCREB (results Hayashida ligation, suggest that activation of glutamate LC transporters (descencin in the LC results in increase of g extracellular glutamate signaling, projection possibly via facilitation of glutamate to sc) release from astrocytes) 2010, 2011 20, 40 and 80 µg/10 Rat, ↓ glutamate (HPLC) and apoptosis Hassanzade µl/rat) dissolved in 1% morphine- (40, 80 comp to morphine, but not h Tween 80 in sterile induced controls) 0.9% normal saline apoptosis and infused icv 8 days 2011 Hama 8 mg/kg single ip Rat, spinal Anti-nociceptive (2ml/kg) or icv (5 µl) cord injury (Sigma) dissolved in a vehicle of 30% (2- hydroxypropyl)-ß- cyclodextrin in saline

283

Riluzole: In vivo studies with central or behavioral effects Reference Dose Model Effect 2011 Single (1st day Rat, ↓ hyperemotional responses Takahashi administration) and olfactory (depression) and glutamate subchronic (7 day bulbectomy (microdialysis) administration) riluzol e treatment (1-10 mg/kg, po) (Sanofi- Aventis) Tablets 50 mg were crushed and suspended uniformly in 0.5% methyl cellulose (MC). The drug was administrated po in a volume of 5.0 ml/kg body weight Gourley 6-60µg/ml (~1.2 – Mice Transport. ↓ depressive symptoms, 2012 13.2 mg/kg/day) in restore hippocampal BDNF and ↑ drinking water, 21 GLT-1(60 µg/ml) days

Riluzole: mechanism of action Reference Dose Measurement Outcome 1991 Pre-synaptic. interacts with the voltage- Benoit & dependent Escande sodium channel, probably by stabilizing the inactivated conformation 1991 10 nmol icv, or ↓ MCD and DTX but not 4-AP seizures Sanger 4 mg/kg oral 1992 Pre-synaptic. Blocks glutamate release Chdramy 1992 interaction with G proteins Doble

284

Riluzole: mechanism of action Reference Dose Measurement Outcome 1992 n added to the Striatal dose-dependent decrease in the uptake of Sameul incubation synaptisomes 3H-dopamine, 3H-GABA and 3H- medium or glutamate into striatal synaptosomes. 3H- after in vivo dopamine uptake more sensitive to riluzole administration than glutamate. inhibited 3H-dopamine uptake competitively and 3H-glutamate uptake non-competitively. After in vivo injection, riluzole did not affect the striatal dopamine, DOPAC, serotonin, 5HIAA, glutamate, aspartate or GABA contents. compound was previously reported to induce a decrease in the spontaneous release of glutamate, serotonin, dopamine and possibly acetylcholine, the hypothesis is put forward that riluzole may, at least at high concentrations, have general effects on the striatal nerve terminals affecting both the uptake and release processes. 1993 Post-synaptic. Block ionic flux through Debono NMDA channels 1993 Pre-synaptic. ↓ potassium-elicited Glu Martin release. 1994 Rat, cultured Post-synaptic. Antagonism of Ca++ entry Hubert granule cells evoked by NMDA, glutamate. even in the presence of TTX to block Na+ channels or veratridine to activate them. Conclude blocks voltage-dependent Na+ channels and NMDARs, g-protein-dependent. 1994 Rat, striatal Blocks GABA uptake Mantz synaptisomes 1995 Rat, motor Post-synaptic. ↓ glutamate and NMDA Estevez neuron- toxicity, but not AMPA or kainate enriched culture 1995 Mary 4, 8 mg/kg/ os Rat, ↓ lesion, like MK801 (but less), and quinolinate lamotrigine (voltage-sensitive NA+ channel HD blocker) had no effect) :: quinolinate is NMDAR agonist 1995 1-100 µmol Culture, neuroprotection Rothstein Glutamate- mediated motor neuron toxicity

285

Riluzole: mechanism of action Reference Dose Measurement Outcome 1996 Doble 1996 Rat, ↓ Pre-synaptic conduction → accounting for Maclver hippocampal ↓ post-synaptic population spikes, slice, CA1 glutamate EPSPs. No effect on GABA inhibition. 1997 Rat, dorsal ↓ Pre-synaptic Ca+ influx by blocking N- Huang horn ganglion and P/Q-, but not L-type high voltage Ca+ channels 1997 Keita Synaptisomes, ↓ DA release induced by veratridine, NMDA, NMDA and kainite, but not KCl or nicotine Kainate and veratridine DA release 1997 Rat, cortical ↓ Pre-synaptic. ↓ action potentials elicited Siniscalchi neurons by current and ↑ threshold for generation of Ca+ spike. No effect on glutamate-elicited depolarization 1997 Cortical Pre-synaptic. ↓ Na++ current, and high- Stefani neurons and low-voltage Ca+-currents 1998 Striatal spiny Pre- and Post-synaptic. ↓ number of, Centonze neurons frequency and amplitude of spikes. ↓ exogenous glutamate-elicited EPSPs. 1998 Rat, cultured Pre-synaptic. Blocked depolarization- Hubert motor neurons evoked Ca+ transients

286

Riluzole: mechanism of action Reference Dose Measurement Outcome 1998 2, 4 mg/kg. Rats, various Not at NMDARs. Riluzole influenced Kretschme Dissolved in glutamatergic neither stereotyped sniffing behavior nor r cremophor drugs locomotion but impaired motor 10%, single ip coordination and attenuated rigidity induced before by blockade of dopamine D1 and D2 behavioral receptor antagonists when given alone. At testing higher doses spontaneous behavioral activity decreased and motor coordination was more impaired. Augmentation of the riluzole effects were observed when NMDA, but not GYKI 52466, was coadministered. The glycine site agonist DCS increased the anticataleptic properties of riluzole. The results indicate that when given alone, riluzole has a behavioral profile resembling that of competitive NMDA receptor antagonists. However, coadministration of Riluzole with NMDA/AMPA receptor ligands suggests that this assumption is incorrect, and that riluzole affects glutamatergic transmission by a more indirect mechanism 1999 Rat, cortical Pre-synaptic. Blocked Veratridine- (Na+ Lingamane synapticsomes and Ca+ dep) but not KCl- (Ca+ dep) ni evoked glutamate release. Conclude that it inhibits synaptic glutamate by blocking presynaptic Na+ channels 2000 3.5 or 48hr Rat ↓ hyperalgesia and Riluzole decreased the Abarca before testing. hyperalgesia concentrations of glutamate and aspartate Drugs were and those of citrulline and arginine in the dissolved in presence or absence of painful stimulation saline _except in the ventral posterolateral thalamic nuclei riluzole, which was dissolved in 10% Tween 80. and administered s.c. in a volume of 1 ml/kg 2000 8 mg/kg 4hr Rat, spinal ↑ glutamate uptake. “significantly Azbill prior cord increased in spinal cord synaptosomes synaptosomes obtained from rats treated with 8 mg/kg (i.p.) of riluzole and sacrificed 4 h later”

287

Riluzole: mechanism of action Reference Dose Measurement Outcome 2000 5, 10, 20 Mice, Did not prevent convulsions of lethality. Brackett mg/kg ip, pre- cocaine- Pre-treatment with competitive or treatment. induced NMDA/glycine site antagonists dose- dissolved in convulsions dependently attenuated cocaine-induced 50% DMSO convulsions and lethality (P,0.05). Pre- and 50% treatment with channel blockers or distilled water. allosteric modulators of the NMDA receptor protected against cocaine-induced convulsions (P,0.05), but were ineffective or less effective than the competitive and glycine site antagonists in preventing death. The glutamate release inhibitor riluzole failed to prevent both the convulsions and lethality induced by cocaine. 2000 2.5-20 mg/kg Mice, cocaine Did not effect expression nor induction of Itzhak pre-treatment and meth- sensitization to cocaine, (20 mg/kg) blocked induced the acute response to METH on day 1 and behavioral the expression sensitization of the sensitized response on day 5 but not the induction of sensitization to METH 2000 Jehle Mouse (Rat Pre-synaptic. ↓ [3H]Glu release up to 77%. and human), ↓ Ach, DA and 5-HT but not NA release neocortical (that’s a bad profile if the concs are slices and therapeutically relevant). Article draws caudate some weird conclusions. 2000 Mu 8 mg/kg pre- Rat TSCI ↑ mitochondrial function and enhance treatment, glutamate and glucose uptake dissolved in a small volume of HCl (0.1 N), diluted in saline at a concentration of 2.5 mg/ ml, and the pH adjusted to 7.4 with NaOH. 2000 Pre-synaptic. ↓ glu release by Na channel Prakriya block

288

Riluzole: mechanism of action Reference Dose Measurement Outcome 2001 Hippo slice, Riluzole produced an extremely potent Peyclit NMDA- concentration-related inhibition of NMDA induced (1 tyrosine mM)-stimulated protein tyrosine phosphorylati phosphorylation (IC 50.560.03 mM, on mean6S.D.), but failed to affect that evoked by phorbol 50 12-myristate 13-acetate (PMA, an activator of protein kinase C, 0.1 and 1 mM). 2001 Westerlaak 2001 Xu bovine Pre-synaptic. ↓ pre-synaptic glu release in adrenal zona part by slowing Kv1.4 K+ inactivation fasciculate cells 2002 Pre-synaptic. Prevents NMDAR Farber - hypofunction toxicity (counterintuitive, but NPG makes sense) 2002 Farber 2002 He Cultured ↑ post-synaptic GABA (at higher concs – hippocampal how high?) neurons 2002 Izumi Rat, retina Riluzole protected against TBOA-induced neuronal damage (but not PDC or THA). Their interpretation is because Riluzole ↓ glutamate release, but it could also be because Riluzole ↑ transport (unless TBOA was saturating at the elevated transport level) 2002 Tutka 2002 Zona 100µM Rat cortical ↓kainate-induced inward currents measured neuron culture by patch-clamp with evidence for non- competitive mechanism of inhibition by reducing open-state of ionic channels 2003 Pre-incubated Rat ALS ↑ glutamate transport in controls and early Dunlop with SOD spinal disease, but not late stage disease which had synaptisomes, cord fewer transporters [not the same as ↑ not transporters; be careful about interpretation] administered in vivo 10 min 2003 Izumi

289

Riluzole: mechanism of action Reference Dose Measurement Outcome 2004 Rat, astrocyte ↑+ ↓ glutamate transport dep on dose. low Frizzo culture concentrations of glutamate (1 and 10 ¹M) riluzole significantly increased glutamate uptake, whereas from 100 ¹M promoted a slight reduction 2004 Rat, Pre-synaptic. ↓ EPSP Meeks hippocampal slice 2004 culture ↓ glutamate accumulation and neuronal Vidwans injury 2004 Synaptosome Pre-synaptic. ↓ glutamate release. Ca+ Wang channel. 2005 2mM through Rat, SCI Three sodium channel blockers, riluzole, McAdoo fiber from mexiletine and QX-314, were administered impact injury to determine whether agents to end of that block firing and conduction of action experiment potentials also (~30 min), reduce the release of glutamate following Tocris SCI. No detectable effect on glutamate release. 2006 Camacho 2006 Wu Striatal protect HD neurons from glutamate- neurons induced cell death. We found that folic acid, gabapentin and lamotrigine did not , but memantine and riluzole were protective. 2007 Kim 4 mg/kg ip Rat, drug- ↓ VGlut1, Riluzole treatment completely after baseline induced inhibited pre-ictal recording seizure spikes and spike-wave discharges 2008 Clonal cells, ↑ glutamate transporter activity. ↑ activity Fumagalli rat of GLT1, GLAST and EAAC1. transporters 2008 Rat neonatal Pre-synaptic. Blocks Na+ (fully) and Ca+ Lamanausk hypoglassal (partially) currents and modulates release as motoneuron via presynaptic NMDARs and PKC slice prep inhibition. Induced repetitive firing was inhibited without changing single action potentials. Informative intro. This result shows that riluzole was not indiscriminately depressing synaptic transmission, but it targeted synapses, which were particularly active.

290

Riluzole: mechanism of action Reference Dose Measurement Outcome 2008 1 µM Hippocampal Does not change NMDAR activation during Rammes “therapeuticall slice neurotransmission y relevant dose” 2009 Rat, cortical ↓ glutamate-induced slowing of Stevenson neuron culture neurofilament transport 2011 Cifra 5 µM 15 min Rat, ↓ TBOA-induced bursting, ↓ the frequency after TBOA motoneuron of spontaneous glutamatergic events, culture reversed ↑ in S100B and prevented late loss of motoneuron staining 2011 Liu 1µM Neuroprogenit ↑ GLT1 (reporter and protein expression) or cell culture and HSF1, and ↓ NMDAR-mediated neuronal death 2011 Induces GDNF Tsuchioka 2012 Mouse striatal ↑ GLT1. Growth factor G5 →↑ GLT1 and Carbone astrocyte removal ↓ GLT1. Riluzole (non-Na+ culture cnannel-dependant) and dexamethasone maintained GLT1, ceftriaxone and CDP-choline did not.

291