Progressive Pathogenesis of Juvenile

Neuronal Ceroid Lipofuscinosis

A thesis submitted for the degree of Doctor of Philosophy in the University of London

Charlie Pontikis

Department of Neuroscience Institute of Psychiatry, King’s College London Denmark Hill London SE5 8AF

1 Statement of Originality

All histological analyses and assessment of progressive neuropathological changes present in Cln3-/- and homozygous Cln3∆ex7/8 mice were carried out by Charlie Pontikis at the Institute of Psychiatry, London, England

All live animal procedures including generation of Cln3-/- mice were carried out at the University of Rochester School of Medicine, NY, USA.

All live animal procedures including generation of homozygous Cln3∆ex7/8 mice were carried out at Massachusetts General Hospital, MO, USA

2 Abstract

Juvenile neuronal ceroid lipofuscinosis (JNCL) belongs to a group of fatal neurodegenerative disorders of childhood. This autosomal recessive disorder is caused by mutations in the CLN3 gene, resulting in the production of an abnormal truncated CLN3 protein that is presumed to be dysfunctional. The recent development of two distinct mouse models of JNCL (Cln3 null mutant [Cln3-/-] and Cln3 ‘knock-in’ [Cln3Δex7/8]) has made it possible to map the progressive development of neuropathological changes associated with JNCL. The studies described in this thesis investigate the sequence of pathological changes during disease progression, and have begun defining the nature and timing of pathological effects upon the CNS in Cln3-/- and homozygous Cln3Δex7/8 mice. In order to investigate these events, unbiased stereological methodology was used to characterize the extent of neuronal loss and glial activation during prenatal and early postnatal development in both mouse models of JNCL. There was a pronounced neurodegenerative phenotype at terminal stages of the disease, which was more subtle during the early postnatal period of development. Neuropathological changes included the widespread deposition of autofluorescent storage material, regional atrophy, loss of selected neuron populations that was preceded by glial activation. The thalamocortical system was consistently targeted by the disease, exhibiting an early loss of thalamic neurons and localised astrocytosis, before effects on target neocortical neuron populations. Taken together with studies in human NCL, our data suggest that highly regionalized interactions occur between neurons and glia at different stages of disease progression. Whether these events contribute directly to subsequent neurodegeneration has not been demonstrated, but the suggestion that neuron–glial interactions differ markedly between CNS regions may provide important clues to the pathogenesis of these disorders.

3 Publications Arising from this Thesis

Pontikis CC, Cella CV, Parihar N, Lim MJ, Chakrabarti S, Mitchison HM, Mobley WC, Rezaie P, Pearce DA, Cooper JD (2004). Late onset neurodegeneration in the Cln3-/- mouse model of juvenile neuronal ceroid lipofuscinosis is preceded by low level glial activation. Brain Res. 1023, 231-242

Pontikis CC, Cotman SL, Macdonald ME, Cooper JD (2005). Thalamocortical neuron loss and localized astrocytosis in the Cln3 (Deltaex7/8) knock-in mouse model of Batten disease. Neurobiol Dis. [Epub ahead of print]

4 Acknowledgements

I am very grateful to my supervisor Dr. Jonathan Cooper for all his valuable advice and support throughout the course of this PhD and his constructive comments during the writing of this thesis. I would also like to thank my second supervisor Dr. Brenda Williams and colleague Dr. Payam Rezaie for all of their help when I strayed into their areas of expertise.

A big thank you to my esteemed colleagues Dr. Ellen Bible and Dr. Ming Lim, who have provided intellectual input as well as stimulating conversation in both the work capacity and on numerous social occasions. I am also thankful for all of the technical assistance provided by Steve Shemilt and Noreen Alexander and their efficient running of the lab, which has enabled data collection to run efficiently.

In addition, I would like to express my gratitude to the soon to be appointed Dr. Jill Weimer AKA ‘The twin’ for continuous collaborative efforts and invaluable help towards the end of this PhD. Thank you also to her supervisor, Dr. David Pearce, for supplying the Cln3-/- mouse tissue used in this thesis and for welcoming me into his lab at the University of Rochester; and to Drs. Sue Cotman and Marcy MacDonald of Harvard Medical School for providing the Cln3∆ex7/8 mouse tissue used in these studies.

Lastly I would like to thank my family for their encouragement and confidence in me, and to all of my friends including my special girlfriend Shama for ample support throughout this PhD.

5 Contents

Statement of Originality...... 2

Abstract...... 3

Publications Arising from this Thesis ...... 4

Acknowledgements ...... 5

Contents ...... 6

List of Figures...... 14

List of Tables ...... 18

Abbreviations ...... 20

Chapter One: Introduction...... 24

1.1 Overview...... 25

1.2 The lysosome ...... 26 1.2.1 Lysosomal membrane ...... 26 1.2.2 Transport of proteins into the lysosome...... 28 a) Lysosomal enzyme processing ...... 28 b) Transport of membrane associated proteins to the lysosome ....31 c) Lysosomal dysfunction...... 32 d) Incorrect processing of lysosome/Mutations within lysosomal enzymes ...... 32

1.3 Lysosomal Storage Disorders...... 33 1.3.1 Enzyme related lysosomal dysfunction...... 33

6 a) Gangliosidoses...... 33 b) Mucopolysaccharidoses...... 35 c) Glycoproteinoses ...... 36 d) Mucolipidoses...... 36 1.3.2 Non-enzyme lysosomal dysfunction ...... 37

1.4 The Neuronal Ceroid Lipofuscinoses: Enzyme and non-enzyme related lysosomal dysfunction...... 38 1.4.1 Storage material...... 39

1.5 NCL proteins ...... 41 a) Lysosomal enzymes ...... 41 b) Novel proteins of unknown function...... 45 c) Rare forms of NCL...... 48

1.6 Juvenile NCL, the most common form of NCL...... 49 i) Structure of CLN3...... 49 ii) Intracellular localisation of CLN3...... 52 iii) Proposed function for CLN3...... 53 iv) Mutations in CLN3...... 54 v) Clinical progression of JNCL...... 54 vi) Potential Interactions of CLN Proteins ...... 55

1.7 Animal models of NCLs ...... 56

1.8 Mouse models of JNCL ...... 61 a) Cln3 null mutant mice (Cln3-/-)...... 61 b) Cln3 ‘knock-in’ mice (Cln3∆ex7/8)...... 61

1.9 Neuronal loss in NCLs...... 62

1.10 Mechanisms of cell death in NCLs ...... 63

1.11 Glial responses in NCLs ...... 64

7 a) The role of microglia in the NCLs ...... 65 b) The role of astrocytes in the NCLs...... 66

1.12 Studies in this thesis: Aims and objectives ...... 68

1.13 Summary of background...... 69

1.14 Hypothesis tested...... 69

Chapter Two: Materials and Methods...... 71

2.1 Introduction...... 72

2.2 Stereology ...... 72 a) Design-based stereology...... 72 b) Stereological estimation of volume...... 73 c) Stereological estimation of cell number...... 74 d) Stereological estimation of mean cellular volume ...... 76 e) Use of stereology in the JNCL CNS ...... 77

2.3 Mouse models of JNCL ...... 77 a) Cln3 null mutant mouse model of JNCL ...... 77 i) Summary of targeting strategy...... 77 ii) Animals in study ...... 78 b) Cln3∆ex7/8 ‘knock-in’ mouse model of JNCL...... 78 i) Summary of targeting strategy...... 78 ii) Animals in study ...... 79

2.4 Histological processing...... 79

2.5 Nissl staining...... 79

2.6 Immunohistochemistry...... 80 a) Immunohistochemistry for interneuron markers...... 81

8 b) Immunohistochemistry for glial markers ...... 82

2.7 Measurements of volume, neocortical and laminar thickness ...... 83

2.8 Measurements of total neuronal number and volume...... 86

2.9 Measurements of interneuron number ...... 88

2.10 Assessment of glial phenotype...... 90 a) Quantitative analysis of glial phenotype ...... 90 b) Measurements of astrocytic number ...... 91

2.11 Statistical analysis...... 91 a) One-way ANOVA...... 91 b) Co-efficient of error...... 92

Chapter Three: Progressive Neuropathological Changes in the Cln3-/- Mouse Model of JNCL...... 93

3.1 Introduction...... 94

3.2 Autofluorescent storage material ...... 94

3.3 Evaluation of regional atrophy in Cln3-/- mice...... 96

3.4 Regional effects upon neocortical thinning and lamination in Cln3-/- mice ...96

3.5 Regional and laminar effects upon neuronal number in Cln3-/- mice ...... 101

3.6 Regional and laminar effects upon interneuronal number in Cln3-/- mice....105

3.7 Astrocytic and microglial responses in Cln3-/- mice ...... 105

3.8 Discussion ...... 113

9 Chapter Four: Neuropathological Changes in the Severely Affected Cln3∆ex7/8 Knock-in Mouse Model of JNCL ...... 118

4.1 Introduction...... 119

4.2 Accumulation of autofluorescent storage material in homozygous Cln3∆ex7/8 mice...... 119

4.3 Assessment of regional atrophy in homozygous Cln3∆ex7/8 mice...... 121

4.4 Regional effects upon neocortical thinning and lamination in homozygous Cln3∆ex7/8 mice...... 121

4.5 Regional and laminar effects upon neuronal number in homozygous Cln3∆ex7/8 mice...... 125

4.6 Survival of GABAergic interneurons in homozygous Cln3∆ex7/8 mice ...... 127

4.7 Regionally restricted astrocytic and microglial activation in homozygous Cln3∆ex7/8 mice...... 129

4.8 Discussion ...... 133

Chapter Five: Early Neuropathological Changes in the Cln3-/- Mouse Model of JNCL ...... 138

5.1 Introduction...... 139

5.2 Evaluation of regional volume in P8 Cln3-/- mice...... 139

5.3 Regional effects upon neocortical thinning and lamination in P8 Cln3-/- mice ...... 141

5.4 Regional and lamina effects upon neuronal number in P8 Cln3-/- mice ...... 144

10 5.5 Astrocytic and microglial responses in P8 Cln3-/- mice...... 146

5.6 Discussion ...... 153

Chapter Six: Neuropathological Changes in Early Postnatal Homozygous Cln3∆ex7/8 Knock-in Mice...... 157

6.1 Introduction...... 158

6.2 Assessment of regional atrophy in P7 homozygous Cln3∆ex7/8 mice ...... 158

6.3 Regional effects upon neocortical thinning and lamination in P7 homozygous Cln3∆ex7/8 mice...... 160

6.4 Regional and lamina effects upon neuronal number in P7 homozygous Cln3∆ex7/8 mice...... 163

6.5 Early astrocytic and microglial responses in P7 homozygous Cln3∆ex7/8 mice...... 163

6.6 Discussion ...... 173

Chapter Seven: General Discussion ...... 177

7.1 Introduction...... 178

7.2 Mouse models of JNCL ...... 179

7.3 Importance of strain backgrounds...... 181

7.4 Accumulation of autofluorescent storage material in mouse models of JNCL ...... 182

7.5 Regionally specific changes in mouse models of JNCL...... 183

11 a) Regionally specific changes in the Cln3-/- mouse model of JNCL ...183 b) Regionally specific changes in the early postnatal Cln3-/- mouse model of JNCL...... 184 c) Regionally specific atrophy in the homozygous Cln3∆ex7/8 mouse model of JNCL...... 185 d) Regionally specific changes in the early postnatal homozygous Cln3∆ex7/8 mouse model of JNCL ...... 186

7.6 Lamina specific changes in mouse models of JNCL ...... 186

7.7 Selective loss of neocortical neurons in mouse models of JNCL...... 187 a) Selective loss of neocortical neurons in the Cln3−/− mouse model of JNCL...... 187 b) Selective loss of neocortical neurons in the homozygous Cln3∆ex7/8 mouse model of JNCL ...... 188

7.8 Selective vulnerability of interneurons in mouse models of JNCL ...... 190

7.9 Loss of thalamic neurons in mouse models of JNCL ...... 192 a) Loss of thalamic neurons in the Cln3−/− mouse model of JNCL ...... 193 b) Loss of thalamic neurons in the homozygous Cln3∆ex7/8 mouse model of JNCL...... 195

7.10 Glial responses in mouse models of JNCL ...... 197 7.10.1 Astrocytic responses in mouse models of JNCL...... 198 a) Astrocytic responses in the Cln3-/- mouse model of JNCL...... 198 b) Astrocytic responses in the homozygous Cln3∆ex7/8 mouse model of JNCL...... 200 7.10.2 Microglial responses in mouse models of JNCL ...... 202 a) Microglial responses in the Cln3-/- mouse model of JNCL .....203 b) Microglial responses in the homozygous Cln3∆ex7/8 mouse model of JNCL...... 205

7.11 Conclusions ...... 207

12 References...... 209

13 List of Figures

Figure 1.1 The transport of newly synthesized lysosomal hydrolases to lysosomes ...... 29

Figure 1.2 The ultrastructural appearances of the abnormal intraneuronal in different forms of the neuronal ceroid lipofuscinoses (NCLs) ...... 41

Figure 1.3 Predicted topologies of the CLN3 protein...... 50

Figure 2.1 Unbiased counting frame within a virtual space of a Nissl stained tissue section ...... 75

Figure 2.2 Representative image of the Nucleator estimation of neuronal size within an unbiased counting of a Nissl stained tissue section ...... 76

Figure 3.1 Accumulation of autofluorescent storage material in Cln3-/- mice .....95

Figure 3.2 Unbiased Cavalieri estimates of regional volume in homozygous Cln3-/- mice ...... 97

Figure 3.3 Neocortical thickness measurements in Cln3-/- mice vs. littermate controls at 18, 14 and 5 months of age ...... 98

Figure 3.4 Laminar-specific changes in neocortical thickness in Cln3-/- mice.....99

Figure 3.5 Unbiased optical fractionator estimates of neuronal number in neocortical laminae and thalamic nuclei in aged Cln3-/- mice...... 102

Figure 3.6 Unbiased optical fractionator estimates of neuronal number in neocortical laminae and thalamic nuclei in Cln3-/- mice...... 103

Figure 3.7 Interneuron counts in Cln3-/- mice...... 106

14

Figure 3.8 Astrocytic responses in Cln3-/- mice ...... 108

Figure 3.9 Immunohistochemical analysis of astrocytic responses in Cln3-/- mice...... 109

Figure 3.10 Microglial responses in Cln3-/- mice ...... 111

Figure 3.11 Immunohistochemical analysis of microglial responses in Cln3-/- mice...... 112

Figure 4.1 Accumulation of autofluorescent storage material in homozygous Cln3∆ex7/8 mice...... 120

Figure 4.2 Unbiased Cavalieri estimates of regional volume in homozygous Cln3∆ex7/8 mice...... 122

Figure 4.3 Subregion- and laminar-specific changes in neocortical thickness in homozygous Cln3∆ex7/8 mice ...... 123

Figure 4.4 Unbiased optical fractionator estimates of neuronal number in homozygous Cln3∆ex7/8 mice ...... 126

Figure 4.5 Interneuron counts in homozygous Cln3∆ex7/8 mice...... 128

Figure 4.6 Astrocytic responses in homozygous Cln3∆ex7/8 mice ...... 130

Figure 4.7 Quantitative assessment of regional astrocytosis in homozygous Cln3∆ex7/8 mice at 12 months of age ...... 132

Figure 4.8 Microglial responses in homozygous Cln3∆ex7/8 mice...... 134

Figure 5.1 Unbiased Cavalieri estimates of regional volume in Cln3-/- mice.....140

15 Figure 5.2 Subregion- and laminar-specific changes in neocortical thickness in Cln3-/- mice ...... 142

Figure 5.3 Unbiased optical fractionator estimates of neuronal number in Cln3-/- mice...... 145

Figure 5.4 Astrocytic responses in Cln3-/- mice ...... 147

Figure 5.5 Quantitative assessment of regional astrocytosis in P8 Cln3-/- mice.149

Figure 5.6 Microglial responses in Cln3-/- mice ...... 151

Figure 5.7 Quantitative assessment of microglial responses in Cln3-/- mice ...... 152

Figure 6.1 Unbiased Cavalieri estimates of regional volume in homozygous Cln3∆ex7/8 mice...... 159

Figure 6.2 Subregion and laminar-specific changes in neocortical thickness in homozygous Cln3∆ex7/8 mice ...... 161

Figure 6.3 Unbiased optical fractionator estimates of neuronal number in homozygous Cln3∆ex7/8 mice ...... 164

Figure 6.4 Astrocytic responses in homozygous Cln3∆ex7/8 mice ...... 166

Figure 6.5 Quantitative assessment of regional astrocytosis in homozygous Cln3∆ex7/8 mice at 7 days of age...... 168

Figure 6.6 Microglial responses in homozygous Cln3∆ex7/8 mice...... 170

Figure 6.7 Immunohistochemical analysis of microglial responses in P7 homozygous Cln3∆ex7/8 mice ...... 172

16 Figure 7.1 Diagrammatic representation of neuronal loss in the thalamocortical system ...... 194

17 List of Tables

Table 1.1 Mouse models of NCL ...... 57

Table 2.1 Details of Cln3-/- mice, homozygous Cln3Δex7/8 mice and age-matched control littermates ...... 79

Table 2.2 Cavalieri based grid areas for estimating regional volume in Cln3-/- mice and age-matched control littermates ...... 84

Table 2.3 Cavalieri based grid areas for estimating regional volume in homozygous Cln3Δex7/8 mice and age-matched control littermates.....85

Table 2.4 Optical fractionator and nucleator based sampling schemes for different neuron populations in Cln3-/- mice and age-matched control littermates ...... 87

Table 2.5 Optical fractionator and nucleator based sampling schemes for different neuron populations in homozygous Cln3Δex7/8 mice and age-matched control littermates ...... 88

Table 2.6 Selected sampling scheme for interneurons in Cln3-/- mice, homozygous Cln3Δex7/8 mice and age-matched control littermates.....89 Table 2.7 Selected sampling scheme for chosen structures in homozygous Cln3Δex7/8 mice and age-matched control littermates ...... 91

Table 3.1 Tabular depiction of significant changes in individual laminar thickness in Cln3-/- mice...... 100

Table 4.1 Tabular depiction of significant changes in individual laminar thickness in homozygous Cln3∆ex7/8 mice...... 124

18 Table 5.1 Tabular depiction of significant changes in individual laminar thickness in Cln3-/- mice...... 143

Table 6.1 Tabular depiction of significant changes in individual laminar thickness in homozygous Cln3∆ex7/8 mice ...... 162

19 Abbreviations

l 3 0,i Cube of the ith point-sampled intercept length +/+ Control µm Micrometer a/p Area associated with each point ABC Avidin biotin complex AFI Amaurotic familial idiocy AGT Angiotensinogen ANCL Adult neuronal ceroid lipofuscinoses Ang Angiotensin ANOVA Analysis of variance ANP Atriopeptin AQP-4 Aquaporin-4 asf Area sampling fraction ATP Adenosine-tri-phosphate AVP Vasopressin BBB Blood brain barrier BLBP Brain lipid-binding protein bp Base pair Ca2+ Calcium ion cAMP Cyclic adenosine monophosphate CE Co-efficient of error cGMP Cyclic guanosine monophosphate CL Centrolateral thalamic nucleus Cl- Chloride CLN Ceroid lipofuscinosis CNS Central nervous system CV Curvilinear profile DAB Diaminobenzidine DG Dentate gyrus DNA Deoxyribonucleic acid E Embryonic

20 EMPR Epilepsy with mental retardation ER Endoplasmic reticulum FP Fingerprint profile FvLINCL Finnish variant late infantile neuronal ceroid lipofuscinosis g Grams GABA γ-amino-butyric acid GAD Glutamic acid decarboxylase GFAP Glial fibrillary acidic protein GFP Green fluorescent protein GLAST1 Sodium-coupled glutamate transporter 1 GLT1 Glutamate transporter 1 GROD Granular osmiophilic deposit Hex Hexaminidase hsf Height sampling fraction IG/ML Inner granular/molecular layer IgG Immunoglobulin G IL6 Interleukin 6 INCL Infantile neuronal ceroid lipofuscinosis iNOS Inducible nitric oxide synthase JNCL Juvenile neuronal ceroid lipofuscinosis kb Kilobase kDa Kilodalton kg Kilogram l Length of intercepts LAM Lamina LAMP Lysosomal associated membrane protein LD Laterodorsal thalamic nucleus Lent Lateral entorhinal cortex LGNd Dorsal lateral geniculate nucleus LIMP Lysosomal integral membrane protein LINCL Late infantile neuronal ceroid lipofuscinosis M Molar m Section periodicity

21 M1 Primary motor cortex M6P Mannose-6-phosphate MD Mediodorsal thalamic nucleus mg Milligram mM Millimolar Mnd Motor neuron disease MPS Mucopolysaccharidoses MRI Magnetic resonance imaging mRNA Messenger ribonucleic acid n Sample number/Total number of point sampled linear intercepts NA Numerical aperture NCL Neuronal ceroid lipofuscinosis Nclf Neuronal ceroid lipofuscinosis NE Northern Epilepsy NGS Normal goat serum NIH National Institute of Health NO Nitric oxide Nv Unbiased virtual counting spaces P Postnatal PGK Phosphoglycerate kinase pH Potential of hydrogen Pi Number of points counted in each section Po Posterior thalamic nucleus PPT1 Palmitoyl protein thioesterase 1 PV Parvalbumin Q Mean of the total number of cells counted in any of the counting frames within any section RC1 Radial cell 1 RER Rough endoplasmic reticulum RL Rectilinear complex S1BF Primary somatosensory barrel field cortex SOM Somatostatin ssf Sampling fraction

22 t Mean section thickness TBS Tris buffered saline TPP-I Tripeptidyl peptidase I TUNEL Terminal dUDP nick end-labelling V1 Primary visual cortex vLINCL Variant late infantile neuronal ceroid lipofuscinosis VM Ventromedial thalamic nucleus VPM/VPL Ventral posterior thalamic nucleus WM White matter

23

CHAPTER 1

Introduction

24 1.1 Overview

The neuronal ceroid lipofuscinoses (NCL) are a fatal group of inherited neurodegenerative disorders, with an age of onset ranging from childhood to early adulthood (Goebel, 2004). These disorders are characterized by widespread intralysosomal accumulation of autofluorescent material, blindness, marked psychomotor deterioration and uncontrollable seizures and with no effective treatment available ultimately result in premature death (Hofmann and Peltonen, 2001; Cooper, 2003). There are at least seven distinct forms of NCL, of which Juvenile NCL (JNCL) is the most prevalent (Haltia, 2003). JNCL is caused by mutations in the CLN3 gene, resulting in the production of an abnormal truncated CLN3 protein that is presumed to be dysfunctional (International Batten Disease Consortium, 1995). Under normal circumstances, the CLN3 protein is expressed in both neuronal as well as somatic cells; however reports of pathology outside the CNS are rare in JNCL (Østergaard et al., 2005) and the main clinical features of the disease are characteristic of neurological impairment. As such, it seems likely that the CLN3 protein may have an important role within the CNS. However, this precise role and the consequences of CLN3 mutation upon the CNS remain poorly understood. In order to map the sequence of neuropathological changes associated with JNCL, the studies described in this thesis have looked at progressive neuronal and glial phenotypes in two distinct mouse models of JNCL.

The NCLs are just one of several lysosomal storage disorders, which are typically caused by the defective activity of lysosomal proteins and result in the intra-lysosomal accumulation of undegraded metabolites (Neufeld et al., 1975; Futerman and van Meer, 2004). While pathological mechanisms that operate in the NCLs remain unknown, valuable information can be obtained by studying how the normal biology of the lysosome is affected in other storage disorders.

25 1.2 The lysosome

The lysosome is a membrane-bound organelle containing multiple hydrolytic enzymes that work together within an acidic environment to digest biological macromolecules. Materials taken into the lysosome via endocytosis from the cytosol, and internal components undergoing autophagy, are broken down into their constituent parts before they are transferred across the membrane and back into the cytosol for further degradation (Sabatini et al., 2001). Alternatively, these degradation products may be synthesised into other macromolecules (Sabatini et al., 2001). The lysosomal membrane is impermeable to undigested or even partially digested material. However, lysosomal digestive products of low molecular weight can pass through via passive diffusion, or metabolite transporters (Lloyd, 1996). Several hundred lysosomes reside within all cell populations, with the exception of mature erythrocytes. These compartments are heterogeneous, differing morphologically, depending on the material that is being digested and the stage that the degradatory process has reached.

1.2.1 Lysosomal membrane

The lysosomal membrane is essential for the selective passage of various macromolecules between the lysosomal lumen and the cytosol (Winchester et al., 2000). As such, there is a variety of carriers and transporters within the structure of the membrane (Winchester et al., 2000). These components are able to regulate or influence both luminal and cytosolic environments, whilst proton pumps create an acidic intralysosomal pH, which is required for degradation (Mellman et al., 1986). The lysosomal membrane ultimately prevents the efflux of material, which has entered the lysosome via endocytosis or autophagy, and only permits degraded products to pass through to the cytosol. Several components associated with the lysosomal membrane have been identified, such as carrier mediated transport systems, which are specific to cystine (Gahl et al., 1982a, Gahl et al., 1982b) cationic amino acids (Pisoni et al., 1985; Pisoni et al., 1987a) and neutral amino acids of varying size (Pisoni et al., 1987b, Bernar et al., 1986). These are functionally diverse systems and

26 highlight the heterogeneous roles of the lysosome in addition to degradation of proteins.

In order to maintain an optimal acidic pH of 5 within the lysosomal lumen, membrane bound proton pumps, otherwise known as hydrogen ion ATPases, transport protons against their concentration gradient into the cytosol (Mellman et al., 1986). This process functions electrogenically (Harikumar et al., 1983) and requires cytoplasmic ATP, in addition to chloride ions, which are plentiful within the luminal space (Sabatini et al., 2001). Lysosomal proton pumps closely resemble vacuolar ATPases within the endoplasmic reticulum (ER, Rees-Jones et al., 1984), Golgi apparatus (Glickman et al., 1983) and endocytic vesicles (Mellman et al., 1986), due to their sensitivity to selective inhibitors (Okhuma et al., 1982; Harikumar et al., 1983), and their constant acidification of the lumen.

It is important for proton pumps, enzymes and other functional machinery within the lysosome to be protected from lysosomal hydrolases, since these enzymes are highly degradative. Complex oligosaccharides, which are characteristic of the luminal surface of the lysosomal membrane, are thought to act as protective residues from such enzymes, due to their rich poly-N-acetyllactosamines bearing sailic acids (Reijngoud et al., 1977). These acids may also help to establish an acidic pH within the lysosome (Reijngoud et al., 1977). Although details of the cytoplasmic surface of the lysosomal membrane remain unresolved, it is likely that surface receptors may be present, which aid in the fusion of lysosomes with endosomes, in order to help facilitate the transfer of lysosomal hydrolases (Sabatini et al., 2001). An understanding of lysosomal proteins and their normal function is particularly important, as the CLN3 protein is itself expressed within the lysosomal membrane (Mitchison et al., 1997; Golabek et al., 1999; Haskell et al., 2000; Mao et al., 2003a; Kyttala et al., 2004).

27 1.2.2 Transport of proteins into the lysosome a) Lysosomal enzyme processing

Lysosomal enzymes are synthesised in the rough ER (RER), where they are directed from the ribosomes to the reticular lumena. These enzymes then enter the mannose 6- phosphate (M6P) pathway, where a series of post-translational modifications occur, including glycosylation, phosphorylation and structural changes within carbohydrate and protein moieties (Sly et al., 1982). During this period they also acquire M6P, which is a specific lysosomal recognition marker for its receptor within the Golgi compartment. This assimilation is catalysed by N-acetyleglucosamine and N- acetylglucosamiyl phosphodiesterase, which are different enzymes associated with Golgi membrane fractions of higher or intermediate buoyant density (Pohlmann et al., 1982).

There are two distinct M6P receptors, which have been isolated and characterised. Both are integral membrane glycoproteins, but have very different molecular weights, calculated as 215 kDa (Sahagian et al., 1981) and 46 kDa respectively (Hoflack and Kornfeld, 1985). Their binding specificities to various types of phosphorylated oligosaccharides are similar, but not identical, with the larger receptor (M6P1) binding independently of divalent cations, and the smaller (M6P2) requiring their presence for enhanced binding of substrate (Hoflack and Kornfeld, 1985).

The M6P pathway is essential for trafficking of lysosomal enzymes from the rough ER into the lysosome and distinguishing them from secretory proteins (Figure 1.1). Initially, precursors to lysosomal enzymes residing within the RER are transported to the cis Golgi, where mannose residues are phosphorylated. The cis Golgi then develops into the trans Golgi reticulum where phosphorylated enzymes bind to the M6P receptor, drawing them into pits containing a fibrous protein called clathrin (Griffiths et al., 1988). A clathrin-coated vesicle is then generated within the membrane, which separates before clathrin is degraded into its sub-units. The uncoated vesicle merges to the late endosome, which has a low pH resulting in dissociation of the phosphorylated enzyme from the M6P receptor, with subsequent

28 29

Figure 1.1 The transport of newly synthesized lysosomal hydrolases to lysosomes. (Adapted from Alberts et al., 1994)

loss of the phosphate group and recycling of the mannose 6-phosphate receptor to the plasma membrane (Mellman et al., 1986). A transport vesicle containing the enzyme then buds off from the late endosome and merges with the lysosome, where the enzyme is released. Separation of lysosomal enzymes from secretory proteins occurs within the trans Golgi reticulum, where different vesicles are formed containing each class of protein. The secretory proteins are then sent to the plasma membrane and expelled via exocytosis. M6P receptors are also found on the cell surface where they bind phosphorylated secreted lysosomal enzyme. Clathrin-coated vesicles are then formed around this complex, which are internalised before the clathrin is degraded, an early endosome is formed and fusion to the late endosome occurs (Griffiths et al., 1988; Mellman et al., 1986).

Once the lysosomal enzymes are internalised within the lysosomal compartment they undergo a process of maturation. This includes proteolysis, folding and aggregation, with the possibility of dephosphorylation. Proteolytic modification may include several different steps, such as reduction of molecular weight by partial removal of the C-terminus, or cleavage of polypeptides from precursors, with the examples of β- glucoronidase and cathepsin D respectively (Rosenfeld et al., 1982; Erickson et al., 1983). Non-enzymic protein co-factors such as saposins may then be required in order for various lysosomal enzymes to function and act on lipophilic substrates. Moreover, four different saposins derived from the precursor, prosaposin are transported to the lysosome via the M6P recognition pathway (Winchester et al., 2000). Such co-factors may be of relevance to the NCLs, as saposins A and B have been identified as the primary components of autofluorescent storage material in Infantile NCL (INCL) (Tyynelä et al., 1993).

Within the M6P pathway, different classes of endosomes share many structural similarities with each other, although several differences are suggested to exist between the early and late forms of these compartments. These include morphological phenotype, location within the cell, period of acquisition of endocytosed material, biochemical composition and density (Murphy et al., 1991), and relative acidity, with the late endosome harbouring a more acidic environment (Mellman et al., 1986). The aforementioned pathway details the early endosome functioning to generate vesicles which transfer endocytosed material to the late endosome, with subsequent transfer to

30 the lysosome. Alternatively, another pathway may exist, where the early endosome undergoes ‘maturation’, a process that involves its transformation into a late endosome. The late endosome would then mature into a lysosome, as it acquires all the necessary lysosomal hydrolases and membrane proteins that are brought in through clathrin coated vesicles, originating from the trans Golgi network (Brown et al., 1986; Griffiths et al., 1988). Since lysosomal hydrolases are present within endosomal multivisicular bodies, these compartments may represent an intermediate stage between the transformation of endosomes into lysosomes. Although, consistent with the idea that endosomes and lysosomes are separate pre-existing entities, multivesicular bodies would usually transfer endocytosed material from the early endosome to the late endosome, and therefore would be expected to be present in both of these compartments (Griffiths et al., 1991). Furthermore, structural compositions of proteins within early and late endosomes differ such that it is likely for two separate and very stable endosomal compartments may exist (Schmid et al., 1988).

b) Transport of membrane associated proteins to the lysosome

The transport of lysosomal membrane proteins to the lysosome does not involve mannose 6-phosphate markers and, in that sense, differs from lysosomal hydrolase trafficking. There are both direct and indirect pathways for the transport of proteins from the membrane to the lysosome. The direct pathway requires signals that are responsible for endocytic entry of lysosomal membrane proteins, to promote their entry from early/late endosomes and lysosomes. Mutagenesis studies lend support to this theory, as mutant tyrosine residues within the cytoplasmic domain of lysosome membrane proteins cause inadequate/inaccurate targeting of membrane proteins to lysosomes (Williams et al., 1990).

The indirect pathway for delivery of lysosomal membrane proteins into the lysosome is similar to that of lysosomal hydrolases, whereby, the majority of membrane proteins are separated from other proteins in the trans Golgi network before they are delivered to a developing lysosome. Labelling and cell fractionation experiments accredit this view, as newly synthesized lysosomal membrane proteins are found in

31 lysosomes, with such rapid kinetics that it is extremely unlikely that they were able to reach the cell surface before being internalised via endocytosis (Green et al., 1987). Although it is likely that both of these pathways occur, the contribution from each pathway is likely to vary depending on the cell type that is involved and the stages of cell differentiation (Harter et al., 1992; Carlsson et al., 1992).

c) Lysosomal dysfunction

It is suggested that lysosomal dysfunction is an important part of pathogenesis in the NCLs (Hofmann and Peltonen, 2001; Weimer et al., 2002). Defective proteins responsible for maintaining the lysosomal/endosomal M6P pathway, may subsequently lead to the accumulation of characteristic storage material that is observed in the NCLs and other groups of storage disorders (Neufeld et al., 1975; Goebel, 1995; Winchester et al., 2000; Hofmann and Peltonen, 2001). There are several primary and contributory factors that may aid the disruption of lysosomal function. These factors are almost exclusively genetic, involving defective lysosomal enzymes, their co-factors (e.g. saposins), lysosomal membrane proteins and proteins involved in post-translational modification or transport (Neufeld et al., 1975; Mole, 1998; Winchester et al., 2000; Hofmann and Peltonen, 2001). Such defects typically lead to the accumulation of partial degradation products and even whole substrate in the instance of transporter dysfunction (Neufeld et al., 1975; Winchester et al., 2000; Hofmann and Peltonen, 2001). Pathological conditions resulting from defective lysosomes are referred to as lysosomal storage disorders or diseases (Neufeld et al., 1975). The nature of pathological insult to the lysosomal system and the cell types and tissues that are affected generally determine the severity of clinical symptoms.

d) Incorrect processing of lysosome/Mutations within lysosomal enzymes

Improper coding of lysosomal enzymes and/or contributory factors may either prevent their production, or lead to incomplete processing of the protein into the mature form

32 (Hasilik and Neufeld, 1980; Tammen et al., 2001). Abnormal processing may suggest that the mutated enzyme either failed to reach the lysosome, or it was unstable in an acidic environment. In the latter scenario, in vitro enzyme may be stabilised by incubating cells with alkaline (von Figura et al., 1983; Lemansky et al., 1987), or treating them with protease inhibitors (von Figura et al., 1983). However, under either of these conditions it is still not uncommon for the mutant enzyme to remain unbound to the mannose 6-phosphate receptor, thus preventing their secretion. Similar to other incorrectly folded proteins, mutant lysosomal enzymes may then be confined to the ER where they are degraded (Lau et al., 1989).

1.3 Lysosomal Storage Disorders

The precise nature of lysosomal dysfunction in the NCLs and many other lysosomal storage disorders is poorly understood, although there may be some overlap in the normal function of defective machinery within the lysosome (Nishino et al., 2000; Futerman and van Meer, 2004). Lysosomal dysfunction is typically associated with defective enzymes (Neufeld et al., 1975; Futerman and van Meer, 2004); however, it is also possible for this to be accredited to abnormal proteins within the lysosomal membrane (Nishino et al., 2000). In order to understand the spectrum of complex neuropathological changes that are evident amongst the NCLs, it is informative to explore the mechanisms that operate in other lysosomal storage disorders, considering how these defects arise in each disorder. These disorders may typically arise from defects in a lysosomal enzyme, or may be related to dysfunction of transmembrane protein expressed in the lysosomal membrane.

1.3.1 Enzyme related lysosomal dysfunction a) Gangliosidoses

The gangliosidoses belong to a group of storage disorders, which result from the accumulation of the ganglioside GM2 and associated glycolipids in the lysosome

33 (Gravel et al., 2001). Gangliosides are glycosphingolipids composed of a hydrophobic ceramide and hydrophilic oligosaccharide chain, bearing one or more sialic acid residues (Ledee et al., 1984). Possible roles in neuronal recognition (Brady, 1982), myelination (Brady et al., 1973) and synaptogenesis (Brady, 1976) have been speculated, due to an optimal period of ganglioside turnover in the central nervous system during neonatal development, a period just before myelination. Furthermore, greater concentrations of gangliosides are observed in the neocortex, which is rich in grey matter, compared to white matter-rich regions such as the optic chiasm, internal capsule, or corpus callosum (Brady, 1982; Kracun et al., 1984). The catabolism of GM2 requires hexaminidase and neuraminidase in the presence of a substrate-specific cofactor known as GM2 activator. Hexaminidase has two isoenzymes, including HexA and HexB, which are composed of subunits αβ and ββ respectively. HEXA encodes the α subunit of HexA, which is required for catabolism of the GM2/GM2 activator complex into GM3, and HEXB encodes the β subunit of both HexA and HexB, with only the latter enzyme effective in the further breakdown of GM2 cleavage product, GA2 (Gravel et al., 2001).

There are three main forms of gangliosidoses, which arise from defects in genes encoding HexA, HexB and the GM2 activator. These include Tay-Sachs disease, Sandhoff disease and GM2 activator deficiency (Brady, 1982). In Tay-Sachs there is a deficiency in HexA, preventing the cleavage of the N-acetylgalactosamine terminal of GM2 (Sandhoff et al., 1971; Neufeld et al., 1975). Conversely, the activity of the other hexaminidase enzyme, HexB is increased several fold during this period (Brady, 1982). Sandhoff disease exhibits deficiencies in both HexA and HexB, thus GM2 and GA2 accumulation occurs more readily than in unaffected brains. This is due to defects of the β subunit, which is common to both hexaminidase enzymes (Sandhoff et al., 1968; Sandhoff et al., 1971) and is therefore important for the catabolism of these gangliosides, but is not the full story. The neuroaminidase pathway is also intrinsically involved in this process and helps to reduce levels of GM2 by hydrolysing it to GA2 (Brady et al., 1982). However, this neuramidase pathway works as part of a system and cannot clear gangliosides independently. GM2 activator deficiency is a very rare member of the gangliosidoses (Gravel et al., 2001), in which both the α and β subunits are intact and the encoded hexaminidase enzymes remain

34 unaffected. In this disorder, the GM2 activator is diminished, thus resulting in the pathogenic accumulation of GM2 (Conzelmann et al., 1978).

b) Mucopolysaccharidoses

Mucopolysaccharidoses are caused by the deficiency of a variety of enzymes needed to degrade glycosaminoglycans (Mucopolysaccharides) (Dorfman et al., 1976; Neufeld et al, 1978). These include dermatan sulphate, heparan sulphate, keratan sulfate and chondroitin sulphate and are themselves, lysosomal degradation products, which stem from proteolysis of the central protein of proteoglycans. Hyuluronan is another glycosaminoglycan, although it has no protein core (Neufeld et al., 2001). There are a total of eleven known enzyme deficiencies, which may either prevent the catabolism of a single, or combination of glycosaminoglycans and are responsible for seven distinct mucopolysaccharidoses (MPS I-VII). These conditions are all transmitted in an autosomal recessive manner, except for MPS II which is linked to the X chromosome (Dorfman et al., 1976; Neufeld et al., 2001). Partial or incomplete degradation of glycosaminoglycans leads to their accumulation within the lysosome, resulting in cell, tissue and possibly organ dysfunction. The majority of these disorders are caused by small changes within the gene, such as point mutations, although major changes and deletions can occur in MPS II (Neufeld et al., 2001).

Accumulation of gangliosides has also been demonstrated in patients with MPS and has been linked to damage of the central nervous system. Findings come from patients with MPS IH, II, and III, where significant changes in levels of gangliosides were noted in the brain and isolated neuronal preparations (Constantopolous et al., 1978). Minor monosialogangliosides GM1, GM2 and GM3 were also found in the brain at levels similar to that observed in lipid storage disorders (Constantopolous et al., 1978, 1980).

35 c) Glycoproteinoses

Glycoproteinoses is the collective name given to a group of autosomal recessive disorders involving the deficiency of a lysosomal enzyme that is required for catabolism of glycoproteins (Michalski et al., 1999). The presence of oligosaccharide chains covalently bound to a polypeptide backbone defines the structure of a glycoprotein, with its bonds either formed through hydroxyl groups of serine or arginine, or via an unused amino group of asparagine (Thomas et al., 2001). Large amounts of these materials are degraded through normal metabolism, due to their wide distribution and turnover within cells, in particular within the lysosome. Interference of any step in the process may lead to cellular filling and pathology, which are clear signs of disease phenotype (Warner et al., 1983). This degradation process involves the break down of the protein backbone of the glycoprotein, and either removal of sugars or a minor cleavage at nonreducing and reducing termini of Asn-linked oligosaccharides respectively. Proteins are degraded using lysosomal proteases and peptidases, whereas cleavage at the nonreducing termini is processed via several exoglycosidases (Aronson et al., 1989). Lysosomal enzymes involved in this process include neuroaminidase (Sialidase), α-fucosidase, α-mannosidase, β- mannosidase, β-galactosidase and β-N-acetylhexaminidase (Thomas et al., 2001). When defects occur in each of these enzymes, a number of corresponding disorders arise, including sialidosis, fucosidosis, α-mannosidosis and β-mannosidosis (Thomas et al., 2001), which have a range of ages of onset and rates of clinical progression.

d) Mucolipidoses

Mucolipidoses are disorders characterised by abnormal lysosomal enzyme transport within cells originating from the mesenchyma. In these disorders, such enzymes are not targeted correctly to the lysosome, but instead secreted into the extracellular medium (Kornfeld et al., 2001). The main defect is thought to be due to the absence of an uptake signal from patient enzymes and a reduced level of phosphorylation. Evidence for this defect is obtained from cultured fibroblasts that display reduced phosphorylation of the oligosaccharide component of glycoprotein enzymes (Bach et

36 al., 1989; Hasilik et al., 1980). Phosphotransferase is an important enzyme required for synthesis of the mannose 6-phosphate recognition marker and its specificity within the process. This enzyme is composed of three subunits (α2β2γ2) and possesses two domains, one of which has catalytic properties and the other has the unique ability to distinguish between lysosomal enzymes and the numerous luminal glycoproteins within the Golgi apparatus, for selective phosphorylation (Waheed et al., 1982; Kornfeld et al., 1999). Moreover, its substrate, mannose 6-phosphate is responsible for providing the signal for the internalisation and compartmentation of enzymes during their passage to the lysosome (Brown et al., 1986). However, when enzymes are underphosphorylated, they do not partake in this process and are released into the serum and urine (Kornfeld et al., 2001).

I-cell disease (mucolipidosis II) and pseudo Hurler polydystrophy (mucolipidosis III) belong to the family of mucolipidoses, which are genetic disorders of an autosomal recessive nature. These disorders each present with symptoms like Hurler syndrome, although age of onset and clinical severity vary, with I-cell disease being the most aggressive form of the disease and of earlier onset (Reitman et al., 1981). To some extent, these phenotypic similarities may be attributed to a deficiency in the first enzyme of the mannose 6-phosphate biosynthetic pathway (UDP-N- acetylglucosamine-1-phosphotransferase), as demonstrated by studies of fibroblasts from these patients (Reitman et al., 1981).

1.3.2 Non-enzyme lysosomal dysfunction

Danon disease

Danon disease is caused by the deficiency of lysosomal-associated membrane protein- 2 (LAMP2), which belongs to the family of LAMP proteins and is an essential component of the lysosomal membrane (Nishino et al., 2000). Taken together with lysosomal integral membrane proteins (LIMPs), this family of proteins are tightly packed and represent over fifty percent of the total membrane protein in both late endosomes and lysosomes (Marsh et al., 1987). The structurally related LAMP1

37 protein shares 37% amino acid sequence homology with LAMP2 and can be functionally substituted, although, the roles of LAMP2 are somewhat more complex (Eskelinen et al., 2003). Insights into the normal function of this protein have come from LAMP2 null mutant mice, which suggest its involvement in lysosomal biogenesis, enzyme targeting and autophagy (Eskelinen et al., 2003). In these mice, LAMP2-deficient hepatocytes, display elevated levels of a subset of secreted lysosomal enzymes and the improper processing of cathepsin D (Eskelinen et al., 2002). M6P processing is also severely affected, as demonstrated by a severely reduced half life of radiolabelled M6P and the mislocalisation of M6P from to the trans Golgi network to autophagic vacuoles (Eskelinen et al., 2002). The presence of these autophagic vacuoles within these mice is indicative of a connection between LAMP2 and autophagic cell death, and as such is a feature observed in several tissues, although the brain and fibroblasts do not show this accumulation (Eskelinen et al., 200). Clinically, Danon disease is characterized by cardiomyopathy, skeletal myopathy and mental retardation in males, however, of these conditions, only cardiomyopathy is prevalent in females (Sugie et al., 2002).

1.4 The Neuronal Ceroid Lipofuscinoses: Enzyme and non-enzyme related lysosomal dysfunction

Amaurotic familial idiocy (AFI) was first described in 1826 by Christian Stengel, who reported the observations of four Norwegian siblings with characteristics of JNCL, including progressive blindness, epilepsy, cognitive decline, and motor dysfunction (Stengel, 1826). Several years later, Frederick Batten demonstrated an intraneuronal storage process in a group of patients (Batten, 1903), who had very similar clinical observations to those observed by Stengel (Zeman et al., 1970). Due to the histochemical resemblance of storage material to ceroid or lipofuscin, this group of debilitating disorders were subsequently termed neuronal ceroid-lipofuscinosis (NCL, Zeman and Dyken, 1969), in order to distinguish them from Tay Sachs disease, which belongs to a separate group of storage disorders. Today, there are at least seven distinct forms of NCL which are often still referred to collectively as Batten disease. Taken together, the NCLs are the most common childhood neurodegenerative

38 disorders, having an estimated incidence of 1 in 12,500 births (Goebel et al., 1995) and are almost exclusively autosomal recessive, with the exception of extremely rare adult forms (Mole, 1998).

The neuronal ceroid lipofuscinoses (NCLs) are subdivided into four main categories, including infantile, late infantile, juvenile and adult forms, corresponding to age of onset and rate of progression of the disease (Goebel et al., 2004; Goebel and Wisniewski, 2004). To date the total spectrum of NCLs result from mutations in eight genes, of which six genetic loci have been identified through linkage analysis (International Batten Disease Consortium, 1995; Vesa et al., 1995; Sleat et al., 1997; Savukoski et al., 1998; Ranta et al., 1999; Gao et al., 2002; Wheeler et al., 2002) and another via radiolabelling and two-dimensional gel analysis (Sleat et al., 1997). Four of these genes have been described in detail (CLN1, CLN2, CLN3 and CLN5), with over 20 mutations reported thus far in CLN1 and CLN3 (Hofmann et al., 2001). Although most of these mutations lead to typical NCL phenotypes, some mutations result in a slightly milder or delayed disease state that may be attributed to residual function of the encoded protein (van Diggelen et al., 2001; Mole, 2004). Clinical diagnosis of the NCLs is traditionally determined by a combination of clinical presentation and ultrastructural analysis (Wisniewski et al., 2001a; Goebel and Wisniewski, 2004). However, recent efforts also utilize enzymatic assays and assessment of DNA, for increased efficiency and accuracy of diagnosis (Wisniewski et al., 2001b; Goebel and Wisniewski, 2004).

1.4.1 Storage material

The accumulation of autofluorescent storage material is a common neuropathological feature that distinguishes the NCLs from other lysosomal storage disorders and was pivotal in their naming (Zeman and Dyken, 1969). The autofluorescent properties of the storage material, as viewed by epifluorescence, or via confocal microscopy (Hall et al., 1989; Cotman et al., 2002; Bible et al., 2003), are attributable to its major biochemical component known as ceroid lipofuscin. The major protein component of stored material is subunit c of mitochondrial ATP synthase which is present in late

39 infantile, juvenile and adult forms of the disease (Tyynelä et al., 2004; Oswald et al., 2005). However, subunit c deposition it is not a consistent feature in INCL (Hofmann and Peltonen, 2001). Autofluorescent storage material is also rich in protein, lipid and carbohydrate, although the relative proportion of these substances varies due to the heterogeneity of such depositions (Hofmann and Peltonen, 2001; Goebel and Wisniewski, 2004). Ultrastructural examination of these complexes in both autopsy and biopsy material has revealed several different inclusion bodies, which are reported to be present in specific combinations in each sub-type of NCL. These inclusion bodies include granular osmiophilic deposits, curvilinear profiles, fingerprint profiles, and rectilinear complexes (Figure 1.2; Elleder et al., 1999; Hofmann and Peltonen, 2001).

Granular osmiophilic deposits are granular rounded bodies, which are bound by a membranous sac and forms aggregates. These are commonly used to diagnose infantile NCL (CLN1). Curvilinear profiles are stacks of uniformly curved lamellar, which form alternating light and dark lines. These inclusion bodies are membrane bound and are a major constituent of the lysosome in late infantile NCL (CLN2). Fingerprint profiles are a series of parallel lines with a lucent central line. These are present within JNCL (CLN3), LINCL and vINCL (CLN6). Rectilinear profiles are typically oligolamellar with a prevailing straight course, differing from curvilinear profiles by shape and a dense central line. This is the dominant inclusion body of storage material within CLN3 and is also found in Finnish variant LINCL (CLN5) and CLN6 (Elleder et al., 1999; Goebel and Wisniewski, 2004b).

Ultrastructural analysis of autofluorescent storage material is usually performed via examination of peripheral blood leukocytes or other available tissue and has been used for laboratory-based diagnosis of NCL type (Wisniewski et al., 2001a; Haltia, 2003). In conjunction with this approach, diagnoses now include enzymatic assays, which are only possible for infantile and late infantile varieties (Wisniewski et al., 2001b), and analysis of DNA, where infantile, late infantile, juvenile, Finnish variant late infantile, and variant infantile forms can be assessed (Wisniewski et al., 2001a; Mole, 2004).

40

Figure 1.2 The ultrastructural appearances of the abnormal intraneuronal in different forms of the neuronal ceroid lipofuscinoses (NCLs). (A) Granular osmiophilic deposits are typical of INCL, x10 000; (B) Curvilinear profiles are typical of LINCL, x 20 000; (C) Fingerprint bodies are typical of JNCL, x 30 000; (D) Many inclusions are typically present in variant forms of NCL, x 15 000. Taken from Haltia et al., 2003.

1.5 NCL proteins

The CLN genes encode a structurally diverse group of proteins that fall into two main categories, either lysosomal enzymes which are encoded by CLN1 and CLN2 (Vesa et al., 1995; Sleat et al., 1997), or transmembrane proteins of unknown function encoded by CLN3, CLN5, CLN6 and CLN8 (International Batten Disease Consortium, 1995; Savukoski et al., 1998; Ranta et al., 1999; Gao et al., 2002; Wheeler et al., 2002).

a) Lysosomal enzymes

CLN1 – Infantile NCL i) Normal function/ processing Palmitoyl protein thioesterase 1 (PPT1) is a soluble, lysosomal enzyme encoded by the CLN1 gene, which exists in several glycosylated forms (Camp et al., 1993). This enzyme is a soluble hydrolase that is capable of removing long-chain fatty acids from cysteine residues of proteins (Verkruyse et al., 1996) and is essential for degradation within the lysosomal compartment. After PPT1 is transported to the lysosome via the M6P receptor pathway (Hellsten et al., 1997; Verkruyse et al., 1996), thioester bonds

41 are hydrolysed from palmitate or other fatty acid S-acylated proteins, which are then degraded. Palmitoylation is a common post-translational modification within proteins and may regulate certain signal transduction by determining protein- protein/membrane interactions (Casey, 1995; Mumby, 1997). Nevertheless, there is currently little knowledge of the normal in vivo substrates of PPT1.

PPT1 mRNA is shown to be widely distributed within the tissues and organs of humans, rats and mice (Camp et al., 1994; Schriner et al 1996). The levels of PPT1 expression vary considerably between tissues, with the brain, lungs, spleen, and testis having relatively high levels of expression, but lower levels in the liver and skeletal muscle (Schriner et al., 1996). Microscopically, PPT1 in the rat brain has been localised to neocortical neurons, cerebellar Purkinje cells and Bergmann glia, and is expressed in a developmentally regulated manner (Soupanki et al., 1999). A more detailed mapping of PPT1 within the CNS of mice reveals a widespread distribution that was confined to the grey matter, which was totally absent from white matter (Bible, personal communication). During maturation of the human CNS, the temporal expression of this enzyme increases, as does PPT1 mRNA throughout cortical development (Heinonen et al., 2000). At embryonic day seventy six, PPT1 mRNA expression is weak within the ventricular zone, but is substantially stronger in cells migrating towards the cortical plate. As immature neurons are migrating out of the ventricular zone during this period of corticogenesis, PPT1 has been suggested to be involved in establishing laminar organisation (Heinonen et al., 2000).

ii) Mutations in PPT1 INCL is caused by a mutation in the PPT1 encoding region of the CLN1 gene on chromosome 1p32 (Vesa et al., 1995). The most common mutation is a single base pair transversion (tryptophan to arginine), with a carrier frequency of 1 in 70 in the Finnish population (Vesa et al., 1995). To date, 40 different mutations have been reported in this gene, of which 19 are missense, 9 are nonsense, 9 are small deletions or insertions and 3 are mutations affecting splice sites (Mole, 2004). X-ray crystallography of the mature bovine enzyme, which shares 95% homology to human PPT1 (Bellizzi eta al., 2000), has revealed the three dimensional structure of the PPT1

42 protein. This analysis has revealed its monomeric, globular appearance, with three distinct glycosylation sites and a catalytic component consisting of three conserved proteins (Ser115, Asp233 and His289) (Bellizzi et al., 2000). CLN1 mutations that involve the active site of the PPT1 protein may affect its geometry and therefore change the conformation of the binding pocket (Bellizzi et al., 2000). This decrease in enzyme stability may be far more efficacious to disease phenotype than a reduction of catalytic activity (Das et al., 2001). Nonsense mutations are also evident and are more serious as they cause premature truncation of the protein, leading to a total loss of PPT1 activity (Das et al., 1998). As the onset and progression of the disease has been associated with varying levels of enzyme activity, some forms of PPT1 deficiency are clinically indistinguishable from other forms of NCL and require enzyme assay analysis for identification (van Diggelen et al., 2001).

CLN2 – Late Infantile NCL i) Normal function/ processing Tripeptidyl peptidase I (TTP-I) is a lysosomal exopeptidase that is encoded by the CLN2 gene (Sleat et al., 1997) and belongs to the family of recently defined serine- carboxyl proteinases (Wlodawer et al., 2001). During maturation, residues of 179 and 16/19 amino acid acids are separated from the 563 amino acid protein, leaving the mature enzyme with 368 amino acid residues (Liu et al., 1998; Lin et al., 2001). This enzyme is essential for degradation of tripeptides and makes them available for cellular metabolism (Tomkinson, 1999), a feature that is achieved via their cleavage from the N-termini of oligopeptides. Studies of the purified enzyme in bovine and human brain have also suggested an intrinsic endoproteolytic activity which may aid conversion of the 66kDa proenzyme into its mature form of 46 kDa (Junaid et al 2000; Lin et al., 2001). TPP-I has an acidic pH optimum, requiring this environment for normal proteolytic processing and enzymatic activation (Vines et al., 1998; Sohar et al., 1999; Lin et al., 2000). Moreover, amino acids, Ser475, Asp360, and Asp517 are also essential for optimal activity, as inferred by mutagenesis studies (Lin et al., 2000). Natural substrates for TPP-I activity remain unresolved, although subunit c of mitochondrial ATP-synthase is a possible candidate, which accumulates within

43 autofluorescent storage material, with decreasing enzyme activity (Ezaki et al., 1999). The distribution of TPP-I is widespread within a variety of tissue samples and organs (Kida et al., 2001; Kurachi et al., 2001). However, there are notable differences in the spatiotemporal distribution of this enzyme (Kida et al., 2001; Kurachi et al., 2001), particularly in the cerebral cortex, where adult levels are reached much earlier on during development (Kurachi et al., 2001). Prior to neuronal expression, TPP-I is present in endothelial cells, choroid plexus, microglial cells, and ependyma (Kida et al., 2001). This enzyme is initially detectable within neurons during gestation, and then increases postnatally, closely mirroring neuronal differentiation and maturation (Kida et al., 2001).

ii) Mutations in TPP-I LINCL is caused by mutations to the CLN2 gene within chromosome 11p15 (Sleat et al., 1997). To date, 52 different mutations have been identified, with 7 small deletions, 2 small insertions 29 missense mutations, 5 nonsense mutations and 9 mutations effecting intronic sequence or affecting splice sites (Mole, 2004). Out of these, there are two mutations with the greatest prevalence, including IVS5-1 G>C and R208X, which account for 33% and 26% of all LINCL mutations respectively (Mole et al., 1999).

Similar to PPT1, the crystalline structure of a TPP-I related enzyme (Pseudomonas serine-carboxyl proteinase) also contains a catalytic triad (Glu80, Asp84 and Ser287) and a separate residue (Asp170), which is essential for the catalytic activity of the enzyme (Wlodawer et al., 2001). Curiously, the equivalent residues in human TTP-I are very similar, with a triad of Glu272, Asp276 and Ser495 and an Asp360 residue required for enzyme activity (Mole, 2004). Recent findings also implicate glycosaminoglycans (GAGs) as essential for TPP-I activity. This family of proteins essentially aid in the conversion of TPP-I proenzyme into its mature form and lower pH requirement for autoactivation of the enzyme (Golabek et al., 2005). Mutations to CLN2 may either directly affect catalytic activity, or alter the structural integrity of TPP-I, thus affecting potential interactions with GAGs and altogether abolishing enzymic function.

44 b) Novel proteins of unknown function

CLN5 – Finnish Variant Late Infantile NCL i) Normal function/ processing CLN5 encodes a 407 amino acid glycosylated peptide with a molecular weight of 60 kDa that is primarily located in the lysosome (Isosomppi et al., 2002). Due to our limited understanding of the CLN5 protein, little is known about its localisation and function. However, predictions of two hydrophobic membrane spanning regions with an intraluminal loop (Savukoski et al., 1998), coupled to studies within transfected CLN5, suggest that the use of different start methionines may produce both soluble and membrane bound forms of the CLN5 protein (Isosomppi et al., 2002; Vesa et al., 2002). Furthermore, several motifs within CLN5 have been identified using sequence analysis techniques in conjunction with protein prediction programs (Nakai et al., 1992; Solovyev et al., 1994) and suggest posttranslational modification of the protein. These potential modification sites include protein kinase C phosphorylation sites, casein kinase II phosphorylation sites, tyrosine kinase phosphorylation sites, N- myristoylation sites and N-glycosylation sites (Savukoski et al., 1998). The CLN5 protein is only found in vertebrates, where it is synthesized into four precursor forms (Isosompi et al., 2002). Its expression is markedly elevated during neocortical neurogenesis and is also intense in cells of the thalamus as well as in the future Purkinje cell layer of the cerebellum. Therefore, CLN5 is suggested to be important for development of a wide range of maturating neurons (Heinonen et al., 2000; Holmberg et al., 2004).

i) Mutations in CLN5 Finnish variant LINCL (FvLINCL) is caused by a mutation of the CLN5 gene, that is located on chromosome 13 (Savukoski et al., 1994; Pineda-Trujillo et al., 2005). To date, five disease mutations in CLN5 have been identified, three of which result in the premature termination of the polypeptide chain (Isosomppi et al., 2002) and the other two result in amino acid substitutions (Pineda-Trujillo et al., 2005). The most common of these accounts for 94% of cases in Finland and is caused by a two base

45 pair deletion in exon 4 of CLN5, which induces a premature stop codon and results in the generation of a truncated protein (Ranta et al., 2001). This common mutation appears to block the lysosomal targeting of transfected CLN5, suggesting that the pathology of FvLINCL is caused by defective trafficking (Isosomppi et al., 2002).

CLN6 – variant Late Infantile NCL i) Normal function/ processing CLN6 is a gene that is localized to chromosome 15q21-23 and encodes a 36 kDa polypepetide chain of approximately 311 amino acids (Gao et al., 2002; Wheeler et al., 2002). CLN6 is suggested to be a transmembrane protein, due to its predicted five membrane spanning regions (Wheeler et al., 2002). More recently, CLN6 has been shown to reside within the ER of human cells, as demonstrated by cell cultures of HEK293 cells and immunofluorescence microscopy (Mole et al., 2004). CLN6 shares no sequence homology or functional domains with other known proteins (Mole, 2004), although the role of the ER in protein and lipid synthesis present the possibility of involvement in the same biological pathway (Mole et al., 2004). Other clues for CLN6 protein function come from studies of naturally occurring animal models, including the nclf mouse (Bronson et al., 1998) and the South Hampshire and Merino sheep (Broom et al., 2001; Tammen et al., 2001), which are unable to generate a normal Cln6 protein and therefore exhibit an NCL-like phenotype.

ii) Mutations in CLN6 Variant forms of INCL (vINCL) are caused by mutations in the CLN6 gene, which may lead to abnormal processing of the protein and defective function (Gao et al., 2002; Wheeler et al., 2002). So far, a total of eighteen different mutations have been identified in vINCL patients, all of which give rise to a similar pathological phenotype apart from one instance (Sharp et al., 2003). This was reported from a compound heterozygote for a missense mutation and an unidentified mutation that resulted in a protracted disease progression (Sharp et al., 2003). Other mutations include nonsense, small deletions or insertions, and two splice-site mutations, some of which generate

46 truncated proteins. There does not appear to be a founder mutation in CLN6, although one mutation is more prevalent in the population of Costa Rica than the other mutations originating from this country (Sharp et al., 2003).

CLN8 – Epilepsy with Mental retardation i) Normal function/ processing The CLN8 gene is located on chromosome 8p23 and is predicted to encode a protein of 286 amino acids, with a molecular mass of 30 kDa (Ranta et al., 1999). The exact function of this protein remains unclear, although various properties have been hypothesised. Lonka et al. (2000) described the CLN8 protein as likely to be a transmembrane protein, with several membrane spanning domains, due to its predicted hydrophobicity. A series of markers for subcellular organelles were then used in conjunction with an antibody to CLN8, which suggested that localisation of this protein was primarily in the ER, where CLN8 is thought to constitutively cycle between this compartment and the ER-Golgi intermediate complex (Lonka et al., 2000). A further study investigated spatiotemporal distribution of Cln8 in mnd mice, revealing a prominent expression of this protein in the CNS of embryonic tissue, which was greatest in the neocortex and hippocampus of postnatal brains (Lonka et al., 2005). Indeed a role for Cln8 has been suggested in the maturation, differentiation and survival of neuronal populations (Lonka et al., 2005).

ii) Mutations in CLN8 Mutations in the CLN8 gene result in a condition known as progressive epilepsy with mental retardation (EMPR), or alternatively Northern Epilepsy (NE, Lonka et al., 2000). EMPR/NE is a variant form of NCL caused by a single missense mutation of exon 2 in the CLN8 gene, which results in the substitution of an arginine to glycine in all patients studied to date (Ranta et al., 1999; Herva et al., 2000). Defects to this gene have also been associated with motor neuron degeneration in the naturally occurring mnd mouse. This NCL model is caused by a 1 bp insertion in codon 90 and displays

47 autofluorescent inclusion bodies similar to that observed in EMPR/NE (Messer et al., 1986; Cooper et al., 1999; Ranta et al., 1999).

A closely related gene, CLN7 was originally assigned to a small subset of vLINCL cases that formed a cluster in six Turkish families (Wheeler et al., 1999; Mitchell et al., 2001). This gene has not yet been identified, although recent findings suggest that this form of NCL may be due to an allelic variant of the CLN8 gene (Mitchell et al., 2001).

c) Rarer forms of NCL

CLN4 – Adult onset NCL

The most prevalent adult onset form of NCL (ANCL) is Kufs disease which has an autosomal recessive inheritance. Parry disease is another rarer form of ANCL that exhibits autosomal dominant inheritance (Martin et al., 1999). Both of these disorders remain largely uncharacterised, with no gene reported to date. This may be partially because insufficient ANCL families are available to perform genetic linkage studies, to map and subsequently identify the CLN4 gene (Martin et al., 1999). Alternatively, minor mutations in other NCL genes may leave residual levels of biological activity and a delayed disease onset, as has been reported for PPT1 (van Diggelen et al., 2001). Apart from these minor mutations causing delayed onset, attempts to link the genetic defect in ANCL families to other known NCL mutations (CLN1, CLN3 and CLN6) have been unsuccessful, which strongly suggests Kufs and Parry disease to be distinct disorders (Mole, 1998; Nijssen et al., 2003).

CLN9

A CLN9 variant form of NCL has recently been proposed based on clinical course and the distinctive appearance of intracellular inclusions (Schulz et al., 2004). Enzyme assays and molecular testing for known variants of NCL appeared normal,

48 and gene expression profiles were distinctly different from normal and all known variant forms of NCL. However, the mutation responsible for this disease phenotype has not yet been identified (Schulz et al., 2004).

1.6 Juvenile NCL, the most common form of NCL

CLN3

JNCL is the most common form of NCL and is caused by defects in the CLN3 gene (International Batten Disease Consortium, 1995). The incidence of JNCL varies across the globe, but is most prevalent in Iceland, with up to seven per hundred thousand live births affected by the disease (Wisniewski et al., 2001a). In order to understand the complex pathogenic nature of this disorder, it is necessary to have an understanding of structure, localisation, function and instances of mutations within the CLN3 gene.

i) Structure of CLN3 CLN3 is a membrane protein that is composed of 438 amino acids and has a molecular weight of approximately 43 kDa (International Batten Disease Consortium, 1995), but its precise structure and function remain elusive. Due to its hydrophobic nature, antibodies raised against this polypeptide yield unreliable and often conflicting results (Ezaki et al., 2003; Mao et al., 2003a; Kyttala et al., 2004). Thus, multiple techniques must be utilized and great caution taken when interpreting these studies. Several computer prediction models have suggested CLN3 to be an integral membrane protein (Janes et al., 1996; Mitchison et al., 1997; Mao et al., 2003a), a proposition that has been confirmed by its detection in the detergent phase, during partition experiments (Kaczmarski et al., 1999). These groups have also predicted the topology of CLN3 to have between five to eight transmembrane spanning regions (Janes et al., 1996; Mitchison et al., 1997; Mao et al., 2003a). To determine between these models and reveal the orientation of CLN3 within the membrane, flag tags were

49 50

Figure 1.3 Predicted topologies of the CLN3 protein. (A) Based on topology proposed by Mao et al. (Mao et al., 2003a). Five transmembrane domains with the amino terminus residing in the lumen and the carboxy terminus in the cytosol. (B) Based on topology proposed by Ezaki et al. and Kyttala et al. (Ezaki et al., 2003; Kyttala et al. 2003). Six transmembrane domains with both the amino and carboxy termini in the cytosol. (Adapted from Phillips et al., 2005)

used in conjunction with antibodies against different regions of the CLN3 protein (Mao et al., 2003a). Separate studies confirmed the amino terminal tail to be cytosolic (Mao et al., 2003a; Kyttala et al., 2004; Ezaki et al., 2003), although, the membrane spanning loops and amino terminus were highly variable, facing either the cytosolic or lumenal space. Thus, two distinct models have emerged predicting both five (Figure 1.3A) and six transmembrane domains (Figure 1.3B), with the N-terminus facing a different side of the membrane in each case (Mao et al., 2003a; Kyttala et al., 2004; Ezaki et al., 2003). However, it is more likely for the six transmembrane model to hold true, as the CLN3 antibodies used in this study appear to be the most specific to date, according to Western analysis and mass spectroscopy (Ezaki et al., 2003). This model suggests that carboxy and amino termini are cytoplasmic in addition to the fourth loop domain, whereas loops one, three and five are luminal (Kyttala et al., 2004; Ezaki et al., 2003).

The CLN3 protein undergoes a variety of posttranslational modifications, including glycosylation, phosphorylation and lipid modifications (Pullarkat and Morris, 1997; Michalewski et al., 1999; Ezaki et al., 2003). These modifications have been demonstrated on polyacrylamide gels as a change in the predicted molecular mass (Jarvella et al., 1998). Moreover, translated CLN3 was shown to increase from the native form of 43 kDa to a 45 kDa form of the protein (Jarvella et al., 1998). Detailed sequence analysis of the human CLN3 protein suggests four potential N-glycosylation sites (residues 49, 71, 85, and 310), two putative O-glycosylation sites (residues 80 and 256), and two potential glycosaminoglycan sites (residues 162 and 186, Golabek et al., 1999). Further evidence of glycosylation has been obtained from mouse liver and brain extracts that were deglycosylated by PNGase F treatment with subsequent detection via multiple antibodies (Ezaki et al., 2003). This treatment caused the liver specific band, which ran at 60 kDa, and the brain specific band, which ran at 55 kDa, to both shift back to 45 kDa (Ezaki et al., 2003).

CLN3 also has several potential phosphorylation sites, including ten serines and three threonines, although, it is not known whether these residues are phosphorylated or are involved in the function of the protein (Michalewski et al., 1999). Nevertheless, phosphorylation of CLN3 is evidient in vivo, as demonstrated in a Chinese hamster ovary cell line overexpressing GFP-CLN3, which was able to incorporate

51 radiolabelled phosphate (Michalewski et al., 1999). Incubating the cells with a variety of kinases and phosphatases was then used to help identify which of these catalysts may have been responsible for phosphorylation. These studies implicated cAMP- dependant kinase, cGMP-dependant kinase, and casein kinase II, with protein phosphatase 1 and 2a responsible for dephosphorylation (Michalewski et al., 1999).

Lipid modification of the CLN3 protein is largely unexplored, although a putative farnesylation motif, CQLS has been located within the C terminus (Pullarkat and Morris, 1997) and is suggested to provide a likely target for functional alteration of this protein (Haskell et al., 2000). Evidence for this process comes from a functional assay of yeast, which examined the effect of naturally occurring point mutations on intracellular localization of CLN3, and their ability to complement CLN3/Btn1p- deficient yeast (Btn1-∆, Haskell et al., 2000). In this study, CLN3 bearing a mutation in the farnesylation motif still trafficked normally, although this mutant protein was functionally compromised (Haskell et al., 2000). More precisely, btn1-∆ yeast were able to grow in the presence of a pH sensitive toxin, a property which was prevented by exposure to wild-type human CLN3, but partially restored with cells harbouring a mutation in the farnesylation motif of CLN3 (Haskell et al., 2000).

ii) Intracellular localisation of CLN3 CLN3 was initially thought to be associated with mitochondria, due to the major contribution of mitochondrial ATPase subunit c to the composition of storage material (Mitchison et al., 1999; Palmer et al., 2002; Cotman et al., 2002) and its presence in the mitochondria of mouse retina photoreceptors (Katz et al., 1997). More recently, localisation of this protein has been demonstrated in a variety of compartments within different cell types. In somatic cells this includes the Golgi (Haskell et al., 1999; Jarvela et al., 1999), ER (Jarvela et al., 1998), lysosomes (Golabek et al., 1999; Haskell et al., 2000; Kyttala et al., 2004), plasma membrane (Mao et al., 2003b) and endosomes (Haskell et al., 2000; Kyttala et al., 2004). In contrast, within neurons CLN3 is reported to be present within lysosomes, endosomes (Kyttala et al., 2004), synaptic vesicles (Haskell et al., 2000), nuclei, cytoplasm (Margraf et al., 1999) and synaptosomes (Luiro et al., 2001). It has thus been proposed that CLN3 traffics

52 through the ER to the Golgi plasma membrane, before residing in the lysosome, where it may subsequently be recycled through the plasma membrane to the endosome and lysosome (Jarvela et al, 1999; Haskell et al., 2000; Mao et al., 2003b). Despite the potential localisation of CLN3 to a variety of cell compartments, the consensus view is that this protein is specifically targeted to the endosomal/lysosomal system, but may also cycle through other pathways (Kyttala et al., 2005).

iii) Proposed functions for CLN3 The CLN3 protein is highly conserved in a variety of organisms, including humans, dogs, rats, mice, Drosophilia, Caenorhabditis elegans, and yeast, thus suggesting a fundamental role for this protein in cellular function (Mole, 2004; Phillips et al., 2005). Complex organisms express this protein in both neuronal and somatic cells (Jarvela et al., 1998; Golabek et al., 1999; Haskell et al., 1999; Jarvela et al., 1999; Margraf et al., 1999; Haskell et al., 2000; Luiro et al., 2001; Mao et al., 2003b; Kyttala et al., 2004). However, mutations in the CLN3 gene resulting in JNCL are thought to specifically target the CNS (International Batten Disease Consortium, 1995). Perhaps there is a pathway common to both somatic and neuronal cells, which may be of lesser importance or is compensated for in somatic cells. Alternatively, neuronal cells may be more sensitive to changes in this pathway, or may utilise this protein for additional cellular processes.

Insights into the precise function of CLN3 have been obtained from Btn1p, the orthologous protein in the yeast Saccharomyces cerevisiae, which shares 39% identity and 59% similarity to CLN3 (Pearce and Sherman, 1997; Pearce and Sherman 1998). When the corresponding gene, BTN1 was deleted, vacuolar pH was lowered during early growth, a property that was reversed by introducing the human CLN3 gene (Pearce and Sherman, 1998). Additional evidence for pH disruption has come from two separate studies using human JNCL cell lines, although lysosomal pH was increased instead of decreasing (Golabek et al., 2000; Holopainen et al., 2001). More recently, deletion of BTN1 in the same strain of yeast resulted in disruption to arginine transport, which was later restored with the introduction of human CLN3 (Kim et al., 2003). The transport or defective transport of small molecules across the

53 lysosomal/vacuolar membrane may be responsible for observed fluctuations in pH. Conversely, it may be altered pH that is causing the transport of arginine to be defective. Deletion of BTN1 also resulted in an upregulation of a novel BTN2 gene, encoding Btn2p (Pearce et al., 1999). This protein is homologous to Hook1, which is a microtubule binding protein capable of regulating endocytosis within Drosophila (Kramer et al., 1996). Potential interactions between CLN3 and Hook1 were assessed in mammalian cells overexpressing human CLN3. An aggregation of Hook1 protein was observed, which may have been mediated by dissociation from the microtubules (Luiro et al., 2004). In addition, receptor mediated endocytosis was found to be defective in CLN3-deficient JNCL fibroblasts, thus connecting CLN3, Hook1 and endocytosis in the mammalian system (Luiro et al., 2004).

iv) Mutations in CLN3 JNCL is caused by a mutation of the CLN3 gene located in the p12.1 region of chromosome 16 (International Batten Disease Consortium, 1995). So far thirty one mutations have been reported in this gene, including four small deletions, four small insertions, four large deletions, seven missense mutations, seven nonsense mutations, four mutations affecting splice sites and one intron change (Mole, 2004). Of these, the most common mutation is a 1.02 kb deletion that is seen in over 85% of JNCL alleles (International Batten Disease Consortium, 1995; Mole, 2004). However, some compound heterozygote JNCL patients are known to carry the 1.02 kb deletion on one allele and one of the less common mutations, deletions or insertions on the other allele (International Batten Disease Consortium, 1995).

v) Clinical progression of JNCL The clinical onset of JNCL occurs between four to seven years of age and is usually first manifested as an impairment of vision. This is followed by the slow development of mental retardation and motor dysfunction, which becomes more pronounced later in the disease process (Hofmann and Peltonen, 2001; Gardiner, 2002; Haltia, 2003; Goebel and Wisniewski, 2004b). Motor dysfunction may be attributable to extrapyramidal, slight pyramidal and cerebellar disturbances. Such abnormalities

54 manifest as Parkinson-like dysfunction and then lead to a nonambulatory state at a highly variable age (Hofmann and Peltonen, 2001; Gardiner, 2002; Haltia, 2003; Goebel and Wisniewski, 2004b). Epileptic seizures are another major symptom to accompany this disease and are typically evident at the age of ten, although these do not occur in all patients (Hofmann and Peltonen, 2001). JNCL patients usually die by the age of thirty, although a more protracted course of the disease has also been described (Wisniewski et al., 1998). There is very little quantitative data regarding the clinical progression of JNCL, but efforts are underway to provide scoring scales for judging a range of clinically relevant parameters (Kohlschutter, et al. 1988; Kwon et al., 2005). Such scales will be invaluable for providing a detailed picture of disease progression and for judging the efficacy of potential therapeutic agents in clinical trials.

vi) Potential Interactions of CLN Proteins It is unclear why the NCLs, which are a group of genetically heterogeneous disorders, display similar clinical features and biochemical and pathological profiles, albeit with markedly different ages of onset (Hoffman and Peltonen, 2001; Cooper, 2003). Such phenotypic similarities may either suggest a common biological pathway for the NCL proteins, or that these proteins may interact with one another. Despite this suggestion, attempts to demonstrate an NCL protein interaction have been unsuccessful using a yeast two-hybrid system (Zhong et al., 2000), although a more recent study used coimmunoprecipitation to suggest a potential CLN5 interaction with CLN2 and CLN3 (Vesa et al., 2002). In vitro binding assays provided further evidence for a CLN3/CLN5 interaction and also revealed that these two proteins may interact directly with one another and do not require assisting proteins (Vesa et al., 2002). CLN3 is a lysosomal membrane protein like CLN5 and has been implicated in vacuolar pH homeostasis in yeast (Pearce et al., 1999). Since both CLN3 and CLN5 patients display an increase in lysosomal pH (Halopainen et al., 2001), it may be possible for CLN5 to interact with CLN3 in a manner that regulates lysosomal pH within human cells (Vesa et al., 2002).

55 Although the proposed interactions between CLN proteins may be possible, the lack of reliable antibodies for these proteins suggest that caution must be taken when interpreting these immunoprecipitation studies. To provide supportive data for CLN protein interactions more specific antisera and other corroborating methodology must be used. An appropriately designed yeast two-hybrid model of protein interaction may potentially resolve this issue (Benedict et al., 2005; Yliannala et al., 2005), whereby individual loops of the CLN proteins may themselves be targeted. Detailed information of the interacting partners of each NCL protein is likely to improve our current understanding of disease mechanisms and lead to a better insight into NCL pathogenesis.

1.7 Animal models of NCLs

Detailed quantitative information about neuropathological changes in the NCLs is almost entirely lacking and the limited amount of data that is available is restricted to post-mortem material (Braak and Goebel, 1978; Braak and Goebel 1979). In contrast, animal models provide the opportunity to study progressive pathogenesis, which can be explored using a combination of naturally occurring and experimentally generated animal models. Thus far, a selection of mouse models is available for the six genetically identified forms of NCL. These include a Cln1 knock-out (PPT1-/-, Gupta et al., 2001) and a Cln1 knock-in (Ppt1∆ex4, Jalanko et al., 2005) for INCL; a Cln2 knock-out (TPP-I-/-) for LINCL (Sleat et al., 2004); two Cln3 knock-out (Cln3-/-, Mitchison et al., 1999; Katz et al., 1999) and a Cln3∆ex7/8 knock-in (Cotman et al., 2002) for JNCL; a Cln5 knock-out (Cln5-/-) for FvINCL (Kopra et al., 2004); and spontaneous mutants for CLN6 (nclf) (Bronson et al., 1998) and CLN8 (mnd) (Bronson et al 1993), which are variant forms of LINCL (Table 1.1). Collectively, these mice each exhibit NCL-like phenotypes, which includes the widespread accumulation of autofluorescent storage material in both somatic and neuronal cells, and varying degrees of neurodegeneration, which depends on the subtype of NCL that is modelled (Cooper, 2003; Mitchison et al., 2004).

56 Mouse model CNS Neuropathology Neurological behavioural disorder References

Gene and Design, strain and mutation Brain atrophy Reactive gliosis General Motor Seizures protein product Infantile NCL mouse models

CLN1/PPT1; Knock-out, Mixed C57BL/6J- Neocortical/subcortical Widespread Viable, fertile. Lack of grooming Progressive gait abnormality from 5 Frequent myoclonic Gupta et al., 2001; Palmitoyl 129/Sv. Neo gene plus in- atrophy. Selective astrocytosis, from 4-5 mo. Aggression and mo. Progressing to hind-limb seizures from 3-4 Bible et al., 2004 protein frame stop codon in exon 9 neocortical thinning and localized dermatitis. Survival 50% by 7 paralysis. Clasping phenotype from mo. Some tonic- thioesterase 1 laminar effects. microglial mo., all dead by 10 mo. 21 wks, 100% by 8 mo. clonic convulsions Now congenic on C57Bl/6 Pronounced interneuron activation loss

CLN1/PPT1; Knock-in, Mixed C57BL/6J- Marked neocortical Widespread Viable, fertile. Dead by 6-9 mo. Progressive motor abnormalities. Myoclonic jerks and Jalanko et al., 2005 Palmitoyl 129/Sv. Deletion of exon 4, atrophy. Loss of astrocytosis, Paralysis of hind limbs at 5 mo. seizures from 3-4 protein loxP site left in intron interneurons in several localized mo. thioesterase 1 CNS regions microglial activation

Late-infantile NCL mouse models

CLN2/TPP1; Knock-in, Mixed C57BL/6J- Degeneration in multiple Pronounced Viable, fertile. Dead by 3-6 mo. Tremors, abnormal gait, ataxia. Myoclonic jerks and Sleat et al., 2004; Tripeptidyl 129/Sv. Insertion point CNS regions astrocytosis seizures by 4 mo. peptidase I mutation Arg446His exon 11, LoxP site left on intron

CLN5; CLN5 Knock-out, Mixed C57BL/6J- Atrophy in several CNS Widespread Viable, fertile No seizures by 8 Kopra et al., 2004

57 129/Sv. Neo gene replaces regions, but not astrocytosis, months Hansen et al., 2005 exon 3 cerebellum at 12 mo. localized microglial activation

CLN8 (mnd); Spontaneous mutant Atrophy of the neocortex, Marked GFAP Viable, fertile. Increased activity Progressive motor deterioration from Seizure like Bronson et al., 1993; CLN8 C57BL/6.KB2/Rn substrain. but not cerebellum at 9 staining at 9 mo. and aggression. Learning and 4-5 mo. leading to ataxia, paralysis. behaviour at Bronson et al., 1998; c.267-268insC. Frameshift mo. Progressive loss of memory deficits. Dermatitis. Immobile by 9-12 mo. Rotarod and terminal stages Boyce et al., 1999; after Val89 neocortical and Dead by 10-12 mo. various other deficits, abnormal Cooper et al., 1999; hippocampal interneurons clasping from 5 mo. Ranta et al., 1999; Bolivar et al., 2002

CLN6 (nclf); Spontaneous mutant, Mixed Neocortical atrophy. Marked GFAP Viable, fertile. Clinically normal From 8 mo. develop spastic rear limb Death of a few mice Bronson et al., 1998; CLN6 C57BL/6J/10J, Progressive loss of staining at 6 mo. until 8 mo. Dead by 12 mo. paresis progressing to paralysis, by terminal seizures Gao et al., 2002; C3HeB/FeJLe. c.307insC neocortical and typical of upper motor neuron reported Wheeler et al., 2002 frameshift after Pro120 hippocampal interneurons degeneration. Abnormal clasping

Table 1.1 Mouse models of NCL. (Adapted from Mitchison et al., 2004)

Mouse model CNS Neuropathology Neurological behavioural disorder References

Gene and Design, strain and mutation Brain atrophy Reactive gliosis General Motor Seizures protein product

Juvenile NCL mouse models

CLN3; CLN3 Knock-out, Mixed C57BL/6J- Delayed neocortical Astrocyte/microglia Viable, fertile. Autoimmune Motor deterioration seen by 16 mo. with Altered threshold Mitchison et al., 1999; 129/Sv. Neo gene replaces exon atrophy and thinning. markers upregulated response. Dermatitis from 16 mo. reduced activity, stiff and slow gait for seizure Pontikis et al., 2004 1-6 Progressive loss of pre-symptoms with irritation and scratching. Dead generation neocortical and by 18-20 mo. Now available on 129/Sv and hippocampal C57BL/6 backgrounds interneurons

CLN3; CLN3 Knock-out, Mixed C57BL/6J- Viable, fertile. Lower mating By 25 wks stiff and slow gait No seizures up to 25 Katz et al., 1999 129/Sv. Neo gene replaces exon success in knockouts. Some wks 7-8 mutants show circling behaviour 58 from 8 wks

CLN3; CLN3 Knock-in, Mixed 129/Sv-CD1. Neocortical thinning. Marked GFAP staining Viable, fertile. Reduced lifespan Abnormal gait at 10-12 mo. 42% show Seizures not Cotman et al., 2002; Deletion of exon 7-8, loxP site Loss of neocortical and in brain at 10 and 12 from 7 mo., 80% survival at 10-12 clasping phenotype at 10-12 mo. prominent Pontikis et al., 2005 left in intron 6 thalamic relay neurons mo. More subtle mo. at 12 mo. microglial response Now available on C57Bl/6

Table 1.1 cont. Mouse models of NCL. (Adapted from Mitchison et al., 2004)

The PPT1 null mutant mouse generated by Gupta et al. (2001) displays motor abnormalities, spasticity, myoclonic jerking and seizures, leading to premature death by the age of 10 months (Gupta et al., 2001). Pathological changes within the CNS also include neocortical and subcortical atrophy with disruption of individual laminae, loss of neocortical and hippocampal interneurons and widespread astrocytosis, which is accompanied by localized microglial activation (Bible et al., 2004). More recently, a Ppt1∆ex4 mouse has been generated by targeted deletion of exon 4 of the mouse Cln1 gene (Jalanko et al., 2005). This model displays similar neuropathological abnormalities to the PPT1 null mutant mouse, including an abundance of autofluorescent storage material, a regional specific interneuron loss and a reduction of brain mass, but death occurs at an earlier age than in PPT1-/- mice (Jalanko et al., 2005). Further characterisation of these mice will be required in order to determine the precise sequence of events during disease progression and provide an in depth comparison of clinical phenotypes.

A mouse model of LINCL was recently generated via targeted disruption of the Cln2 gene, resulting in loss of TPP-I activity (Sleat et al., 2004). Similar to PPT1, TPP-I is a soluble lysosomal enzyme, and TPP-I dysfunction is typically associated with a more protracted course of disease (Hofmann and Peltonen, 2001; Goebel, 2004). Surprisingly, the lifespan of TTP-I null mutant mice is extremely short, with the median survival of approximately four and a half months of age. This is accompanied by tremors, ataxia, and extensive neuronal pathology, including loss of cerebellar Purkinje cells, widespread axonal degeneration and the progressive accumulation of autofluorescent storage material within the lysosomal-endosomal compartment (Sleat et al., 2004).

The presence of Cln5 null mutants (Kopra et al., 2004) and naturally occurring nclf (Bronson et al., 1998) and mnd mice (Bronson et al., 1993) have provided us with useful tools for gaining insight into neuropathological changes within different variant forms of LINCL. In Cln5 null mutant mice (Cln5-/-) these events include a progressive visual impairment from three months of age and the widespread accumulation of autofluorescent storage material in the CNS and somatic cells without prominent brain atrophy (Kopra et al., 2004). This is accompanied by a profound loss of interneurons and the altered expression of various genes associated with

59 neurodegeneration and immune response (Kopra et al., 2004). Nclf and mnd mice harbour frameshift truncated mutations, which abolish the function of genes orthologous to CLN6 and CLN8 respectively (Gao et al., 2002; Ranta et al., 1999). Extensive characterization of mnd mice has revealed severe brain atrophy, loss of interneuron populations and retinal cells, motor and cognitive deficits (Cooper et al., 1999; Bolivar et al., 2002). These mice also exhibit a series of complex changes in the expression of glutamate receptor expression, which may be compensatory in response to elevated glutamate levels, as demonstrated by metabolic profiling (Griffin et al., 2002). Nclf mice display a similar neuropathological profile to that of mnd mice, with a similar deposition rate of autofluorescent storage material, retinal atrophy, motor abnormalities, muscle weakness and premature death (Bronson et al., 1998).

Larger animal models of NCL may represent a closer approximation of the human disease phenotype than mouse models, because of their similarly sized and more complex CNS that more closely resembles the human brain. These large animal models include naturally occurring canine (Koppang, 1988; March et al., 1995; Taylor et al., 1993; Katz et al., 2005; Melville et al., 2005), ovine (Jolly et al., 1976; Broom and Zhou, 2001; Tammen et al., 2001), equine (Url et al., 2001), and bovine models of NCL (Harper et al., 1988; Mertinus et al., 1991; Tammen et al., 2005). Despite exhibiting NCL-like phenotypes, the underlying gene defects in many of these larger animals remain largely unknown, although there has been a recent flurry of papers identifying mutations in canine and bovine models of NCL (Katz et al., 2005; Mellvile et al., 2005; Tamen et al., 2005). Of all these models, the New Zealand South Hampshire sheep is best characterized, having a gene defect located on chromosome 7q13-15, which is analogous to human chromosome 15q21-23 and therefore corresponds to human CLN6 (Jolly et al., 1976; Broom and Zhou, 2001). These sheep display widespread accumulation of autofluorescent storage material, significant loss of GABAergic interneurons in the neocortex and cerebellum and exhibit severe gliosis early in development when the sheep are presymptomatic (Oswald et al., 2001; Oswald et al., 2005).

60 1.8 Mouse models of JNCL a) Cln3 null mutant mice (Cln3-/-)

The murine Cln3 gene shares 82% sequence identity and 85% amino acid similarity with human CLN3 (Lee et al., 1996). Cln3-/- mice were generated via targeted disruption of the Cln3 gene and as expected exhibit a JNCL-like phenotype (Katz et al., 1999; Mitchison et al, 1999). In one model, disruption to this gene was achieved via the by replacement of a large portion of exon 7 at the beginning of intron 8, using a neo cassette (Katz et al., 1999). The other model involved deletion of exons 2-6 and most of exon 1 via replacement with a neomycin resistance gene that was transcribed in reverse orientation from a mouse PGK promoter (Mitchison et al., 1999). These animals display widespread intracellular accumulation of autoflourescent pigment, behavioural abnormalities, and premature death (Katz et al., 1999; Mitchison et al, 1999). Further characterization of the Cln3-/- mice has revealed an altered threshold for seizure generation coupled with the selective loss of neocortical interneuron populations (Mitchison et al., 1999; Kriscenski-Perry et al., 2002). Similar to individuals with JNCL, these mice also raise autoantibodies to glutamic acid decarboxylase (GAD65) that inhibit the activity of this enzyme, resulting in elevated levels of glutamate at the expense of depleted levels of interneuronal transmitter GABA (Chattopadhyay et al,. 2002a).

b) Cln3 ‘knock-in’ mice (Cln3∆ex7/8)

Cln3 null mutant mice do not make any functional Cln3 protein, rather than bearing the 1.02 kb deletion in the Cln3 gene that is seen in over 85% of JNCL alleles (International Batten Disease Consortium, 1995). More recently, a Cln3∆ex7/8 mouse has become available which accurately reproduces this common mutation of JNCL patients (Cotman et al., 2002). This model was generated by introducing or ‘knocking-in’ an identical 1.02kb genomic DNA deletion, removing exons 7 and 8 and the surrounding non-coding DNA (∆ex7/8 allele) (Cotman et al., 2002). These mice are phenotypically more aggressive than Cln3 null mutants, with a progressive

61 neurologic disease and premature death of homozygous Cln3∆ex7/8 mice from 7 months onwards (Cotman et al., 2002). Widespread deposition of ATPase subunit c is present throughout the CNS as early as embryonic day 19.5 (E19.5). This precedes extensive accumulation of autofluorescent storage material by ten months of age, which is most prominent in the hippocampus, cerebellum, neocortex, thalamus and liver (Cotman et al., 2002). A reduction of photoreceptors in the retina and severe reactive gliosis in motor cortex, hippocampus, cerebellum and midbrain is also evident during this period (Cotman et al., 2002).

1.9 Neuronal loss in NCLs

The loss of selective neuronal populations is a consistent feature of many neurodegenerative disorders including the NCLs (Mitchison et al., 2004). Previous descriptions of the neocortex of human NCL patients have outlined the loss of small stellate neurons and inhibitory synapses, which are suggested to be GABAergic (Braak and Goebel, 1978; Braak and Goebel, 1979). It is now widely accepted that interneurons are affected in the NCLs, as documented in various animal models of the disease and post-mortem human material (Braak and Goebel, 1978; Braak and Goebel, 1979; Bible et al., 2004; Pontikis et al., 2004; Tynellä et al., 2004; Oswald et al., 2005). Such changes vary tremendously in different subpopulations of these cells, with the relative sparing of calretinin-positive inteneurons (D’Orlando et al., 2002; Cooper, 2003; Bible et al., 2004; Tynellä et al., 2004).

As mentioned briefly, an autoantibody to glutamic acid decarboxylase (GAD65) has recently been identified in Cln3 null mutants (Chattopadhyay et al., 2002a) and in JNCL patients (Chattopadhyay et al., 2002b; Chattopadhyay et al., 2002b). This autoantibody is capable of inhibiting the activity of GAD65, which is an essential enzyme for the synthesis of GABA from glutamate. Therefore, the presence of anti- GAD65 could potentially explain the targeting of GABAergic interneurons in both JNCL patients and Cln3-/- mice. Nevertheless, autoantibodies to GAD65 are absent from all other forms of NCL (Chattopadhyay et al., 2002b), despite the severe loss GABAergic interneurons seen in these disorders (Cooper et al., 1999; Bible et al.,

62 2004; Tyynelä et al., 2004; Oswald et al., 2005). Thus, it is likely that GAD65 autoantibodies to only play a minor contributory role, if any, in the loss of these interneuron populations.

In addition to the loss of GABAergic interneurons, pyramidal cells are also affected in the CNS of the NCL patients (Tyynelä et al., 2004) and in NCL mouse models (Mitchison et al., 2004; Bible et al., 2004). This neuronal loss has been described within the hippocampal formation, where earlier onset forms INCL and LINCL display a complete breakdown of hippocampal architecture (Tyynelä et al., 2004). In cases of JNCL, neuronal loss is most pronounced in the hilus with decreasing severity moving through the CA3, CA2 and relatively sparing of neurons in CA1 (Tyynelä et al., 2004). These patterns of neuronal loss do not correlate with the extent and distribution of autofluorescent storage material deposition, thus suggesting storage material not to be directly causative of damage to these cells (Tyynelä et al., 2004).

1.10 Mechanisms of cell death in NCLs

There has been much debate over the precise cause of neuronal death in the NCLs. For instance, the filling of cells with characteristic autofluorescent storage material causes severe hypertrophy and may possibly be damaging to neurons. However, the extent of this deposition and neuronal loss do not appear to be linked and these may be independent events (Tyynelä et al., 2004; Oswald et al., 2005). Whatever the underlying causes, apoptotic neuron death has been suggested in the NCLs (Lane et al., 1996; Seigel et al., 2002; Koike et al., 2003) and in a variety of other neurodegenerative disorders (Offen et al., 2000; Pompl et al., 2003). However, it cannot be excluded that multiple mechanisms of cell death may operate in such diseases, and may be extremely context dependent (Graeber and Moran, 2002).

Apoptotic cell death has been reported in the retina of Cln3 null mutant and Cln3∆ex7/8 mouse models of JNCL (Cotman et al., 2002; Seigel et al., 2002) and in cathepsin D null mutant mice, which also exhibit an NCL-like phenotype (Koike et al., 2003). By using the terminal dUDP nick end-labelling (TUNEL) staining method, apoptotic cell

63 death has also been observed in the neurons of INCL and JNCL patients and in naturally occurring canine and ovine models of NCL (Lane et al., 1996). CLN3 has recently been proposed to have a role in the prevention of apoptosis, as the deletion of the CLN3 gene in immortalized JNCL lymphoblasts may cause these cells to grow at a slower rate and display increased sensitivity to etopside-induced apoptosis (Persaud- Sawinet al., 2002).This function has been localized to domains clustered around the C-terminus and involves deletion mutants in highly conserved regions of the CLN3 gene (Persaud-Sawin et al., 2002). Despite this finding, there is no direct data to support an apoptotic mechanism of cell death in the CNS of Cln3-deficient or nclf mouse models of NCL (Bronson et al., 1998; Cotman et al., 2002, Mitchison et al., 2004). Indeed, ultrastructural characteristics of cell death in Cln3 null mutant mice are more suggestive of an autophagic mechanism, including the presence of double- membrane autophagosomes, which are intracellular vacuoles containing cytoplasmic components (Mitchison et al., 2004 Yuan et al., 2003). This process is responsible for the bulk degradation of dysfunctional or damaged components within the cytoplasm, or intracellular organelles within several cell types (Klionsky and Emr, 2000). Autophagic cell death can either accompany apoptosis or one process can precede or trigger the other (Xue et al., 1999). It is therefore unclear whether these processes act alone or in combination in the different forms of NCL, and these events are currently under investigation.

1.11 Glial responses in NCLs

All forms of NCL exhibit glial cell activation at the end stages of disease (Haltia et al., 1973a; Haltia et al., 1973b; Goebel et al., 1999; Haltia, 2003), although more recently reactive changes have also been reported earlier in pathogenesis (Oswald et al., 2005). Reactive gliosis is also common to several other degenerative disorders of the CNS, including multiple sclerosis (Windhagen et al., 1996), Alzheimer disease (McGeer et al., 1998) and Parkinson disease (Youdim et al 1994). Typically astrocytosis and microglial activation may serve as an early indication to areas that are severely affected by disease (Gehrmann et al., 1995; Kreutzberg et al., 1996; Raivich et al., 1999). Within the NCLs, reactive gliosis is regionally selective for both

64 microglial activation and astrocytosis, as derived from patient autopsy material (Tyynelä et al., 2004). This study revealed microglial activation to be associated with regions of greatest neuronal loss, whereas astrocytosis was present within regions where neuronal loss was less evident (Tyynelä et al., 2004).

a) The role of microglia in the NCLs

Microglia are complex immune cells of the CNS thought to contribute to both chronic neurodegeneration and oxidative stress (Gebicke-Haerter, 2001). These cells are capable of producing a variety of cytoactive factors, as well as reactive oxygen and nitrogen species, which include nitric oxide (NO), superoxide and glutamate (Koutsilieri et al., 1999), a transmitter that is excitotoxic at higher concentrations. The pathological involvement of NO has been explored in cathepsin D null mutant mice that exhibit microglial activation in both the neocortex and in the thalamus (Nakanishi et al., 2001). These morphologically transformed microglia markedly express inducible nitric oxide synthase (iNOS), which was blocked with an iNOS inhibitor (Nakanishi et al., 2001). Subsequent TUNEL staining revealed this treatment to significantly reduce the number of apoptotic cells within the thalamus, thus suggesting NO production via iNOS to be damaging to tissue (Nakanishi et al., 2001). Indeed, as microglia express iNOS when they become phagocytic, it is not uncommon for cell death to occur in the local area surrounding such phagocytic activity (Koutsilieri et al., 2002).

The actual role of microglia is proposed to range from homeostatic regulation to the repair of damaged cells (Kreutzberg, 1996). Evidence for this comes from the immediate response of microglia to physiological stress, their ability to secrete neurotrophic factors and cytokines, their increase in mobility as they migrate towards damaged cells, their transformation into phagocytes when neuronal damage occurs and their clearance of dying cells when repair is no longer possible (Kreutzberg, 1996; Heppner et al., 1998; Hirt et al., 2000; Liu and Hong, 2003). The concept of microglial mediated neurodegeneration vs. neuroprotection has been extensively investigated in the facial nerve axotomy paradigm (Guntinas-Lichius et al., 1994;

65 Schiefer et al., 1999; Kella et al., 2001; Hurly and Coleman, 2003; Moran and Graeber, 2004). Following neuronal injury, resident microglia are activated and increase in number within the facial nucleus (Graeber et al., 1988a; Graeber et al., 1988b). Some of these cells maintain close cell to cell contact with motor neurons and move along proximal parts of neuronal dendrites (Schiefer et al., 1999). Activated microglia that cover the cell bodies of chromatolytic motorneurons may subsequently cause deafferentation of these neurons in a process referred to as synaptic stripping (Graeber and Kreutzburg, 1988). This synaptic stripping may offer neuroprotection by removing excitatory input in order to allow for neuronal recovery (Blinzinger and Kreutzburg, 1968), whereas perineuronal ensheathment may facilitate targeted delivery of growth factors from activated microglia to injured neurons (Batchelor et al., 2002). Further evidence for a neuroprotective or regenerative role of microglia come from studies of spinal cord injury (Rabchevsky and Streit, 1997). The engraftment of cultured microglial cells into the injured spinal cord is capable of promoting neurite growth into such microglial grafts (Rabchevsky and Streit, 1997). Again, the release of trophic factors from microglial populations could potentially aid in the recovery of these damaged neurons (Rabchevsky and Streit, 1997; Batchelor et al., 2002).

b) The role of astrocytes in the NCLs

There is an abundance of astrocytes within the brain, which together with their numerous leaflet like processes, cover most synapses within the CNS and may constitute as much as one third of total brain mass (Kandel et al., 1991). These resident immune cells of the CNS are highly complex and have a diverse spectrum of proposed roles, including the optimization of interstitial space for synaptic transmission by controlling water and ionic homeostasis, regulation of extracellular glutamate concentrations, and modulation of synaptic transmission (Simard and Nedergaard, 2004).

Glutamate may be of particular importance for interactions between astrocytes and neurons. Clearance of extracellular glutamate occurs via high affinity glutamate

66 transporters that are enriched in astrocytic processes (Rothstein et al., 1996; Duan et al., 1999). In response to stimuli, both GLT1 and GLAST1 subtypes modulate via a slow regulatory mechanism or via a rapid process taking only minutes (Davis et al., 1998; Duan et al., 1999). Glutamate is then transported into astrocytes in an electrogenic manner before its conversion into glutamine by an astrocyte specific enzyme glutamine synthase (Hallermayer et al., 1981; Robinson and Dowd, 1997). Subsequent to its release into the extracellular space, glutamine can either be used as a fuel for neurons or be recycled back into glutamate for modulation of neuronal transmission (Simard and Nedergaard, 2004). Alternatively, metabotropic receptors located on astrocytes are activated during neuronal activity (Poitry-Yamate et al., 2002), which leads to an increased turnover of inositol triphosphate and the release of intracellular stores of Ca2+, independently of extracellular stores (Cornel-Bell et al., 1990). Such changes in [Ca2+] causes the Ca2+-dependant release of glutamate from astrocytes and modulation of synaptic activity (Haydon, 2001).

The hormonal regulation of water and ion homeostasis within the CNS is governed by centrally released neuropeptides such as vasopressin (AVP), atriopeptin (ANP), angiotensinogen (AGT) and angiotensin (Ang) II, all of which influence osmoregulation and the ionic environment within astrocytes (Simard and Nedergaard, 2004). Of these peptides, AVP and ANP have opposing effects on glial volume, whereby, AVP increases the water content of glial cells and ANP decreases it (Latzkovits et al., 1993). Additionally, ANP restricts chloride (Cl-) uptake in a dose dependent manner, an effect which is abolished by AVP in cultured astrocytes (Katay et al., 1998). A distinct astrocytic perivascular system also appears to be in place within the CNS (Simard and Nedergaard, 2004). This system has several potassium channels (Dietzelk et al., 1989; Ordaz et al., 2004) and aquaporin-4 (AQP-4), which is a tetrameric water channel that is localized within astrocytic endfeet and as such is important for osmoregulation within the brain (Nielson et al., 1997; Amiry- Moghaddam et al., 2003; Simard et al., 2003; Simard and Nedergaard, 2004).

Taken together, the emerging intricacies of astrocytic function suggest an essential role within the CNS that ultimately governs a homeostatic balance between neurons and their external environment (Simard and Nedergaard, 2004). This role may be housekeeping under normal circumstances, although it also possible that during

67 disease progression, astrocytic function may be affected and therefore contribute to disease phenotype within the NCLs.

1.12 Studies in this thesis

Aims and objectives

The neuronal ceroid lipofuscinosis (NCLs) are a group of fatal neurological disorders which primarily affect children (Cooper, 2003). Pathologically these disorders are incompletely characterised and the limited amount of data that is available is restricted to post mortem tissue. However, the recent development of mouse models of JNCL (Mitchison et al, 1999; Katz et al., 2001; Cotman et al., 2002) makes it possible to identify progressive pathological changes within the CNS of these disorders. Detailed characterisation of Cln3-/- and Cln3∆ex7/8 mice will allow identification of specific subpopulations of neurons or glia, which are affected at different stages of disease. These studies will provide clear pathological landmarks of disease progression in the Cln3-/- and Cln3∆ex7/8 mice. Since therapeutic interventions are likely to be most effective in the early stages of disease, it will be particularly important to identify when the CNS is first impacted by this disease.

The aims of this research are to a) plot progression of pathological changes during disease progression and b) define the nature and timing of pathological effects upon the CNS in these two major models of JNCL. To achieve this goal, unbiased stereological methodology and immunohistochemical analysis shall be used to characterize the extent of neuronal loss and glial activation during postnatal development in null mutant and knock in mouse models of JNCL (Cln3-/-, Cln3∆ex7/8 mice).

68 1.13 Summary of background

The NCLs are the most common group of inherited neurodegenerative disorders of childhood with an incidence of up to 1 in 12,500 live births (Goebel et al., 1995). These lysosomal disorders are characterised by widespread intralysosomal accumulation of autofluorescent material, blindness, marked psychomotor deterioration and uncontrollable seizures and ultimately premature death (Hofmann and Peltonen, 2001; Cooper, 2003). However, each form of NCL is caused by mutations in an individual ‘CLN’ gene and results in different ages of onset and rates of progression (Hofmann and Peltonen, 2001).

Detailed information about progressive neurodegenerative changes in the NCLs is almost entirely lacking, although it has always been assumed that the development of the CNS in affected individuals is normal before the effects of disease subsequently become evident. However, recent data has shown significant changes in gene expression (Chattopadhyay et al, 2002a; Elshatory et al., 2003) and reactive changes in populations of glia (Pontikis et al., 2004) in JNCL, which occur significantly earlier than the onset of clinical symptoms or neuronal loss.

Cloning of the CLN3 gene has led to the production of genetically engineered null mutant mouse models (Mitchison et al, 1999; Katz et al, 1999), which are generated via targeted disruption of Cln3 and exhibit a JNCL phenotype. These animals display a lower threshold for seizure generation, a widespread intracellular accumulation of autoflourescent pigment and loss of GABAergic interneurons (Mitchison et al, 1999; Pontikis et al., 2004). More recently, a homozygous Cln3∆ex7/8 mouse has become available which accurately reproduces the common mutation in JNCL patients and also exhibits an NCL phenotype (Cotman et al., 2002).

1.14 Hypotheses tested

We hypothesise that we can detect pathological changes at different stages of disease pathogenesis by measuring neuronal morphology at a regional and cellular level and

69 examining resident glial and neuronal populations. These data will be essential in determining when the CNS is first detected and for identifying the cellular components that are affected in this disorder.

70

CHAPTER 2

Materials and Methods

71 2.1 Introduction

The studies in this thesis survey the progression of a range of pathological phenotypes looking at the early postnatal period and subsequently at later stages of disease progression in two distinct mouse models of JNCL. The core methodology used for this analysis was unbiased, design-based stereological measurements of regional volume, as well as number and volume of selected neuronal populations. Such unbiased estimations represent a more accurate and efficient methodology than semi- quantitative analysis, which is often subject to observer bias.

2.2 Stereology a) Design-based stereology

Stereology can be used to quantify any range of parameters within an area of interest. In the context of the brain, this sampling strategy efficiently estimates regional volumes, cell numbers, mean cellular volumes and length densities of biological structures (Schmitz and Hof, 2005). Traditionally, such methods were used to produce three-dimensional information based upon two-dimensional sections (Weibel, 1979, 1980). However, recent design-based methods enable similar information to be obtained from thicker three-dimensional sections. The use of design-based stereological methods implies that the probes and sampling scheme used are designed so that size, shape, spatial orientation, and spatial distribution need not be considered (West, 2002). Thus, systemic errors in calculations are eliminated and these approaches provide a more reliable means of data collection than qualitative efforts, which are compromised by observer bias (Gundersen et al., 1988; West, 1993, 2002). As such, design-based stereological methods enable valid comparisons between different histological studies within a chosen condition or disorder, with no assumptions made about the size or shape of brain structures as a result of disease.

Stereological analysis requires a systematic uniform random sampling of a structure in order to obtain an accurate estimate of the parameter that is being analysed, for

72 example the number of neurons or their size. The starting location, where data is first gathered, is randomly determined and subsequent fields sampled in a systematic and uniform manner. To guarantee that there is an equal chance for sampling of all parts of a structure (Gundersen and Jensen, 1987), this sampling method must be applied to a series in which the entire selected region is present (Gundersen, 1986; Gundersen and Jensen, 1987).

Although design-based stereology can limit biasedness of corresponding estimates (Guillery and Herrup, 1997), there are several other factors that must be considered. The precise protocols that have been used to process the selected tissue must be kept constant between studies, as minor variations may lead to the sample being incomparable. Such variations can either be attributable to fixation, embedding, cutting or the staining of a given tissue. Differences in fixation and embedding protocols can result in tissue shrinkage to different extents (Bauchot, 1967; Dorph- Petersen et al., 2001). Similarly, during the cutting procedure, shrinkage along the plane of the section may be variable (West et al., 1991; Dorph-Peterson et al., 2001). It should also be mentioned that unless adequate precautions are taken when using immunohistochemical or Nissl methods (Cooper et al., 1988), neurons may be inadequately stained, thus leading to underestimation of neuronal number.

b) Stereological estimation of volume

An unbiased estimation of regional volume can be determined according to the Cavalieri principle (Cavalieri, 1665). This volume is calculated by multiplying the total area of the region of interest in all sections, by the distance between consecutive sections (Gundersen and Jensen, 1987). To measure these profile areas, the number of points from a grid overlay that cover the chosen region are multiplied by the area between them (Gunderson and Jensen, 1987). Overprojection of estimated volume is always a possibility, due to bias selection of sampling sites and the relative opacity between the region of interest and surrounding matrix in thicker sections (Howard and Reed, 1998a). Thus such factors must be considered during experimental design. Nevertheless, this can be almost entirely eliminated by cutting the region of interest

73 into an exhaustive series of sections, all of uniform thickness and then selecting a systematic random series with a random starting point for analysis (Schmitz and Hof, 2005). The volume of a given structure can be estimated by applying the following formula

m a vTˆ = ..∑ Pi p i=1

where Pi is the number of points counted in each section, a/p is the area associated with each point, m is section periodicity, and t is the mean section thickness (Cavalieri, 1665; Gundersen and Jensen, 1987).

c) Stereological estimation of cell number

It is not usually feasible to count every neuron in a given structure. Instead, a ‘fractionator’ can be used to make an unbiased estimation of total neuronal number based on only a small sample of the region of interest (Schmitz and Hof, 2005). This sampling is conducted by selecting an appropriate fraction of microscopic fields, counting the number of neurons within each sampling frame that resides within the region of interest and then applying a sampling probability. Before neurons can be counted, unbiased counting spaces referred to as ‘optical dissectors’ must be applied to each of the microscopic fields (West, 1993). These must be placed in a systematic and random manner, in a series of systematically randomised sections through the structure. The design-based stereological analysis of total neuronal number can then be calculated by multiplying the total number of neurons within all unbiased virtual counting spaces (Nv) by the reciprocal value of the sampling probability (West et al., 1991). Sampling probability is influenced by the number of analyzed sections vs. total number of sections (Sampling fraction, ssf), the area of unbiased virtual counting frames versus the profile area of a given structure within each section (Area sampling fraction, asf), and the unbiased height of the virtual counting frame relative to the

74 mean thickness of sections after histological processing of the tissue (height sampling fraction, hsf) (Schmitz and Hof, 2005).

Figure 2.1 Unbiased counting frame within a virtual space of a Nissl stained tissue section. (Image taken with permission from www.microbrightfield.com).

The ‘optical fractionator’ is the combination of the ‘fractionator’ with ‘optical dissector’ (Antunes, 1995; West et al., 1991) and may offer greater accuracy than non-stereological methods, due to the use of an unbiased counting frame within a virtual space in section thickness (Figure 2.1). This counting frame has a known area, with an ‘acceptance’ line (green) and a ‘forbidden’ line (red) of which certain rules apply. Cells that are cut by the ‘forbidden’ line are not counted, and cells that either fall within the box or are crossed by the ‘acceptance’ line are counted unless they are also cut by the ‘forbidden’ line. This set of rules must be used in conjunction with set criteria for identifying these neurons, such as correct morphology, size and the presence of a nucleus or nucleolus. Taken together, an unbiased count can be made within a defined unit area (Howard and Reed, 1998b). The total number of cells in a given structure can be estimated by applying the following formula

111 NQˆ = ... hsf asf ssf where Q is mean of the total number of cells counted in any of the counting frames within any section, hsf is the height sampling fraction, asf is the area sampling fraction and ssf is the section sampling fraction (West, 1991; Howard and Reed, 1998b).

75 d) Stereological estimation of mean cellular volume

The design-based stereological estimation of number-weighted mean particle volume may initially require ‘optical dissector’/ ‘optical fractionator’ method to be used for particle selection. A stereological probe such as ‘nucleator’ can then be selected to sample particle size in a specific cell population within a given structure (Gunderson, 1988) (Figure 2.2). In accordance to previously outlined criteria (see above), cell populations within counting frames can be isolated and random points chosen within a definable subspace of each cell (i.e. nucleus or nucleolus). From these points, an isotropic direction is generated and the distances in each direction out from the point to the boundary of the cell wall are recorded (Howard and Reed, 1998c). To increase the accuracy of the estimator, measurements can be made in multiple systematic random directions from each sampling point.

Figure 2.2 Representative image of the Nucleator estimation of neuronal size within an unbiased counting of a Nissl stained tissue section (Image taken with permission from www.microbrightfield.com).

The mean cell volume in the number-weighted distribution can then be estimated from these measurements by applying the following formula

n ˆ ππ33 vlvi=⋅00, = ⋅∑ l 33⋅n i=1

76 where n is the total number of point sampled linear intercepts, l is the length of the

l3 intercepts and 0,i is the cube of the ith point-sampled intercept length (Gundersen and Jensen, 1985; Gundersen, 1988).

e) Use of stereology in the JNCL CNS

Unbiased, design-based stereological methodology was applied to both Nissl (2.5) and immunostained (2.6) tissue sections. Tissue was obtained from two separate mouse models of JNCL that use different targeting strategies to produce either a -/- dysfunctional Cln3 gene (Cln3 , Mitchison et al., 1999), or to reproduce the most common 1kb CLN3 deletion mutation in human JNCL (Cln3∆ex7/8, Cotman et al., 2002).

2.3 Mouse models of JNCL a) Cln3 null mutant mouse model of JNCL i) Summary of targeting strategy Cln3-/- mice were originally generated by homologous recombination, as described previously (Mitchison et al., 1999). In order to inactivate Cln3, exons 2-6 and a large portion of exon 1, including the start codon, were replaced by a neomycin resistance gene that was transcribed in reverse order from a mouse PGK promoter. These cells were used to establish chimeras of 129/Sv and Black Swiss outbred lines, and heterozygous offspring were inbred to generate Cln3-/- homozygotes. Brain and kidney samples from Cln3-/- homozygotes were subsequently genotyped by RT-PCR, revealing the absence of Cln3 exons 1 and 2 and expression of the neo cassette (Mitchison et al., 1999). These mice were then backcrossed to make a congenic 129Sv congenic line that was used in this thesis.

77 ii) Animals in study Cln3-/- mice and control (+/+) littermates on the 129S6/SvEv background were maintained at the University of Rochester School of Medicine, as described previously (Pontikis et al., 2004). Mice of both sexes were used for this analysis since previous studies have revealed no significant difference in NCL-like phenotype between male and female mice (Mitchison et al., 1999). Callosal dysgenesis is well documented in mice on the 129S6/SvEv background (Wahlsten, 1982), and occurred with equal frequency in either Cln3-/- and control mice used in this study. To account for this, equivalent numbers of acallosal mice of either genotype were compared at each age.

b) Cln3∆ex7/8 ‘knock-in’ mouse model of JNCL i) Summary of targeting strategy Cln3 ‘knock-in’ (Cln3∆ex7/8) mice were originally generated by homologous recombination (Cotman et al., 2002). These mice reproduce the most common 1kb CLN3 deletion mutation in human JNCL (Cotman et al., 2002), with the 1kb deletion including exons 7 and 8 and the surrounding non-coding DNA. This deletion mutation was introduced into embryonic stem cells by replacing the 1kb segment with a ‘floxed’ PGneo cassette and was then used to generate Cln3∆ex7/8neo mice. Matings with hCMV-cre mice, expressing Cre recombinase, produced offspring with a deleted floxed neo cassette. In order to delete the hCMV-cre transgene, mice were intercrossed and cre-negative offspring were then bred together. These mice bear the Cln3∆ex7/8 allele present in 85% of JNCL alleles and were maintained on a 129Sv/Sv/CD1 background. Subsequent analysis of homozygous Cln3∆ex7/8 mice revealed alternatively spliced mRNAs, including a variant predicting non-truncated protein, as well as mutant Cln3∆ex7/8 (Cotman et al., 2002).

78 ii) Animals in study Homozygous Cln3∆ex7/8 mice and control (+/+) mice were maintained at the Molecular Neurogenesis Unit, Massachusetts General Hospital, Charlestown, MA (Cotman et al., 2002 ). All mice used in this study were littermates produced by crossing heterozygous Cln3∆ex7/8 mice that were from an F4 CD1 backcross generation. Mice of both sexes were used for this analysis since we have previously observed no significant difference in JNCL-like phenotype between male and female Cln3∆ex7/8 mice (Cotman et al., 2002). All perfusion procedures were carried out in accordance with NIH guidelines and under the animal care committee regulations at the Massachusetts General Hospital.

Mouse model Age Sample number (n) 18 months 3 +/+, 3 Cln3-/- Cln3-/- mice 14 months 3 +/+, 3 Cln3-/- 5 months 3 +/+, 3 Cln3-/- P8 3 +/+, 3 Cln3-/- 12 months 5 +/+, 5 Cln3∆ex7/8 Cln3∆ex7/8 mice P7 7 +/+, 5 Cln3∆ex7/8

Table 2.1 Details of Cln3-/- mice, homozygous Cln3∆ex7/8 mice and age-matched control (+/+) littermates

2.4 Histological processing

For histological analysis, all mutant mice and appropriately aged control mice were fixed by transcardial perfusion. Mice were first deeply anaesthetized with sodium pentobarbitone (100mg/kg) and transcardially perfused with vascular rinse (0.8%

NaCl in 100mM NaHPO4) followed by a freshly made and filtered solution of 4% paraformaldehyde in 0.1M sodium phosphate buffer, pH=7.4. Brains were subsequently removed and weighed before overnight post-fixation and cryoprotection at 4oC in a solution containing 30% sucrose in Tris buffered saline (TBS: 50mM Tris, pH 7.6) containing 0.05% NaN3. Brains were then bisected and serial 40µm frozen

79 coronal sections through each hemisphere were cut on a Leitz 1321 freezing microtome. Sections were collected, one per well, into 96 well plates containing a cryoprotectant solution (TBS/ 30% ethylene glycol/15% sucrose/ 0.05% sodium azide) and stored at -40oC prior to histological processing.

2.5 Nissl staining

To provide direct visualization of neuronal morphology every sixth section through the CNS was stained with the Nissl dye cresyl violet. Briefly, sections were mounted onto gelatin-chrome alum coated Superfrost microscope slides (VWR, Dorset, UK), air dried overnight and incubated for 20 minutes at 55oC in 0.05% Cresyl Fast Violet and 0.05% acetic acid (VWR), rinsed in distilled water and differentiated through a graded series of alcohols before clearing in xylene (VWR) and coverslipping with DPX (VWR). These Nissl stained sections were subsequently used for stereological analysis of regional volume, neocortical thinning, neocortical and thalamic neuron number, as detailed below.

To visualize autofluorescent storage material selected sections from each animal were mounted upon Superfrost microscope slides (VWR) and immediately coverslipped with Vectashield aqueous mounting medium (Vector Laboratories, Peterborough, UK) (Bible et al., 2004; Griffey et al., 2004). Sections were viewed under conventional epifluorescence illumination on a Zeiss Axioskop2 MOT microscope (Carl Zeiss Ltd., Welwyn Garden City, UK) and images recorded at multiple wavelengths using a Zeiss Axiocam and Axiovision software (Carl Zeiss Ltd.).

2.6 Immunohistochemistry

Immunohistochemical staining is dependant on the interaction between an antibody and its antigenic binding site on a protein, carbohydrate or lipid (Miller, 2000). However, extreme caution must be taken during the staining protocol in order to prevent non-specific background staining. Such staining may arise due to the non-

80 specific uptake of an antigen during tissue fixation, particularly when high affinity plasma proteins such as immunoglobulins are involved (Miller, 2000). Cross- reactivity is also possible when monoclonal or polyclonal antibodies incorrectly bind to similar antigens on separate molecules (Watanabe et al., 1996). Alternatively, non- immunological binding of the antibody can occur via hydrophobic and electrostatic forces to different parts within the tissue, but this is typically recognized as a uniform background colour (Kraehenbuhl and Jamieson, 1974). To test for specificity of the antibodies involved, and account for false positives, a negative and a positive control should be immunohistochemically processed along with your test sections. Negative controls must show no staining in the absence of primary antiserum, or by adsorption of the primary antibody with the relevant antigen. Conversely, the use of a section with known positivity can be used to confirm that the primary antiserum is specific to a particular antigen and further validates the negative control (Miller, 2000).

Although these precautions may limit inaccuracies during immunostaining, several other parameters may influence densities of immunoreactivities and the use of these sections for quantitative image analysis. These parameters include 1) the thickness of histological sections 2) the dilution range of antisera between studies 3) the type or composition of buffers used for rinsing, or diluting antisera and the chromogen di- aminobenzidine (DAB, Grube, 2004). Thus, where possible these variables need to be standardised in order for more accurate comparisons between studies.

a) Immunohistochemistry for interneuron markers

Interneurons are consistently affected in the CNS of human NCL (Tyynelä et al., 2004), as well as in animal models of NCL (Cooper et al., 1999; Mitchison et al., 1999; Oswald et al., 2001; Cooper, 2003; Bible et al., 2004; Kopra et al., 2004; Mitchison et al., 2004; Jalanko et al., 2005). To survey this neuronal phenotype in a murine model of JNCL, adjacent one-in-six series of sections were immunohistochemically stained for representative interneuron populations, including the neuropeptide somatostatin (SOM) and the calcium binding protein parvalbumin

(PV). Sections were incubated for 15 minutes in 1% H2O2 in TBS, rinsed in TBS and

81 blocked for 40 minutes in TBS/0.3% Triton X-100/15% normal goat serum (NGS) before overnight incubation at 4oC in one of the following polyclonal primary antisera (rabbit anti-SOM, Peninsula Laboratories, Belmont, CA, USA, 1:2000, rabbit anti-PV, Swant, Bellinzona, Switzerland, 1:5000) diluted in TBS with 10% normal goat serum and 0.3% Triton X-100. Sections were then rinsed in TBS and incubated for 2 hours with secondary antiserum (biotinylated goat anti-rabbit IgG, Vector Laboratories, Peterborough, UK, 1:1000) in TBS/0.3% Triton X-100/10% NGS. Following rinsing in TBS, sections were incubated for 2 hrs in an avidin-biotin-peroxidase complex in TBS (Vectastain Elite ABC kit, Vector Laboratories). Sections were next rinsed in TBS and immunoreactivity was visualized by incubation in 0.05% DAB (Sigma,

Dorset, UK) and 0.001% H202 in TBS for 10 minutes (a time which represents saturation for this reaction). Sections were then transferred to excess ice-cold TBS, and were rinsed again, mounted, air dried, cleared in xylene and coverslipped with DPX (VWR).

b) Immunohistochemistry for glial markers

To assess the extent of glial activation, adjacent one-in-six series of free-floating frozen sections were immunohistochemically stained using the standard immunoperoxidase protocol described above for detection of astrocytic (GFAP, S100β) and microglial (F4/80) markers. Staining was conducted using the following primary antisera (polyclonal rabbit anti-GFAP, DAKO, Cambridge, UK, 1:1000, polyclonal rabbit anti-S100β, DAKO, 1:1000, monoclonal rat anti-F4/80, Serotec, Oxford, UK, 1:100) and appropriate normal sera (normal swine serum [GFAP and S100β] and normal rabbit serum [F4/80]). Sections were then rinsed in TBS with subsequent incubation in secondary anti-serum (swine anti-rabbit [GFAP and S100β], Vector Laboratories, 1:400 and mouse adsorbed rabbit anti-rat [F4/80], Vector Laboratories, 1:1000) followed by avidin-biotin-peroxidase complex (Vectastain Elite ABC kit, Vector Laboratories). Immunoreactivity was visualised by a standard DAB reaction and sections were mounted onto slides, air dried, cleared in xylene and coverslipped with DPX (VWR).

82 2.7 Measurements of volume, neocortical and laminar thickness

A range of functionally diverse structures were selected along the rostrocaudal extent of the brain for stereological analysis. These structures include the neocortex, hippocampus, striatum, thalamus, hypothalamus and cerebellum, some of which have already been implicated in the NCLs (Rinnie et al., 2002; Goebel and Wisniewski, 2004) and were all defined according to Paxinos and Franklin (2001). Unbiased Cavalieri estimates of regional volume were made from each animal, with no prior knowledge of genotype (Gudersen and Jensen, 1987). A sampling grid with appropriate spacing was superimposed over Nissl stained sections (Tables 2.2, 2.3) and the number of points covering the relevant areas counted using a x2.5 (NA 0.075) objective. Regional volumes were expressed in µm3 and the mean volume of each region calculated for Cln3-/- mice, homozygous Cln3∆ex7/8 mice and their appropriate control strains.

Neocortical thickness measurements were made on the same one-in-six series of Nissl stained sections for primary motor (M1), primary somatosensory (S1BF), lateral entorhinal (Lent) and primary visual cortex (V1) as defined by Paxinos and Franklin (2001). These regions were chosen as they represent a range of neocortical subfields that portray a diversity of function and may be compromised in the NCLs. The length of perpendicular lines extending from the white matter to the pial surface was measured by placing 10 evenly spaced lines on each of three consecutive sections spanning each neocortical region. These measurements were obtained blind to genotype, which was achieved by obscuring identifiable markings on each slide with opaque tape and then applying codes in a randomised order. A x5 (NA 0.12) objective was subsequently used to view Nissl stained sections. Results were expressed as the mean neocortical thickness in µm per region for mutant mice and appropriately aged controls.

83 Animal Age Region Grid area (µm²) Neocortex 500 Cln3-/- mice 18 months Hippocampus 250 Striatum 175 Thalamus 175 Hypothalamus 175 Cerebellum 500 Hippocampus 200 Cln3-/- mice 14 months Striatum 200 Thalamus 200 Hypothalamus 150 Cerebellum 550 Hippocampus 200 Cln3-/- mice 5 months Striatum 200 Thalamus 200 Hypothalamus 150 Neocortex 500 Cln3-/- mice P8 Hippocampus 150 Striatum 150 Thalamus 150 Hypothalamus 200 Cerebellum 500

Table 2.2 Cavalieri based grid areas for estimating regional volume in Cln3-/- mice and age- matched control littermates. These grid sizes were determined after pilot studies and adjusted to give CE values of less than 0.1 (Gundersen and Jensen, 1987).

Individual laminar thickness measurements were also performed in M1, S1BF, Lent, and V1 using the same Nissl stained sections and referring to landmarks in Paxinos and Franklin (2001) and Hof et al. (2000). Laminae I to VI were present in all of these neocortical subfields excluding the M1, which lacks lamina IV. As with neocortical thickness measures, the thickness of each individual lamina was measured via a series

84 of 10 perpendicular lines drawn in each of three consecutive sections using a x10 (NA 0.3) objective.

Animal Age Region Grid area (µm²) Neocortex 500 Cln3∆ex7/8 mice 12 months Hippocampus 200 Striatum 300 Thalamus 200 Hypothalamus 200 Cerebellum Lateral 400 (IG/ML + WM) Cerebellum Vermis 400 (IG/ML + WM) Neocortex 500 Cln3∆ex7/8 mice P7 Hippocampus 200 Striatum 300 Thalamus 200 Hypothalamus 200 Cerebellum Lateral 300 (IG/ML + WM) Cerebellum Vermis 300 (IG/ML + WM)

Table 2.3 Cavalieri based grid areas for estimating regional volume in homozygous Cln3∆ex7/8 mice and age-matched control littermates. These grid sizes were determined after pilot studies and adjusted to give CE values of less than 0.1 (Gundersen and Jensen, 1987). IG = inner granular layer; ML = molecular layer; and WM = white matter of the cerebellum.

Results were expressed as the mean laminar thickness in µm per region for mutant mice and appropriately aged controls. All volume and thickness analyses were carried out using StereoInvestigator 4.36 software (Microbrightfield Inc., Williston, VT, USA), on a Zeiss Axioskop2 MOT microscope (Carl Zeiss Ltd) linked to a DAGE- MTI CCD-100 camera (DAGE-MTI Inc., Michigan City IN, USA).

85 2.8 Measurements of total neuronal number and volume

To examine neuronal number and volume within individual neocortical laminae and thalamic nuclei, we used StereoInvestigator 4.36 software to obtain unbiased optical fractionator estimates of neuronal number (West et al., 1991) and nucleator estimates of neuronal volume (Gunderson, 1988) from Nissl stained sections. Immunoreactive neurons were sampled using a series of counting frames distributed over a grid and superimposed onto the section. For unbiased sampling a random starting section was chosen, followed by every sixth Nissl stained section thereafter.

The boundaries of M1, S1BF, Lent and V1 regions and each individual lamina were defined as described above. Neuronal morphology and patterns of cell distribution was used as a guide to identify specific neuronal populations within a chosen structure. Lamina IV was identifiable via the presence of a homogeneous layer of darkly stained granular cells. Similarly, lamina V was identified by a layer of morphologically uniform cells, although these were large and pyramidal shaped. A combined measurement was taken for laminae II + III, which consisted of both small granular and large pyramidal cells. These layers were distinguishable from neighbouring laminae by contrasting organisation of neurons. The ventral posterior thalamic nucleus (VPM/VPL) was identified by medium sized excitatory cells with round or oval morphologies that were aggregated in clusters and separated by cell sparse areas. Projection neurons within the dorsal lateral geniculate nucleus (LGNd) were much lager and helped identify this structure from surrounding thalamic nuclei.

The rostral and caudal limits of the hippocampal formation were determined by the first and last sections to contain the region of interest. Hippocampal cell layers CA1, CA2/CA3, dentate gyrus, radiatum and oriens were defined according to Paxinos and Franklin (2001). The transition between CA1 and CA2/3 was identified by the differing morphology of cells in these two layers, CA1 neurons being slightly larger and more loosely packed than neurons in CA2 and CA3. The oriens and radiatum were again defined according to Paxinos and Franklin (2001), with the oriens located between CA1 and white matter, and the radiatum directly below CA1. Only neurons with a clearly identifiable nucleus were sampled and all counts were carried out using

86 a x100 oil objective (NA 1.4). Appropriate sampling schemes were applied to the regions of interest, as detailed below (Tables 2.4, 2.5).

Animal Age Region Grid area Frame area (µm²) (µm²) M1-LAM V 122500 3159.23 Cln3-/- mice 18 months V1-LAM IV 22500 1095.82 S1BF-LAM IV 40000 1095.82 S1BF-LAM V 122500 3159.23 VPM VPL 122500 3382.55 LGN 22500 3518.87 S1BF-LAM IV 40000 1095.82 Cln3-/- mice 14 months S1BF-LAM V 40000 1095.82 VPM VPL 122500 3382.55 LGND 22500 3518.87 S1BF-LAM IV 40000 1095.82 Cln3-/- mice 5 months S1BF-LAM V 40000 1095.82 VPM VPL 122500 3382.55 M1-LAM V 122500 3382.55 Cln3-/- mice P8 S1BF-LAM IV 40000 1095.82 S1BF-LAM V 40000 1095.82 VPM VPL 122500 3382.55

Table 2.4 Optical fractionator and nucleator based sampling schemes for different neuron populations in Cln3-/- mice and age-matched control littermates. Various laminae (LAM) were assessed for the presence of neurons in the primary motor (M1), primary somatosensory (S1BF) and primary visual cortex (V1) of Nissl stained sections.

87

M1-LAM V 122500 2141.84 Cln3∆ex7/8 12 months S1BF-LAM II + III 122500 2141.84 mice S1BF-LAM IV 40000 1095.82 S1BF-LAM V 122500 2141.84 VPM VPL 122500 3382.55 M1-LAM V 122500 3402.75 Cln3∆ex7/8 P7 Lent-LAM II 62500 3402.75 mice S1BF-LAM IV 40000 1095.82 S1BF-LAM V 40000 1095.82 VPM VPL 122500 3382.55

Table 2.5 Optical fractionator and nucleator based sampling schemes for different neuron populations in homozygous Cln3∆ex7/8 mice and age-matched control littermates. Various laminae (LAM) were assessed for the presence of neurons in the primary motor (M1), primary somatosensory (S1BF) and lateral entorhinal cortex (Lent) of Nissl stained sections.

2.9 Measurements of interneuron number

The number of GABAergic interneurons expressing SOM or PV in S1BF and M1 was determined using the design-based optical fractionator method (West et al., 1991). Immunoreactive neurons were sampled as detailed above. The rostral and caudal limits of selected neocortical sub-regions were defined according to Paxinos and Franklin (2001). The lateral to medial extent of these chosen structures was identified by comparison with an adjacent series of Nissl stained sections and anatomical reference points. A x40 oil objective (NA 1.30) was then used to count clearly identifiable immunoreactive neurons, which fell within the dissector frame. Appropriate sampling schemes were applied to the regions of interest, as detailed below (Table 2.6).

88

Animal Age Antigen Region Grid area Frame area (µm²) (µm²) Parvalbumin- M1 Cln3∆ex7/8 12 months Parvalbumin S1BF 62500 20191.1 mice Somatostatin M1 Somatostatin S1BF Parvalbumin S1BF 122500 Cln3-/- mice 18 months Parvalbumin V1 62500 20191.1 Somatostatin S1BF 122500 Somatostatin V1 62500

Table 2.6 Selected sampling scheme for interneurons in Cln3-/- mice, homozygous Cln3∆ex7/8 mice and age-matched control littermates. The primary motor (M1) and primary somatosensory (S1BF) and primary visual cortex were assessed for the presence of interneuronal markers, parvalbumin and somatostatin in appropriately stained sections.

Due to the comparatively low abundance of interneurons present in the hippocampus vs. the neocortex, stereological methods prove inefficient at estimating hippocampal interneuron numbers without sampling the entire tissue (Bible et al., 2004). Instead, counts of the number of interneurons expressing SOM or PV were made in the five most rostral sections of a similar series through defined hippocampal subfields (For SOM; hilus, CA3/2/1, stratum oriens and radiatum. For PV; dentate gyrus, CA3/2, CA1, stratum oriens and radiatum). Counts were carried out under a x20 (NA 0.5) objective and only positively stained cells with clear neuronal morphology were counted. The number of interneurons in each hippocampal subfield was expressed as the mean number of neurons per section.

89 2.10 Assessment of glial phenotype

a) Quantitative analysis of glial phenotype

The expression of glial markers GFAP and F4/80 was assessed using a semi- automated thresholding image analysis system. This analysis was performed blind to genotype, as detailed above and conducted in Cln3-/- mice at eight days, 5, 14 and 18 months, and in Cln3∆ex7/8 mice at seven days and 12 months of age. Forty non- overlapping images were captured, on three consecutive sections, through each of the neocortical regions M1, S1BF and V1, the cerebellum, striatum, thalamus, hippocampal dentate gyrus and a combined measurement for the stratum oriens and CA1. These regions of interest were chosen, as detailed above. All RGB images were captured via a live video camera (JVC, 3CCD, KY-F55B), mounted onto a Zeiss Axioplan microscope using a x40 objective and stored permanently onto a CD-ROM as JPEG files. All parameters including lamp intensity, video camera setup and calibration were maintained constant throughout image capturing.

Images were subsequently analyzed using either Optimas 6.2 (Media Cybernetics, Berkshire, UK), or Image-Pro Plus 4.0 (Media Cybernetics, Berkshire, UK) image analysis software, with an appropriate threshold that selected the foreground immunoreactivity above background. This thresholding method was based on the optical density of the immunoreactive product and was accurately defined by averaging the highest and lowest immunoreactivities within a sampled population of immunohistochemical marker (per colour/filter channel selected) and measured on a scale from 0 (100% transmitted light) to 255 (0% transmitted light) for each pixel. The selected threshold was then applied as a constant to all subsequent images analyzed per batch of animals and reagent used to determine the specific area of immunoreactivity for each antigen in each region. Each field measured 120µm wide, with a height of 90µm. Therefore, the total area compiled from 40 fields in each corresponded to 432000µm2. Macros were recorded to transfer the data to a Microsoft Excel spreadsheet and was subsequently analysed statistically. Data were plotted graphically as the mean percentage area of immunoreactivity per field ± SEM for each region.

90 b) Measurements of astrocytic number

To quantify changes in astrocyte number we obtained design-based optical fractionator estimates of S100β-positive soma, together with nucleator estimates of astrocyte volume. Immunoreactive astrocytes were sampled in chosen structures using a series of counting frames distributed over a grid and superimposed onto the section using StereoInvestigator 4.36 software, exactly as detailed for neuronal counts. The rostral and caudal limits of selected neocortical sub-regions were defined according to Paxinos and Franklin (2001). A x40 oil objective (NA 1.30) was then used to count clearly identifiable immunoreactive astrocytes, which fell within the dissector frame. An appropriate sampling scheme was applied to the regions of interest, as detailed below (Table 2.7).

Animal Age Region Grid area Frame area (µm²) (µm²)

Cln3∆ex7/8 mice 12 months S1BF 122500 20191.1 CA1 + Oriens

Cln3∆ex7/8 mice P7 M1 202500 20191.1

V1

Table 2.7 Selected sampling scheme for chosen structures in homozygous Cln3∆ex7/8 mice and age- matched control littermates. The hippocampal CA1 & stratum oriens, primary somatosensory (S1BF), primary motor (M1) and primary visual cortex (V1) were assessed for the presence S100β-positive soma.

2.11 Statistical analysis a) One-way ANOVA

The statistical significance of differences between genotypes of all quantitative data was assessed using a one-way analysis of variance (ANOVA) (SPSS 11.5 software,

91 SPSS Inc, Chicago, IL, USA), with statistical significance considered at P< 0.05. One-way ANOVA compares the variability of the mean numerical value within a group (within-group variation), to mean differences of numerical value between individuals from separate groups (between-group variation) (Petrie and Sabin, 2000). These groups are defined by an exposure to multiple categories such as genotypes (i.e. wild type or diseased phenotype). Thus, within-group variation would apply to individual animals of a particular genotype within the context of this study. One-way, statistical analysis is so termed as the exposure groups are only classified by a single variable, whereas two or even three way analysis is also possible and as such can involve multiple variables (Kirkwood and Sterne, 2003). For a one way ANOVA, the null hypothesis would be that the mean outcome does not differ between-groups or within-groups. However, in this scenario, if the mean differs between the groups, then the in-between group variation will be larger than the within-group variation and statistical significance may be evident (Kirkwood and Sterne, 2003).

b) Co-efficient of error

The mean co-efficient of error (CE) for all individual optical fractionator and nucleator estimates was calculated according to the method of Gundersen and Jensen (1987) and was less than 0.1 in all these analyses. Quantitative analysis, as conducted by design-based stereological methods, is only an estimate and not an exact measurement, thus results may vary if the same stereological study were to be independently repeated. The mean co-efficient of error takes such discrepancies into account, and estimates the variation, had the stereological measurements been repeated infinitum (Schmitz, 2005).

92

CHAPTER 3

Progressive Neuropathological Changes in the Cln3-/- Mouse Model of JNCL

93 3.1 Introduction

The major aim of this thesis is to map progressive neuropathological changes over time in mouse models of juvenile neuronal ceroid lipofuscinoses (JNCL). Detailed quantitative information about the neuropathological phenotype of JNCL is almost entirely lacking and the limited amount of data that is available is restricted to post- mortem material (Braak and Goebel, 1978; Braak and Goebel 1979). In contrast, animal models provide the opportunity to study progressive pathogenesis, which has now been explored in two distinct mouse models of JNCL, each one generated by a separate targeting strategy. The results described in this chapter are obtained from Cln3-/- mice, which have been engineered to bear a null Cln3 gene and are therefore unable to produce functional Cln3 protein. These mice have been examined for neuronal and glial phenotypes at various time-points, ranging from pre-symptomatic (5 months of age), to affected (14 months of age) and terminal (18 months of age) stages of disease. This detailed analysis of neuropathological landmarks in aged mice served to identify which CNS structures are targeted by disease. These structures were then explored at progressively younger ages in order to determine when they are first affected and the rates at which disease phenotype progresses. A combination of unbiased, design-based stereological methodology and thresholding image analysis of immunohistochemically stained sections were used to investigate the phenotype of Cln3-/- mice at each age.

3.2 Autofluorescent storage material

The accumulation of autofluorescent storage material is characteristic of the NCLs (Hofmann and Peltonen, 2001) and was assessed in Cln3-/- mice at 18 months of age (Figure 3.1). As expected, these mice exhibited a widespread intracellular accumulation of storage material within neuronal soma throughout the CNS (Figures 3.1B-D, F-H). This storage material was punctate in appearance and fluoresced at multiple wavelengths, as viewed by conventional epifluorescence microscopy (Figure 3.1). However, 18 month old control mice also displayed significant accumulation of age-related lipopigment (Figures 3.1A, E). As such, the relative differences in

94

Figure 3.1 Accumulation of autofluorescent storage material in Cln3-/- mice. (A-H) Representative images of unstained coronal sections through the primary motor cortex (A-D) and cerebellum (E-H) viewed by epifluorescence using rhodamine (Rho) or FITC filter sets. Sections from 18 month old Cln3-/- mice display a subtle, but widespread intracellular accumulation of storage material (B-D, F-H), that fluoresces with both filter sets and appears yellow in merged images (D, H). In contrast, sections from littermate controls (+/+) display a lower level of autofluorescence in the neocortex (A), and within Purkinje neurons of the cerebellum (E).

95 autofluorescence were subtle between mice of either genotype, although storage material was consistently more prominent within Cln3-/- mice (Figure 3.1).

3.3 Evaluation of regional atrophy in Cln3-/- mice

To explore progressive neurodegenerative changes in regional volume within the CNS of Cln3-/- mice, we carried out a stereological survey at 5, 14 and 18 months of age (Figure 3.2). The Cavalieri method (Chapter 2) was used to obtain unbiased estimates of volume for the cortical mantle, hippocampus, striatum, thalamus and cerebellum in Nissl stained sections. At 5 months of age, there were no significant differences in the volume of any analyzed region of the CNS between Cln3-/- and control mice (Figure 3.2A). With the exception of the striatum at 14 months, there was no significant effect upon regional volume at either 14 (Figure 3.2B) or 18 months of age (Figure 3.2C) in Cln3-/- mice.

3.4 Regional effects upon neocortical thinning and lamination in Cln3-/- mice

Although the volume of the cortical mantle was unaffected in aged Cln3-/- mice, such measurements of neocortical volume cannot discriminate between events in individual neocortical subfields. Therefore, thickness measurements were made in a series of functionally distinct neocortical regions, including the primary motor (M1), somatosensory barrel field (S1BF), primary visual (V1) and lateral entorhinal (Lent) cortex (Figure 3.3). Before 18 months of age there was no significant change in the thickness of the S1BF (Figures 3.3A, B). In contrast, 18 month old Cln3-/- mice displayed highly variable changes in thickness between these subfields (Figure 3.3C). Rostrally, there was significant atrophy of both the M1 and the S1BF. In contrast, V1 and Lent, exhibited significant increases in thickness when compared to age matched controls.

96

Figure 3.2 Unbiased Cavalieri estimates of regional volume in homozygous Cln3-/- mice.

There was no significant regional atrophy in Cln3-/- vs. littermate controls (+/+) at 5 (A), 14 (B) and 18 (C) months of age. Regions examined include cortical mantle (NeoCtx), hippocampus (Hipp), thalamus (Thal), hypothalamus (Hypoth), striatum (CPu) and whole cerebellum (Cb total). * p<0.05, one way ANOVA.

97

Figure 3.3 Neocortical thickness measurements in Cln3-/- mice vs. littermate controls (+/+) at 18 (A), 14 (B) and 5 (C) months of age. Neocortical thinning was absent from primary somatosensory barrel field (S1BF) cortex at 5 (A) and 14 (B) months. (A) 18 month Cln3-/- displayed pronounced thinning of both primary motor (M1) and S1BF cortex, whereas primary visual (V1) and lateral entorhinal (Lent) cortex exhibited a significant increase in overall thickness. ** p<0.01, *** p<0.001, one way ANOVA.

98

Figure 3.4 Laminar-specific changes in Neocortical thickness in Cln3-/- mice (A-F). Laminar thickness measurements in primary somatosensory barrel field (S1BF, A-C) cortex at 5 (A), 14 (B) and 18 months (C), and in primary motor (M1, D), primary visual (V1, E) and lateral entorhinal cortex (Lent, F) at 18 months. Analysis of these neocortical regions revealed a complex series of changes in thickness of individual laminae in 5, 14 and 18 month old Cln3-/- mice compared with littermate controls (+/+). * p<0.05, ** p<0.01, *** p<0.001, one way ANOVA. These data are also summarized in Table 3.1.

99

Age Region I II/III III IV V VI Total Thickness

18 months M1 NS NS na na ↓ ↓ ↓

18 months V1 NS ↑ na NS ↑ ↑ ↑

18 months Lent ↑ ↑ NS NS NS NS ↑

18 months S1BF NS ↓ na NS ↓ NS ↓

14 months S1BF NS ↓ na ↑ NS NS NS

5 months S1BF NS ↓ na ↓ ↑ ↓ NS

Table 3.1 Tabular depiction of significant changes in individual laminar thickness in Cln3-/- mice. Measurements taken in the primary motor (M1), primary somatosensory barrel field (S1BF), primary visual (V1) and lateral entorhinal cortex (Lent) of 5, 14 and 18 month old Cln3-/- mice vs. littermate controls (+/+).↑ = significantly thicker in Cln3-/-; ↓ = significantly thinner in Cln3-/-; na = not applicable, M1 has no lamina IV. A combined measurement of laminae II and III was made in all regions except Lent where these laminae were measured separately.

100 To determine whether these changes in neocortical thickness were due to lamina- specific events, individual laminar thickness was measured in each neocortical subfield. These measurements revealed a complex series of changes in laminar thickness at different stages of disease progression (Figure 3.4). Although the overall thickness of S1BF was unaffected in Cln3-/- mice at 5 and 14 months vs. control littermates, a series of contrasting changes were observed in laminar thickness. Unexpectedly, 5 month old Cln3-/- mice exhibited significant changes in the thickness of all laminae, apart from lamina I (Figure 3.4A), with significant thinning of laminae II, III, IV and VI and an increased thickness of lamina V. At 14 months of age, these changes in laminar thickness followed a similar trend to those in 18 month old mice, with the exception of lamina IV, which was significantly thicker and lamina V which appeared to be largely unchanged (Figure 3.4B). At 18 months of age, laminae II, III and V displayed the greatest fluctuations in thickness in Cln3-/- mice compared with littermate controls. Consistent with overall changes in neocortical thickness (Figure 3.3), S1BF and M1 displayed significant thinning that was restricted to laminae II, III and V (S1BF, Figure 3.4C), and laminae V and VI (M1, Figure 3.4D). More caudally, the increased thickness of VI was due to selective increases in the thickness of laminae II, III, V and VI (Figure 3.4E), with similar affects in laminae I and II of Lent (Figure 3.4F).

3.5 Regional and laminar effects upon neuronal number in Cln3-/- mice

To investigate whether these changes in individual laminar thickness were the result of selective effects on particular neuron populations, unbiased stereological estimates of neuronal number (optical fractionator) and size (nucleator) were obtained from the same series of Nissl stained sections prepared from Cln3-/- mice (Chapter 2). This analysis focused upon S1BF and M1, two cortical regions which subserve contrasting functions, but were consistently affected in 18 month old Cln3-/- mice with thinning of lamina V in both regions. Optical fractionator estimates revealed a significant loss of lamina V pyramidal neurons in S1BF and M1 of 18 month old Cln3-/- mice (Figure 3.5A). Although the loss of lamina V neurons in S1BF was prominent in aged mutant

101

Figure 3.5 Unbiased optical fractionator estimates of neuronal number in neocortical laminae and thalamic nuclei in aged Cln3-/- mice. (A) Optical fractionator estimates of neuronal number revealed the significant loss of lamina V pyramidal neurons in primary motor (M1) and primary somatosensory barrelfield (S1BF) cortex, as well as lamina IV granule neurons in primary visual cortex (V1); however, this loss was absent from lamina IV in S1BF of 18 month Cln3-/- mice when compared to littermate controls (+/+). There was also a significant reduction in the number of neurons in the ventral posterior thalamic nucleus (VPM/VPL) and dorsal lateral geniculate nucleus (LGNd) of mutant mice. (B) Nucleator estimates of neuronal volume revealed no significant atrophy or hypertrophy in any laminae of Cln3-/-. * p<0.05, ** p<0.01, one way ANOVA.

102

Figure 3.6 Unbiased optical fractionator estimates of neuronal number in neocortical laminae and thalamic nuclei in Cln3-/- mice. Optical fractionator estimates of neuronal number did not reveal significant loss of either lamina IV granule neurons or lamina V pyramidal neurons in primary somatosensory barrelfield cortex (S1BF) of Cln3-/- mice when compared to 5 (A) and 14 (B) month old control littermates (+/+). (A) At 5 months, neuronal loss was absent from laminae IV and V of S1BF, and the ventral posterior thalamic nucleus (VPM/VPL) in both mutant and control mice. (B) At 14 months there was a significant reduction in the number of neurons in VPM/VPL, which was not exhibited in the dorsal lateral geniculate nucleus (LGNd) of mutant mice. * p<0.05, one way ANOVA.

103 mice, this trend was less apparent at 14 (Figure 3.6B) and 5 months of age (Figure 3.6A) and did not reach statistical significance.

To determine whether the loss of pyramidal neurons in the S1BF had any effect on thalamic relay neurons that receive sensory feedback from these cells, unbiased stereological estimates of neuronal number and size were obtained in the ventral posterior thalamic nucleus (VPM/VPL, chapter 2). A significant loss of these relay neurons was observed in 18 month old Cln3-/- mice (Figure 3.5A). These neurodegenerative changes were also evident in younger animals at 14 (Figure 3.6B) and even 5 months of age (Figure 3.6A). Since neurons within the VPM/VPL relay sensory information to granule neurons within lamina IV of SIBF, the survival of these neurons was also assessed. However, no significant loss of S1BF granule neurons was evident at any age examined, suggesting that the loss of VPM/VPL neurons not only occurs relatively early, but is independent of effects upon their target neurons.

To determine whether the loss of somatosensory relay neurons extended to thalamic neurons relaying other sensory modalities, the number and volume of neurons within the dorsal lateral geniculate nucleus (LGNd) that relays information to the primary visual cortex (Livingstone and Hubel, 1988; Sincich and Horton, 2004), was also assessed by unbiased stereology (Chapter 2). This analysis revealed a significant loss of large projection neurons in LGNd of Cln3-/- mice at 18 months of age (Figure 3.5A), that was not evident at 14 months of age (Figure 3.6B). This loss of visual relay neurons in LGNd occurred in the absence of significant effects upon their neocortical target neurons in lamina IV of V1 in 18 month old Cln3-/- mice (Figure 3.5B).

In contrast to the significant effects upon neuron number, nucleator estimates of neuronal volume revealed only subtle changes at 5, 14 (data not shown) and 18 months of age (Figure 3.5B), which were not significant in any neocortical lamina or thalamic nucleus examined.

104 3.6 Regional and laminar effects upon interneuronal number in Cln3-/- mice

The loss of GABAergic interneurons is a common pathological phenotype in human NCLs (Tyynelä et al., 2004), as well as other murine models of NCL (Cooper et al., 1999; Mitchison et al., 1999; Cooper, 2003; Bible et al., 2004; Kopra et al., 2004; Mitchison et al., 2004; Jalanko et al., 2005). In order to assess whether similar changes are evident in 18 month Cln3-/- mice, the survival of parvalbumin (PV) and somatostatin (SOM) positive interneuron populations was surveyed in the neocortex (Figure 3.7A) and hippocampus (Figures 3.7B, C) of both mutant and control littermates.

Primary somatosensory and visual cortex. Optical fractionator estimates revealed a general trend to a reduced number of interneurons immunoreactive for PV and SOM in S1BF and V1. This loss of interneurons only reached significance for PV-positive cells in V1, whereas effects upon SOM positive cells were less pronounced (Figure 3.7A).

Hippocampus. Interneuron populations immunoreactive for these markers were then counted in the hippocampal formation of homozygous Cln3-/- mice. Similar to changes in the neocortex, there was an overall trend for a reduced number of SOM-positive interneurons in the hippocampus of Cln3-/- mice. Significant cell loss was restricted to specific subfields, including PV (Figure 3.7B) and SOM-positive interneurons (Figure 7C) in the stratum oriens and radiatum. Although cell loss was observed in almost all of the remaining hippocampal subfields of mutant mice, this was more subtle and did not reach statistical significance (Figures 3.7B, C).

3.7 Astrocytic and microglial responses in Cln3-/- mice

As described previously, there is a pronounced glial response in human NCLs (Haltia et al., 1973a, Haltia et al., 1973b; Braak and Goebel, 1978; Braak and Goebel, 1979; Haltia, 2003; Tyynelä et al., 2004), as well in other mutant mice modeling the human condition (Bronson et al., 1998; Cooper et al., 1999; Gupta et al., 2001; Bible et al.,

105

Figure 3.7 Interneuron counts in Cln3-/- mice. (A) Optical fractionator estimates of interneuron number revealed a significant loss of parvalbumin (PV) positive interneurons in the primary visual (V1), but not in the primary somatosensory barrelfield cortex (S1BF) of 18 month old Cln3-/- vs. control littermates (+/+). There was no significant loss of somatostatin-(SOM) positive interneurons in either S1BF or V1. (B-C) Counts of hippocampal interneuron number revealed a significant loss of interneurons immunoreactive for parvalbumin in the oriens (B) and for somatostatin in the radiatum (C). There was no significant loss of PV and SOM-immunopositive cells in any other subfield of 18 month old Cln3-/- compared to age matched controls. Hippocampal subfields include the dentate gyrus (DG), hilus (Hi), CA3/2/1, stratum oriens (Oriens) and stratum radiatum (Radiatum). * p<0.05, ** p<0.01, one way ANOVA.

106 2004; Jalanko et al., 2005). To examine glial responses in Cln3-/- mice, 18 month old mutants were assessed for the distribution of astrocytic marker GFAP and microglial marker F4/80. The expression of these glial markers was quantified using thresholding image analysis (Chapter 2) in neocortical subregions M1, S1BF and V1; subcortical structures, thalamus and striatum; and hippocampal subfields dentate gyrus (DG) and stratum oriens together with CA1.

Qualitative assessment of GFAP-positive astrocytes: At 18 months of age, control mice displayed few scattered GFAP-positive astrocytes in the neocortical grey matter, which were almost exclusively located in lamina V1 and in lamina I adjacent to the pial surface (Figures 3.8A, C, D). Age matched Cln3-/- mice exhibited a more pronounced astrocytosis, with a tendency for clustering in both the superficial and deeper cortical laminae, although some GFAP-positive astrocytes were also observed in intermediate laminae (Figures 3.8B, E, F). Higher magnification of astrocytes revealed a protoplasmic appearance, although a subtle transformation into a fibrous hypertrophied morphology was occasionally noted, with a slightly greater incidence in Cln3-/- vs. control mice (Figures 3.8 C-F).

Subcortically, this astrocytic response was more dispersed within the striatum and hippocampus of both Cln3-/- and control mice. GFAP immunoreactivity was particularly pronounced within the hippocampus, although this staining pattern did not differ obviously between animals of either genotype (data not shown). GFAP- immunopositive astrocytes were also intensively stained within the thalamus of Cln3-/- mice when compared to controls (Figures 3.8G, H). Unlike in the striatum and the hippocampus, astrocytes within the thalamus of Cln3-/- mice were more localized and regionally restricted, with clusters of GFAP-positive astrocytes observed in parts of the laterodorsal (LD), ventromedial (VM) and ventrolateral (VL) thalamic nuclei.

Quantitative analysis of GFAP immunoreactivity: GFAP The relative level of GFAP immunoreactivity varied markedly between CNS regions at 18 months of age (Figure 3.9). Astrocytosis was pronounced in M1 of Cln3-/- mice, as demonstrated by a highly significant increase in GFAP-positive astrocytes, which was almost double that observed in age matched controls. In contrast, there were no appreciable differences between genotypes in the extent of GFAP immunoreactivity in S1BF and V1. A series

107

Figure 3.8 Astrocytic responses in Cln3-/- mice. (A-F) Immunohistochemical staining for

GFAP revealed astrocytosis in the neocortex of 18 month old Cln3-/- (B) vs. control littermates (+/+, A). Within the primary motor cortex (M1) GFAP-positive astrocytes formed clusters in both deep (F) and superficial laminae of mutant mice (E), although clusters were also formed in intermediate laminae (B). Astrocytic clusters were smaller and far less numerous in control mice, and were almost exclusively localized to deeper (D) and superficial laminae (C). Compared to controls (G) the thalamus of Cln3-/- exhibited pronounced astrocytosis that was mainly confined to individual thalamic nuclei (H), with GFAP-positive astrocytes prominent in laterodorsal (LD), ventromedial (VM) and ventrolateral (VL) thalamic nuclei, but more subtle in the ventral posterior (VPM/VPL), mediodorsal (MD), posterior (Po) and centrolateral (CL) thalamic nuclei.

108

Figure 3.9 Immunohistochemical analysis of astrocytic responses in Cln3-/- mice. Quantitative image analysis revealed the significantly increased expression of GFAP within the primary motor cortex (M1) and subcortical structures, the striatum (Cpu) and thalamus of 18 month Cln3-/- vs. control littermates (+/+). In contrast, hippocampal subfields displayed significantly reduced levels of GFAP in mutant mice compared with controls. Areas surveyed include neocortical M1, somatosensory barrelfield (S1BF) and visual (V1); subcortical CPu and thalamus; hippocampal dentate gyrus (DG), hippocampal stratum oriens and CA1 (Oriens CA1). * p<0.05, *** p<0.001, one way ANOVA.

109 of subcortical structures including the striatum and thalamus displayed significant increases in GFAP-positive staining in Cln3-/- mice, although this was more prominent in the thalamus. In marked contrast, the hippocampal dentate gyrus, stratum oriens and CA1 of Cln3-/- mice exhibited a significantly reduced expression of GFAP when compared to control littermates.

Qualitative assessment of F4/80 positive microglia: Compared to control littermates, F4/80 immunoreactivity in 18 month Cln3-/- was more prominent in both neocortical and subcortical structures (Figure 3.10). The distribution of F4/80 positive microglia was widespread in animals of either genotype, but there were subtle changes in cellular morphology (Figures 3.10C, D). F4/80 positive microglia in the neocortex of control mice had small soma and numerous thin ramified processes extending into the neuropil (Figure 3.10C). In marked contrast, Cln3-/- mice displayed partial transformation to brain macrophage-like morphology, with enlarged soma and numerous thickened processes that were more intensely stained (Figure 3.10D). The incidence of microglia with this morphology was not different between individual neocortical laminae. However, there were no F4/80-positive cells with full brain macrophage morphology in the neocortex, or any other region of the CNS.

In subcortical structures, there were stark differences in the intensity of F4/80 immunoreactivity between animals of either genotype (Figures 3.10E-F). F4/80 positive staining of microglia was pronounced in Cln3-/- mice, in marked contrast to a paler staining in control littermates. In either genotype, this staining was typically widespread and did not form clusters of F4/80 positive cells (Figures 3.10E-F). Similar to observations in the neocortex, there was a greater incidence of partial transformation of microglia, within subcortical structures of Cln3-/- mice vs. control littermates. The incidence of such changes in cell morphology was comparable between subcortical structures and did not localize to specific subfields.

Quantitative analysis of F4/80 immunoreactivity: Consistent with qualitative observations, thresholding image analysis revealed a widespread increase in F4/80 immunoreactivity in almost all neocortical and subcortical structures in aged mutants vs. control littermates (Figure 3.11). Indeed, the upregulation of F4/80 immunoreactivity was highly significant in S1BF of the neocortex; subcortical

110

Figure 3.10 Microglial responses in Cln3-/- mice. Immunohistochemical staining for F4/80 revealed partial activation of microglia in 18 month old Cln3-/- mice than age matched controls (+/+). F4/80-positive microglia in the primary motor cortex (M1) had enlarged soma and numerous ramified processes that were darkly stained in Cln3-/- homozygotes (B, D). In marked contrast, microglia had smaller soma and thin ramified processes in littermate controls (A, C). This difference between genotypes was even more pronounced in the thalamus (Thal), with greater evidence for partial activation in Cln3-/- mice (F) vs. age matched controls (E). Thalamic nuclei include ventral posterior (VPM/VPL), mediodorsal (MD), laterodorsal (LD), ventromedial (VM), posterior (Po), and centrolateral (CL).

111

Figure 3.11 Immunohistochemical analysis of microglial responses in Cln3-/- mice. F4/80 immunoreactivity was significantly increased within all analysed regions, except for the primary visual cortex (V1), which was significantly reduced in 18 month Cln3-/- vs. control littermates (+/+). Areas surveyed include neocortical primary motor (M1), somatosensory barrelfield (S1BF) and V1; subcortical striatum (CPu) and thalamus; hippocampal dentate gyrus (DG), hippocampal stratum oriens and CA1 (Oriens CA1). *** p<0.001, one way ANOVA.

112 structures, thalamus and striatum; and hippocampal subfields dentate gyrus (DG) and stratum oriens with CA1. Neocortical V1 was the notable exception and displayed a significant reduction of F4/80 immunoreactivity in mutant mice, whereas a small increase in F4/80 staining was exhibited within M1, but this did not reach significance.

3.8 Discussion

Characterization of both neurodegenerative and reactive phenotypes in Cln3-/- mice has enabled detailed mapping of neuropathological events in this robust mouse model of JNCL. Because these analyses were conducted over a range of defined time-points, it is also possible to document the sequence of neuropathological changes during pathogenesis.

Results from these studies demonstrated the significant accumulation of autofluorescent storage material in the CNS of severely affected Cln3-/- mice, with a more subtle accumulation of age-related lipopigments in controls. Despite these intracellular events, regional atrophy was negligible in Cln3-/- brains, even at the terminal stages of disease. Nevertheless, closer examination of aged mutants revealed a complex array of events, with contrasting changes in neocortical thickness in individual subfields. This neocortical thinning was a late event in disease pathogenesis, which only became apparent at 18 months of age. However, measurements of individual cortical laminae revealed that numerous effects were apparent as early as 5 months of age. These changes were again complex, with particular laminae contributing to greater extents to changes in overall neocortical thickness of each neocortical subfield. In particular, the thickness of lamina V consistently followed patterns of overall neocortical thickness in M1, S1BF and V1 at terminal stages of disease. Taken together with data of neuronal number, these findings highlight the selective vulnerability of lamina V pyramidal neurons to pathological changes in JNCL. These may include the presence of autofluorescent storage material within these cells, or fluctuations in pH associated with an absence of functional Cln3 protein.

113

To determine whether these laminar specific effects were due to the loss of particular neuron populations and/or neuronal volume, a stereological survey of neuronal number/size was conducted in selected laminae (Chapter 2). A consistent feature of 18 month old Cln3-/- mice was the selective loss of lamina V pyramidal neurons in M1 and S1BF, reflecting the thinning of this lamina in these neocortical regions. Earlier in pathogenesis, no loss of lamina V pyramidal neurons was evident at either 5 or 14 months of age, although significant thinning of this lamina was already apparent. As such, individual laminar atrophy precedes neuronal loss by several months. Although this could plausibly be due to neuronal atrophy, our data revealed no change in the size of lamina V neurons between animals of either genotype. Therefore, the underlying causes for laminar and overall neocortical thickness are likely to be multifactorial, most likely occurring from multiple pathological events that may include effects upon cell packing or dendritic arborization.

A significant loss of neurons was also evident within sensory relay nuclei of the thalamus, relaying both somatosensory (VPM/VPL) and visual (dLGN) information. Taken together with changes to S1BF, these data reveal that the thalamocortical system is compromised in aged Cln3-/- mice, and that these effects extend to nuclei that relay more than one sensory modality. The loss of thalamic relay neurons in Cln3- /- mice is a relatively early event in pathogenesis occurring as early as 5 months of age. In contrast, lamina IV granule neurons were unaffected in mutant mice at any age, and significant loss of lamina V pyramidal neurons was only evident in 18 month old Cln3-/- mice. These data provide the first evidence for a selective loss of somatosensory relay neurons in VPM/VPL, which occurs independently of effects on their target neurons (lamina IV), but results in a delayed loss of lamina V feedback neurons of S1BF. The loss of large projection neurons in dLGN was a much later event and only became apparent at 18 months of age. These data suggest that even within the thalamus, loss of different neuronal populations occurs at variable rates or at different time-points during the course of disease.

As mentioned previously (Chapter 1), GABAergic interneurons are also vulnerable within the NCLs (Cooper et al., 1999; Mitchison et al., 1999; Cooper, 2003; Bible et al., 2004; Kopra et al., 2004; Mitchison et al., 2004; Tyynelä et al., 2004; Jalanko et

114 al., 2005). In contrast, the loss of interneurons was subtle in Cln3-/- mice, although these changes were more pronounced in individual neocortical regions and hippocampal subfields. As in other mouse models of NCL, the severity of effects on interneuron populations also depended upon which phenotypic marker these interneurons normally express. Indeed, patterns of interneuron loss in the NCLs have been shown to vary tremendously in different subpopulations of these cells, with the relative sparing of calretinin-positive inteneurons (D’Orlando et al., 2002; Cooper et al., 2003; Bible et al., 2004; Tynella et al., 2004). It has been suggested that these selective effects may depend upon the relative ability of these calcium binding proteins to buffer against long-term excitotoxicity (D’Orlando et al., 2002), but this hypothesis remains unproven in the NCL CNS.

Although the precise mechanisms of neuronal injury and cell loss in JNCL are elusive, neurons are capable of synthesizing chemokines under certain pathological conditions (Harrison et al., 1998; Flugel et al., 2001; Loddick and Rothwell, 2002). Certainly, chemokine production occurs in injured neurons with the subsequent recruitment of microglia into the surrounding area (Harrison et al., 1998; Flugel et al., 2001). It is unclear however if the upregulation of glial markers observed in aged Cln3-/- mice represent a degenerative response to similar cues (Gebicke-Haerter, 2001), a regenerative response (Kreutzberg, 1996; Simard and Nedergaard, 2004), or perhaps a combination of the two. It will be important to investigate each of these possibilities to determine to precise role of these glial responses in JNCL pathogenesis.

The expression of GFAP immunopositive astrocytes was consistently more pronounced in superficial and deeper neocortical laminae in aged mutants. Such patterns of atrocytosis suggest that neuropathological events which selectively occur within lamina V may release cues that spread to lamina VI, where astrocytes subsequently attempt to maintain the extracellular environment. In contrast, neocortical distribution of F4/80 was generalized, but was notably pronounced in S1BF, a subfield where loss lamina V neurons may have been subject to the potentially neurotoxic affects of activated microglia.

115 Distinct changes in GFAP expression were also noted in subcortical structures of aged Cln3-/- mice. Thus, regardless of degenerative effects on thalamic relay neurons, astrocytes may be attempting to prevent further cell loss. Since these changes to GFAP-positive astrocytes were accompanied by an increased expression of F4/80- positive microglia, it is possible that similar cues are controlling astrocytes and microglia. However, it is in unclear whether astrocytes are simply offering neuroprotection from neurotoxic affects of activated microglia, or whether both glial populations are attempting to protect neurons from unknown pathological insults.

In marked contrast to other regions of the mutant CNS, astrocytosis was severely reduced in the hippocampus, whereas F4/80 immunoreactivity was characteristically upregulated. These findings suggest that astrocytes may themselves be targeted as part of the disease, or that the cues responsible for mediating astrocytosis have subsided in the hippocampus. Furthermore, the contrasting effects upon GFAP and F480 staining imply that separate cues are responsible for inducing astrocytosis and microglial activation within this structure. Indeed, since astrocytosis may be insufficient to protect neurons from neurotoxic affects of activated microglia or other pathological insults, this potentially hostile environment and lack of trophic support from astrocytes may provide a possible explanation for the loss of selectively vulnerable interneuron populations in the hippocampus.

The extent of morphological transformation of microglia varied markedly in different CNS regions, but regional differences in astrocytic hypertrophy were less pronounced in mutant mice. This subtle morphological transformation of astrocytes in the presence of GFAP upregulation and selective neuronal loss may further highlight the concept of astrocytes being targeted in the disease. Conversely, there was morphological evidence for partial activation of microglia in mutant mice, which was consistent with the increased expression of F4/80 in the mutant CNS. Although partially transformed, microglia rarely reached their full extent to reveal brain macrophage morphology, not even in areas where neuronal loss and/or astrocytosis were pronounced (Chapter 3; Pontikis et al., 2004). Therefore, from these observations we can infer that microglia may also be targeted in the disease (Pontikis et al., 2004), and that cues controlling the activation of astrocytes and microglia are likely to be distinct.

116

As discussed in this chapter, the relationship between the expression of glial markers and neuronal loss is highly complex, and these events do not always correspond with one another. In order to further investigate the consequence of glial changes, F4/80 or GFAP could be conditionally knocked out in Cln3-/- mice and these mice subsequently analyzed for changes in neuronal number. Such experiments would help determine the true extent of glial involvement in neuronal loss and provide an ideal system to assess other neuropathological events within the mutant JNCL CNS.

117

CHAPTER 4

Neuropathological Changes in the Severely Affected ∆ex7/8 Cln3 Knock-in Mouse Model of JNCL

118 4.1 Introduction

The aim of this chapter is to describe the nature of neuropathological effects upon the CNS of a second mouse model of JNCL. Until recently, Cln3 null mutant mice (Cln3- /-) represented a unique in vivo model for studying JNCL pathogenesis (Katz et al., 1999; Mitchison et al, 1999). The results described in this chapter were obtained from a second, and more recently developed, mouse model of JNCL. These Cln3 ‘knock- in’ mice (Cln3∆ex7/8) accurately reproduce the 1.02 kb deletion in the CLN3 gene that is seen in over 85% of JNCL alleles (International Batten Disease Consortium, 1995). These mice also exhibit an NCL-like phenotype, which may be more aggressive than that observed in Cln3-/- mice, with a progressive neurological disease and premature death of homozygous Cln3∆ex7/8 mice from 7 months onwards (Cotman et al., 2002).

To characterize the CNS of homozygous Cln3∆ex7/8 mice, we examined neurodegenerative and reactive phenotypes that are present in other mouse models of NCL, including Cln3-/- mice (Mitchison et al., 1999; Pontikis et al., 2004). These mouse models display a range of effects upon cortical thinning and neuronal loss between sensory and motor cortex (Cooper et al., 1999; Mitchison et al., 1999; Bible et al., 2004). As ~20% of Cln3∆ex7/8 mice die by the age of 12 months (Cotman et al., 2002), this age was selected as appropriate for studying terminal stages of JNCL pathogenesis. A combination of unbiased, design-based stereological methodology and quantitative thresholding image analysis was then used to investigate the neuropathological profile of these aged Cln3∆ex7/8 mice.

4.2 Accumulation of autofluorescent storage material in homozygous Cln3∆ex7/8 mice

Macroscopic examination of the brains of 12 month old Cln3∆ex7/8 homozygotes did not reveal an obvious degenerative phenotype, with the cortex and cerebellum of affected mice appearing indistinguishable from controls (data not shown). Consistent with these observations, the brains of homozygous Cln3∆ex7/8 mice were lighter than age matched controls, but this reduction did not reach statistical significance (control

119

Figure 4.1 Accumulation of autofluorescent storage material in homozygous Cln3∆ex7/8 mice. (A-H) Representative images of unstained coronal sections through the primary motor cortex (A- D) and cerebellum (E-H) viewed by epifluorescence using rhodamine (Rho) or FITC filter sets. Sections from 12 month old homozygous Cln3∆ex7/8 mice display widespread intracellular accumulation of storage material (B-D, F-H), that fluoresces with both filter sets and appears yellow in merged images (D, H). In contrast, sections from littermate controls (+/+) display a low level of background tissue fluorescence and many fewer scattered deposits of storage material within the neocortex (A), and a low level of autofluorescence within Purkinje neurons of the cerebellum (E).

120 494±43mg; Cln3∆ex7/8 456±39mg, p=0.128, n=5). Microscopically, homozygous Cln3∆ex7/8 mice exhibited widespread intracellular accumulation of autofluorescent storage material within neuronal soma throughout the CNS (Figure 4.1). In mutant mice this storage material was present as punctate material that fluoresced at multiple wavelengths, as viewed by conventional epifluorescence microscopy (Figures 4.1B-D, F-G). In contrast, control littermates displayed a low level of background autofluorescence and an age-related and sparsely scattered accumulation of autofluorescent material (Figure 4.1A) that was more pronounced in Purkinje neurons of the cerebellum (Figure 4.1E).

4.3 Assessment of regional atrophy in homozygous Cln3∆ex7/8 mice

Since homozygous Cln3∆ex7/8 brains did not display an overt phenotype macroscopically, we carried out a stereological survey of regional volume and neocortical thickness in Nissl stained sections to look for more subtle neurodegenerative changes. Cavalieri estimates of regional volume revealed that although many CNS regions including the neocortex, hippocampus and cerebellum were reduced in size in Cln3∆ex7/8 mice, the thalamus was the only structure that displayed significantly reduced volume in homozygous Cln3∆ex7/8 mice (Figure 4.2).

4.4 Regional effects upon neocortical thinning and lamination in homozygous Cln3∆ex7/8 mice

Because measurements of overall neocortical volume cannot discriminate between events in different neocortical subfields, we carried out a series of neocortical thickness measurements of primary motor (M1), somatosensory barrel field (S1BF), primary visual (V1) and lateral entorhinal (Lent) regions (Figure 4.3). Homozygous Cln3∆ex7/8 mice showed variable effects upon cortical thickness in different regions. There was significant thinning of S1BF and Lent cortex, but a significant increase in the thickness of M1 (Figure 4.3A). In contrast, V1 exhibited less pronounced thinning

121

Figure 4.2 Unbiased Cavalieri estimates of regional volume in homozygous Cln3∆ex7/8 mice. Unbiased Cavalieri estimates of regional volume revealed no significant regional atrophy in Cln3∆ex7/8 homozygotes vs. littermate controls (+/+) at 12 months of age, with the exception of the thalamus (Thal). Regions examined include cortical mantle (Neocortex), hippocampus (Hipp), Thal, hypothalamus (Hypoth), striatum (CPu), whole cerebellum (Cb total), cerebellar vermis (Cb vermis) and lateral hemisphere (Cb lat). * p<0.05, one way ANOVA.

122

Figure 4.3 Subregion- and laminar-specific changes in neocortical thickness in homozygous Cln3∆ex7/8 mice. (A) Neocortical thickness measurements. Compared with littermate controls (+/+), thinning of the cortical mantle in Cln3∆ex7/8 homozygotes was more pronounced in primary somatosensory barrel field (S1BF) and lateral entorhinal (Lent) cortex than primary visual (V1), or primary motor (M1) cortex which exhibited a small, but significant increase in overall thickness. (B-E) Laminar thickness measurements in these neocortical regions revealed a complex series of changes in thickness of individual laminae in 12 month old homozygous Cln3∆ex7/8 mice compared with littermate controls (+/+). * p<0.05, ** p<0.01, *** p<0.001, one way ANOVA. These data are also summarized in Table 4.1.

123

I II/III III IV V VI Total Thickness

M1 NS ↓ na na ↑ NS ↑

S1BF ↓ ↓ na ↑ ↓ ↓ ↓

V1 ↓ NS na ↓ NS ↑ NS

Lent ↓ ↓ ↑ ↑ NS NS ↓

Table 4.1 Tabular depiction of significant changes in individual laminar thickness in homozygous Cln3∆ex7/8 mice. Measurements taken in the primary motor (M1), primary somatosensory barrel field (S1BF), primary visual (V1) and lateral entorhinal cortex (Lent) of 12 month old homozygous Cln3∆ex7/8 mice vs. littermate controls (+/+).↑ = significantly thicker in Cln3∆ex7/8; ↓ = significantly thinner in Cln3∆ex7/8; na = not applicable, M1 has no lamina IV. A combined measurement of laminae II and III was made in all regions except Lent where these laminae were measured separately.

124 in homozygous Cln3∆ex7/8 mice that was not significant vs. littermate controls (Figure 4.3A).

To determine whether these changes were due to laminae-specific events, we made measurements of individual laminar thickness (Figures 4.3B-E). Although significant reductions in the thickness of lamina I were consistently seen in S1BF, V1 and Lent (Figures 4.3B, C, D), and the thickness of laminae II and III in S1BF, M1, and Lent (Figures 4.3B, D, E), the remaining laminae in these neocortical regions displayed a complex series of changes in thickness (Table 4.1). The most extreme example was in Lent which displayed significant reductions in the thickness of laminae I and II, but significantly increased thickness of laminae III and IV (Figure 4.3D). The increase in the thickness of M1 in homozygous Cln3∆ex7/8 mice appeared to be largely due to a significant increase in the thickness of lamina V (Figure 4.3E), whereas all other laminae were either unchanged (I and VI) or displayed a small, but significant thinning (II and III).

4.5 Regional and laminar effects upon neuronal number in homozygous Cln3∆ex7/8 mice

To investigate whether these changes in individual laminar thickness were the result of effects upon alterations in neuronal number and/or neuronal size, we obtained optical fractionator estimates of neuronal number (Figure 4.4A) and nucleator estimates of neuronal volume (Figure 4.4B) in Nissl stained sections (Chapter 2). These data were collected for lamina V pyramidal neurons in S1BF and M1, two neocortical regions which displayed contrasting changes in the thickness of this lamina, in addition to lamina IV granule neurons of S1BF and a combined measure of laminae II and III neurons in S1BF (Figure 4.4).

Although there was a trend to reduced neuronal number, lamina V pyramidal neurons were not significantly lost in either M1 or S1BF of homozygous Cln3∆ex7/8 mice (Figure 4.4A). In contrast, there was a significant loss of neurons in lamina II and III of S1BF of mutant mice, together with a significant loss of granule neurons in lamina

125

Figure 4.4 Unbiased optical fractionator estimates of neuronal number in homozygous Cln3∆ex7/8 mice. (A) There was a significant loss of neurons in lamina II and III and lamina IV granule neurons, but no significant loss of lamina V pyramidal neurons in primary somatosensory barrelfield cortex (S1BF) or primary motor cortex (M1) of homozygous Cln3∆ex7/8 mice compared with littermate controls (+/+). There was also a significant reduction in the number of neurons in the ventral posterior thalamic nucleus (VPM/VPL) of mutant mice. (B) Nucleator estimates of neuronal volume revealed no significant atrophy or hypertrophy in any laminae of Cln3∆ex7/8 homozygotes. * p<0.05, *** p<0.001, one way ANOVA.

126 IV of this neocortical region (Figure 4.4A). Nucleator estimates of neuronal volume revealed only minor changes in cell size in individual laminae (Figure 4.4B), none of which reached statistical significance.

To determine the effect of this loss of granule neurons in S1BF upon afferent thalamic neurons, we obtained unbiased stereological estimates of neuronal number and volume in the ventral posterior thalamic nucleus (VPM/VPL). These analyses revealed a significant loss of neurons within VPM/VPL of homozygous Cln3∆ex7/8 mice (Figure 4.4A), although these neurons were not significantly hypertrophied (Figure 4.4B).

4.6 Survival of GABAergic interneurons in homozygous Cln3∆ex7/8 mice

Loss of GABAergic interneurons is a common feature of other murine models of NCL (Cooper, 2003; Mitchison et al., 2004). To determine whether these neuronal populations were also affected in homozygous Cln3∆ex7/8 mice, the expression of parvalbumin (PV) and somatostatin (SOM) were surveyed in the neocortex (Figure 4.5A) and hippocampus (Figures 4.5B, C). These markers are usually expressed in representative interneuron populations (Freund and Buzsáki, 1996), that are consistently affected in mouse models of infantile NCL (Bible et al., 2004; Jalanko et al., 2005), juvenile NCL (Mitchison et al., 1999) and variant late infantile NCL (Cooper et al., 1999; Kopra et al., 2004).

Primary motor and somatosensory cortex. Optical fractionator estimates revealed a general trend to a reduced number of interneurons immunoreactive for PV and SOM in M1 and S1BF, but this loss of interneurons did not reach significance for interneurons stained for either antigen in these neocortical regions (Figure 4.5A).

Hippocampus. We next examined interneuron populations that are immunoreactive for these markers in the hippocampal formation of homozygous Cln3∆ex7/8 mice. There was an overall trend to reduced number of SOM-positive interneurons in homozygous

127

Figure 4.5 Interneuron counts in homozygous Cln3∆ex7/8 mice. (A) Persistence of neocortical interneurons in aged Cln3∆ex7/8 homozygotes. Optical fractionator estimates of parvalbumin (PV) and somatostatin-(SOM) positive interneuron number in the primary motor (M1) and somatosensory barrelfield (S1BF) cortex revealed no significant loss of these neurons in 12 month old Cln3∆ex7/8 homozygotes compared with littermate controls (+/+). (B-C) Survival of hippocampal interneurons in aged homozygous Cln3∆ex7/8 mice. Counts of hippocampal interneuron number revealed no significant loss of interneurons immunoreactive for either parvalbumin (B) or somatostatin (C) in any subfield of 12 month old Cln3∆ex7/8 and age matched controls. Hippocampal subfields include the dentate gyrus (DG), hilus (Hi), CA3/2/1, stratum oriens (Oriens) and stratum radiatum (Radiatum).

128 Cln3∆ex7/8 mice, although this did not reach significance in any hippocampal subfield (Figure 4.5C). Effects upon the number of PV-positive interneurons were more variable between hippocampal subfields in homozygous Cln3∆ex7/8 mice, but none of these changes reached statistical significance (Figure 4.5B).

4.7 Regionally restricted astrocytic and microglial activation in homozygous Cln3∆ex7/8 mice

To examine glial responses in homozygous Cln3∆ex7/8 mice, we first surveyed the distribution of cells immunoreactive for the astrocytic markers GFAP and S100β, in addition to the microglial marker F4/80. Subsequently, we used image analysis to quantify the expression of these markers in the striatum, neocortical subregions M1 and S1BF; and hippocampal subfields Hilus, CA1, CA2 and CA3 using standard methods (Bible et al., 2004). Reactive astrocytosis characteristically involves hypertrophy of astrocytes (thickening of processes and changes in cell body volume) which may be independent of cell proliferation. Therefore, we also obtained optical fractionator and nucleator estimates the number and size of astrocytes immunoreactive for S100β, a calcium binding protein that is expressed predominantly in astrocytes (Boyes et al., 1986).

Qualitative assessment of GFAP and S100β-positive astrocytes: In the neocortical grey matter of control mice only few GFAP-positive astrocytes were present, predominantly in lamina VI and in lamina I adjacent to the pial surface (Figure 4.6A). In marked contrast, homozygous Cln3∆ex7/8 mice displayed a profound astrocytosis across the cortical mantle with intensely GFAP-immunoreactive astrocytes present in both superficial and deeper neocortical laminae (Figure 4.6B), with positively stained astrocytes extending in clusters across all laminae. At higher power these GFAP- positive astrocytes exhibited the typical morphology of protoplasmic astrocytes in control brains, but in Cln3∆ex7/8 homozygotes GFAP-positive astrocytes displayed marked hypertrophy with numerous thickened and branched processes associated with the morphology of fibrous astrocytes (Figures 4.6A, B; Graeber and Kreutzberg, 1986). In marked contrast to GFAP staining, S100β immunoreactivity within the

129

Figure 4.6 Astrocytic responses in homozygous Cln3∆ex7/8 mice. (A, B) Immunohistochemical staining for GFAP revealed profound astrocytosis in the neocortex of 12 month old Cln3∆ex7/8 homozygotes (B) compared with age matched controls (+/+, A). Within the primary motor cortex (M1) many hypertrophic GFAP-positive astrocytes bearing numerous thickened processes were present in both deep and superficial laminae of mutant mice (B). In contrast, there were no obvious differences in the distribution or number of S100β-positive astrocytes in M1 of animals of either genotype (C, D). Compared to controls (E) the thalamus of Cln3∆ex7/8 homozygotes exhibited pronounced astrocytosis that was confined to individual thalamic nuclei (F), with GFAP-positive astrocytes prominent in the ventral posterior (VPM/VPL), mediodorsal (MD) and lateral regions of the laterodorsal (LD) thalamic nuclei, but virtually absent in the adjacent ventromedial (VM), posterior (Po) and centrolateral (CL) thalamic nuclei.

130 neocortex did not differ obviously between animals of either genotype and was present within numerous intensely stained cell soma that were distributed evenly across laminae (Figures 4.6C, D).

Subcortically this astrocytic response was far less pronounced in homozygous Cln3∆ex7/8 mice, with reduced staining for GFAP within the hippocampus and cerebellum of these mutant mice (data not shown). However, the thalamus of Cln3∆ex7/8 homozygotes exhibited prominent GFAP immunoreactivity with astrocytic hypertrophy, in contrast to a more subtle staining in control mice (Figure 4.6E, F). However, this staining for GFAP in mutant mice was not uniformly distributed but was most intense within individual thalamic nuclei, most notably within the ventral posterior (VPM/VPL), lateral regions of the laterodorsal (LD) and the mediodorsal (MD) thalamic nuclei, but virtually absent in the adjacent ventromedial (VM), posterior (Po) or intralaminar thalamic nuclei (Figure 4.6F).

Quantitative analysis of astrocytosis: Consistent with these morphologic observations, quantitative image analysis revealed a significant and widespread increase in GFAP expression in both superficial and deeper neocortical layers (Figure 4.7A). This upregulation was not confined to any neocortical region with similar increases in GFAP staining in M1, S1BF, V1 and Lent (Figure 4.7A). In contrast to these events in the neocortex, quantitative analysis revealed a variable expression of GFAP in the hippocampus (Figure 4.7B), cerebellum (Figure 4.7C) and subcortical structures of Cln3∆ex7/8 homozygotes. Unexpectedly, the hippocampal dentate gyrus, stratum oriens and CA1 of homozygous Cln3∆ex7/8 mice exhibited significantly reduced GFAP expression compared to control littermates (Figure 4.7B). GFAP expression in homozygous Cln3∆ex7/8 mice was similarly reduced in the grey matter of cerebellar lateral hemisphere and vermis, but was unchanged in the white matter within these regions of the cerebellum (Figure 4.7C).

To determine whether this altered GFAP expression reflected changes in the number of astrocytes we next obtained unbiased stereological estimates of the number of S100β-positive astrocytes. These optical fractionator data were collected within one representative region that displayed increased GFAP expression (S1BF) and one that

131

Figure 4.7 Quantitative assessment of regional astrocytosis in homozygous Cln3∆ex7/8 mice at 12 months of age. (A) Thresholding image analysis revealed the widespread and significantly increased expression of GFAP in the superficial and deep neocortical laminae of Cln3∆ex7/8 homozygotes compared with littermate controls (+/+) at 12 months of age. (B-C) In contrast, hippocampal (B) and cerebellar grey matter (C) display significantly reduced levels of GFAP in mutant mice compared with controls. Areas surveyed include neocortical primary motor (M1), somatosensory barrelfield (S1BF) and visual (V1); striatum (CPu), hippocampal dentate gyrus (DG), hippocampal stratum oriens and CA1 (Oriens CA1); cerebellar white matter (WM) and molecular and granule cell layers (Mol+Gr) in the lateral hemispheres (Lat) and vermis (Verm). (D) Optical fractionator estimates revealed no significant change in the number of S100β-positive astrocytes in either S1BF or oriens and CA1 of homozygous Cln3∆ex7/8 mice, compared with littermate controls.*** p<0.001, one way ANOVA.

132 displayed decreased GFAP expression (stratum oriens and CA1 of the hippocampus), but did not reveal any significant change in astrocyte number within either region of homozygous Cln3∆ex7/8 mice compared with control littermates (Figure 4.7D).

F4/80-positive microglia: Compared to control littermates, F4/80-immunoreactive microglia were consistently more prominent in homozygous Cln3∆ex7/8 mice, with these differences again more prominent in the neocortex than other CNS regions (Figure 4.8). In the neocortex of control littermates the widespread distribution of microglia was revealed by pale F4/80 immunoreactivity within the cell soma and thin processes extending into the neuropil (Figures 4.8A, C). In contrast, throughout the cortical mantle of Cln3∆ex7/8 homozygotes there were intensely F4/80-immunoreactive microglia with numerous ramified processes, often with more prominent cell soma (Figures 4.8B, D). In homozygous Cln3∆ex7/8 mice these F4/80-immunoreactive microglia were present with no particular focus and across all laminae, although F4/80 positive cells frequently displayed more morphological signs of activation in deeper laminae with fewer thickened processes.

Differences in the relative intensity of F4/80 staining between genotypes were less marked in subcortical structures (Figures 4.8E, F), although homozygous Cln3∆ex7/8 mice again displayed more morphological evidence of microglial activation in these regions. Although complete morphological transformation to brain macrophage-like morphology in homozygous Cln3∆ex7/8 mice was seldom seen, these mice frequently displayed F4/80-immunoreactive microglia with enlarged soma and numerous short thickened processes (Figure 4.8F), compared to the threadlike and ramified processes of microglia in control mice (Figure 4.8E).

4.8 Discussion

Characterization of the neurodegenerative and glial phenotypes of aged homozygous Cln3∆ex7/8 mice has enabled the detailed mapping of neuropathological events in this genetically accurate model of JNCL. Since these mice are nearing terminal stages of the disease process, this study may provide key information of which parts of the

133

Figure 4.8 Microglial responses in homozygous Cln3∆ex7/8 mice. Immunohistochemical staining for F4/80 revealed graded activation of microglia in 12 month old homozygous Cln3∆ex7/8 mice compared with littermate controls (+/+). At this age, F4/80-positive microglia with numerous ramified processes in the primary motor cortex (M1) were more darkly stained in Cln3∆ex7/8 homozygotes (B, D) than littermate controls (A, C). This difference between genotypes was less pronounced in the striatum (CPu), although numerous partially activated microglia with enlarged soma and short thickened processes were evident in homozygous Cln3∆ex7/8 mice (F) compared to the microglia with thin ramified processes that were present in littermate controls (E).

134 JNCL CNS are affected, and may subsequently lead to the more effective targeting of potential therapeutic interventions.

Results from these analyses demonstrated a widespread accumulation of autofluorescent storage material throughout homozygous Cln3∆ex7/8 brains, although atrophy was very mild, only reaching significance in the thalamus. Despite the lack of global atrophy of the neocortex, thickness changes were observed in almost all neocortical subfields. Closer examination of these subfields revealed changes to extend to individual laminae. Indeed, the most noticeable of these thickness changes was of lamina V, which consistently followed patterns of overall neocortical thickness in M1 and S1BF of aged mutants.

To determine whether these changes in laminar thickness resulted from effects on neuronal survival and/or neuronal volume, a stereological analysis of neuronal populations was undertaken within a sample of affected laminae. Changes in neuronal number and volume did not correspond directly with effects upon in laminar thickness in mutant mice. Instead, neuronal populations from S1BF were selectively targeted in the mutant CNS, with the sparing of pyramidal neurons in lamina V. Moreover, neuronal volume was similar in animals of either genotype. Thus, laminar thickness changes are likely to result from a complex series of pathological events, such as altered packing between cells and effects on dendritic arborisation, each of which would require further investigation. In this manner, these parameters could be investigated using stereological analysis of Nissl stained sections and threshold image analysis of Golgi staining or intracellular filling techniques.

The most pronounced loss of somatosensory neurons was evident in laminae II and III, which supply commissural and association projections to other areas of neocortex (Amaral, 2000). Neuronal loss within these laminae may potentially have significant consequences for coordinating neuronal activity between hemispheres and neocortical regions. There was also a loss of lamina IV granule neurons in S1BF, which is where most of the thalamic input terminates (Lopez-Bendito and Molnar, 2003). As there was an additional loss of thalamic relay neurons in VPM/VPL, the loss of these granule neurons may have deleterious effects upon their afferent neuronal populations. Conversely, loss of thalamic relay neurons may affect the survival of

135 their targets in the neocortex. The precise cause of neuronal loss in the thalamocortical system is unknown. However, the lack of neurotrophic support from target neurons has been implicated in the survival of afferent neuron populations in the developing brain (Oppenheim, 1989). Indeed, neurons become more dependent on trophic factors for survival during development (Rodriguez-Tebar et al., 1989) and their dependency has been shown to coincide with the timing of target innervation (Johnson et al., 1986; Cohen-Cory and Fraser, 1994).

Since the loss of GABAergic interneurons is common to human JNCL (Braak and Goebel, 1978; Braak and Goebel, 1979; Tyynelä et al., 2004) and the Cln3-/- mouse model of the disease (Chapter 3; Pontikis et al., 2004), representative populations of these cells were also quantified within the neocortex and in hippocampal subfields of aged Cln3∆ex7/8 homozygotes. In marked contrast to observations in human and murine JNCL, interneuron populations were largely spared in Cln3∆ex7/8 mice. Perhaps this survival of interneurons may be characteristic of these mice and reflect the different strain background, or alternatively, this phenotype may take longer to develop in this model of JNCL.

Neurodegenerative events that have been described in homozygous Cln3∆ex7/8 mice were accompanied by a widespread astrocytosis and more graded microglial activation. GFAP immunoreactivity revealed prominent astrocytosis in the neocortex, striatum and individual thalamic nuclei, with numerous hypertrophied astrocytes with thickened branched processes. However, there was little evidence of astrocytic proliferation or hyperplasia, with no difference in the number of S100β-positive astrocytes between animals of either genotype. A reduction of GFAP-positive staining in Cln3∆ex7/8 mice is therefore more likely to represent a downregulation of this marker on a cell to cell basis, rather than a loss of astrocytes themselves. Importantly, patterns of GFAP immunoreactivity were not uniform across all regions of the mutant CNS, but different in magnitude and direction between CNS regions. For example, in contrast to events in the neocortex, astrocytic responses were diminished in the hippocampus and cerebellum. Taken together with data from S100β, these findings argue for a down regulation of the cues that regulate astrocytosis within the hippocampus, rather than a loss of astrocytes within this region.

136 Cues mediating astrocytosis are poorly understood in JNCL, but there was a close correlation between the extent of astrocytosis and neuronal loss within the neocortex and thalamus of Cln3∆ex7/8 mice. Since F4/80-positive staining was also prominent in the neocortex, it is possible that injured neurons may have secreted cues that were able to recruit microglia into the surrounding area, where they were subsequently activated and capable of secreting excitotoxic levels of glutamate (Heppner et al., 1998; Koutsilieri et al., 1999; Loddick and Rothwell, 2002). Indeed, it has already been demonstrated that glutamate levels are elevated within the JNCL CNS (Chattopadhyay et al., 2002a). As such, the astrocytic response within the neocortex may represent an attempt to clear extracellular glutamate (Rothstein et al., 1996; Duan et al., 1999) and offer neuroprotection. In contrast, subcortical structures that exhibited astrocytosis were not accompanied by an increased expression of F4/80, as described in the neocortex. Nevertheless, there was morphological evidence for partial activation of microglia in mutant mice, although this rarely reached its full extent to reveal brain macrophage morphology. The inability for microglia to become fully activated, even in regions of astrocytosis or neuronal loss suggests that microglia may themselves be targeted in the disease. Furthermore, cues controlling the activation of astrocytes and microglia may be similar in the neocortex, but are likely to be distinct in subcortical structures.

Although astrocytosis is thought offer neuroprotection (Simard and Nedergaard, 2004) and microglial activation has been implicated in degeneration within the CNS (Gebicke-Haerter, 2001), microglia are also capable in assisting neuronal recovery, but only when repair is still a feasible option (Kreutzberg, 1996). Thus the precise relationship between changes in glial populations and neuronal loss is unclear within the context of homozygous Cln3∆ex7/8 mice. Instead, neuropathological events are likely to be multifactorial, and the complex balance and interplay of these events may subsequently lead to neuronal loss.

137

CHAPTER 5

Early Neuropathological Changes in the Cln3-/- Mouse Model of JNCL

138 5.1 Introduction

The second aim of this thesis is to define the nature and timing of pathological effects upon the CNS of Cln3-/- and homozygous Cln3∆ex7/8 mouse models of JNCL. Very little is known about the early stages of JNCL pathogenesis. It is unknown whether there is a period of normal development prior to the subsequent progression of neuropathological changes that are evident in this disorder. Equally, it is unknown precisely when these pathological changes first become apparent. As such, early evaluation of disease phenotypes will help us to determine when the first neuropathological events occur, and reveal whether development progresses normally in this disorder.

This chapter addresses these goals in Cln3-/- mice at 8 days of age, as this represents an early postnatal period during which the cerebellum is still developing (Wang and Zoghbi, 2001) and many events are still underway in the cortex, i.e. patterns of connectivity are being fine tuned (Sur and Leamey, 2001) and radial glia are differentiating either into astrocytes or neurons (Voigt, 1989; Gotz et al., 2002; Tramontin et al., 2003). In order to assess these parameters, we examined neurodegenerative and reactive phenotypes that are known to be present in older Cln3- /- mice (Chapter 3), as well as other mouse models of NCL (Cooper et al., 1999; Mitchison et al., 1999; Gupta et al., 2001; Bible et al., 2004; Kopra et al., 2004; Pontikis et al., 2004; Janlanko et al., 2005). A combination of unbiased, design-based stereological methodology and thresholding image analysis of immunohistochemical markers was used to investigate appropriate phenotypes within the early postnatal CNS of these mutant mice.

5.2 Evaluation of regional volume in P8 Cln3-/- mice

Macroscopically, the brains of P8 Cln3-/- mice had no obvious phenotypic abnormalities and appeared indistinguishable from controls (data not shown). In order to explore more subtle changes within the CNS, we first carried out a stereological survey of regional volume in Nissl stained sections. The Cavalieri method (Chapter 2)

139

Figure 5.1 Unbiased Cavalieri estimates of regional volume in Cln3-/- mice. Unbiased Cavalieri estimates of regional volume revealed significant atrophy of the neocortex and hippocampus (Hipp), but all other regions were spared in P8 Cln3-/- vs. littermate controls (+/+). Regions examined include cortical mantle (Neocortex), hipp, thalamus (Thal), hypothalamus (Hypoth), striatum (CPu), whole cerebellum (Cb). ** p<0.01, *** p<0.001, one way ANOVA.

140 was used to obtain unbiased estimates of volume for the cortical mantle, hippocampus, striatum, thalamus and cerebellum (Figure 5.1). Unexpectedly, the cortical mantle and hippocampal formation exhibited significantly reduced volume in P8 Cln3-/- mice vs. littermate controls. In addition, there was a general trend for all other regions of the P8 Cln3-/- CNS to be smaller than in littermate controls, but these reductions did not reach statistical significance.

5.3 Regional effects upon neocortical thinning and lamination in P8 Cln3-/- mice

Since measurements of overall neocortical volume cannot discriminate between changes within individual neocortical subfields, a series of thickness measurements were made in neocortical regions at different rostrocaudal levels of the CNS. These measurements were performed in the primary motor (M1), somatosensory barrel field (S1BF), primary visual (V1) and lateral entorhinal (Lent) cortex (Figure 5.2A, Table 5.1). Cln3-/- mice exhibited variable degrees of reduced neocortical thickness, which was significant in all subfields examined. However, thinning of the neocortex was more pronounced rostrally in M1, with a slightly more modest atrophy in S1BF and V1. In marked contrast, Lent displayed a significant increase in thickness in Cln3-/- mice when compared to control littermates.

To determine whether these changes in neocortical thickness were due to laminae specific events, measurements were made for individual laminar thickness in these neocortical regions (Figures 5.2B-E, Table 5.1). Deeper laminae (V and V1) generally followed patterns of overall neocortical thickness, with significantly reduced thickness in M1 (Figure 5.2E), S1BF (Figure 5.2B) and V1 (Figure 5.2C), but a significant increase in thickness in Lent (Figure 5.2D). These events in deeper laminae provided the greatest contribution to changes in overall neocortical thickness in Cln3-/- mice vs. age matched controls. Lamina IV in P8 mutants mirrored the thickness changes exhibited by lamina V and VI in all neocortical subfields except for V1, which was significantly thicker than in control mice. A complex series of contrasting changes were observed for the thickness of laminae II and III in Cln3-

141

Figure 5.2 Subregion- and laminar-specific changes in neocortical thickness in Cln3-/- mice. (A) Neocortical thickness measurements. Compared with littermate controls (+/+), thinning of the cortical mantle in Cln3-/- mice was more pronounced in primary motor (M1), primary somatosensory barrel field (S1BF) and primary visual cortex (V1) than lateral entorhinal (Lent), which exhibited a significant increase in overall thickness. (B-E) Laminar thickness measurements in these neocortical regions revealed a complex series of changes in thickness of individual laminae in P8 Cln3-/- mice compared with littermate controls. * p<0.05, ** p<0.01, *** p<0.001, one way ANOVA. These data are also summarized in Table 5.1.

142

I II/III III IV V VI Total Thickness

M1 ↑ ↓ na na ↓ ↓ ↓

S1BF ↑ ↑ na ↓ ↓ ↓ ↓

V1 ↑ ↑ na ↑ ↓ ↓ ↓

Lent NS NS NS ↑ ↑ ↑ ↑

Table 5.1 Tabular depiction of significant changes in individual laminar thickness in Cln3-/- mice. Measurements taken in the primary motor (M1), primary somatosensory barrel field (S1BF), primary visual (V1) and lateral entorhinal cortex (Lent) of P8 Cln3-/- mice vs. littermate controls (+/+).↑ = significantly thicker in Cln3-/-; ↓ = significantly thinner in Cln3-/-; na = not applicable, M1 has no lamina IV. A combined measurement of laminae II and III was made in all regions except Lent where these laminae were measured separately.

143 deficient mice. These laminae were significantly thinner in M1, thicker in S1BF and V1, and unchanged in Lent (Table 5.1). Lamina I was also unaffected in Lent, but appeared thicker in all other neocortical subfields in Cln3-/- mice, when compared to control littermates. Since neocortical subfields located more rostrally in the brain were typically thicker in superficial laminae, and thinner in deeper laminae, this may represent a problem with neuronal migration that is delayed rostrally and more advanced caudally in the brain. Alternatively, cues for neuronal packing may themselves be dysfunctional to varying extents along the rostrocaudal gradient.

5.4 Regional and laminar effects upon neuronal number in P8 Cln3-/- mice

To investigate whether changes in individual laminar thickness in P8 Cln3-/- mice were due to effects upon neuronal number and size, we obtained optical fractionator estimates of neuronal number (Figure 5.3A) and nucleator estimates of neuronal volume (Figure 5.3B) in Nissl stained sections (Chapter 2). These analyses focussed upon lamina V of M1 and components of the thalamocortical system which appeared thinner in P8 Cln3-/- mice and exhibited selective cell loss in older mutants (Chapter 3).

There was a slight reduction in the number of lamina V pyramidal neurons in M1 of P8 Cln3-/- mice, but this did not reach significance compared to littermate controls (Figure 5.3A). Nucleator estimates of neuronal volume demonstrated only minor changes in the size of these neurons (Figure 5.3B). Neuronal number in the thalamus was also only subtly compromised in Cln3-/- mice, but again did not reach statistical significance (Figure 5.3A). As described previously, neurons from the ventral posterior thalamic nucleus (VPM/VPL) relay sensory information to granule neurons within lamina IV of S1BF. The VPM/VPL also receives sensory feedback from lamina V pyramidal neurons of S1BF. Although different components of this system are selectively vulnerable later in JNCL pathogenesis, these neuron populations were not compromised in terms of neuronal number and size in Cln3-/- mice in the early postnatal period (Figures 5.3A, B).

144

Figure 5.3 Unbiased optical fractionator estimates of neuronal number in Cln3-/- mice. (A) There was no significant loss of neurons in any subfield that was analysed in Cln3-/- mice compared with littermate controls (+/+). (B) Nucleator estimates of neuronal volume did not reveal any significant atrophy or hypertrophy in Cln3-/- mice. Neurons examined include lamina IV granule neurons in primary somatosensory barrelfield cortex (S1BF), lamina V pyramidal neurons in S1BF and primary motor cortex (M1), and neurons in the ventral posterior thalamic nucleus (VPM/VPL).

145 5.5 Astrocytic and microglial responses in P8 Cln3-/- mice

We have established that there are glial responses in the CNS of aged Cln3-/- mice, which are complex and regionally selective (Chapter 3). It is unclear when such changes are first apparent or how these events relate to subsequent neuronal loss. To examine early glial responses in the Cln3-/- mouse, one week old mutant mice were assessed for the distribution of astrocytic marker GFAP and microglial marker F4/80. The expression of these glial markers were quantified using thresholding image analysis (Chapter 2), in neocortical subregions M1, S1BF and V1; subcortical structures, thalamus and striatum; and hippocampal subfields dentate gyrus (DG) and stratum oriens together with CA1 (OrCA1).

Qualitative assessment of GFAP-positive astrocytes: In the neocortical grey matter of P8 control mice, GFAP-positive astrocytes were present, predominantly located in lamina VI and in lamina I adjacent to the pial surface (Figures 5.4A, C). Additional staining of astrocytes was noted throughout intermediate laminae, but was scattered and less abundant. Radial glia were also present in this region and easily identified by their unique morphology and location. Their cell morphology consisted of a long radial fibre spanning the width of the neocortex from the corpus callosum to the pial surface. The soma of these radial glia were not apparent, but may have been obscured by GFAP-positive astrocytic processes within deeper laminae.

Eight day old Cln3-/- mice exhibited a more pronounced astrocytosis than control mice within superficial neocortical laminae, extending into laminae II, but was most prominent in V1 (Figures 5.4B-D). GFAP-positive astrocytes within deeper laminae were evenly distributed and encompassed both lamina V and VI. Apart from their processes, astrocytes were seldom noted in intermediate laminae. GFAP-positive radial glia were more abundant in mutant mice when compared to age matched controls. This increased prevalence of radial glia in mutant mice may represent a delay in the differentiation of these cells into neurons or GFAP-expressing astrocytes. Higher magnification in the neocortical grey matter revealed that GFAP-positive astrocytes were typically protoplasmic in appearance, although there was there was a subtle transformation to a fibrous morphology in animals of either genotype. In

146

Figure 5.4 Astrocytic responses in Cln3-/- mice. (A-D) Immunohistochemical staining for

GFAP revealed that radial glia were more abundant in the neocortex of P8 Cln3-/- mice (B) compared with age matched controls (+/+, A). Furthermore, astrocytosis was more prominent in superficial (D) and deeper laminae of the neocortex in Cln3-/- (B, D) vs. control mice (A, C). Nevertheless, there was similar branching of astrocytic processes in animals of either genotype (C, D). (F) A less pronounced astrocytosis was present within the hippocampus (Hipp) of mutant mice and displayed weaker staining, smaller soma, and thinner processes than controls (E). V1=Primary visual cortex

147 control brains, astrocytes were darkly stained with numerous, thickened processes (Figure 5.4C), whereas P8 Cln3-/- displayed more numerous astrocytic processes, but with weaker staining than in control mice (Figure 5.4D). Conversely, radial glia in mutant mice (Figure 5.4D) were more intensely GFAP-immunoreactive than control littermates (Figure 5.4C).

Subcortically, the astrocytic response in Cln3-deficient mice was less pronounced than in the neocortex. However, GFAP-positive astrocytes were evenly distributed within the striatum and thalamus, and did not differ in patterns of distribution or staining intensity between Cln3-/- and control mice (data not shown). In marked contrast, GFAP-positive astrocytes were particularly pronounced within the hippocampus of control mice (Figure 5.4E), but were stained less intensely in mutant mice (Figure 5.4F). At higher power, hippocampal astrocytes in control mice displayed intensively dark staining, with numerous thickened and branched processes (Figure 5.4E). There was a frequent occurrence of darkly stained halo-like structures that appear to represent perivascular staining of astrocyte foot processes (Figure 5.4E). GFAP-positive astrocytes in Cln3-/- mice were less intensively stained, had thinner processes and markedly smaller soma (Figure 5.4F). Furthermore, perivascular GFAP staining was less apparent in mutant brains, which may suggest that the blood brain barrier has not formed properly at this age, but this hypothesis will require further investigation.

Quantitative analysis of GFAP immunoreactivity: Contrasting patterns of GFAP immunoreactivity were evident between neocortical (Figure 5.5A), hippocampal (Figure 5.5B) and cerebellar subfields (Figure 5.5C), with increased GFAP expression in the neocortex and cerebellum, but reduced immunoreactivity in the hippocampal formation of mutant mice. In contrast, GFAP immunoreactivity in subcortical structures was unchanged between animals of either genotype (Figure 5.5B). In the neocortex, M1 and V1 exhibited a significant upregulation of GFAP in Cln3-/- mice, which was more pronounced in V1, with a ~3 fold increase in GFAP-positive staining in mutant mice. Such changes were more subtle in S1BF of mutant mice and did not reach statistical significance. GFAP immunoreactivity was also significantly increased in almost all parts of the cerebellum in mutant mice, the sole exception being the grey

148

Figure 5.5 Quantitative assessment of regional astrocytosis in P8 Cln3-/- mice. (A) Thresholding image analysis revealed the significantly increased expression of GFAP in the primary motor (M1) and visual (V1), but not in the somatosensory barrel field cortex (S1BF) of P8 Cln3-/- compared with littermate controls (+/+). (C) Similarly, this increased expression of GFAP was present within all cerebellar subfields, except for grey matter in the lateral hemispheres (Mol+Gr Lat). (B) In contrast, hippocampal subfields exhibited significantly reduced levels of GFAP in mutant mice compared with controls. Areas surveyed include neocortical M1, S1BF and V1; subcortical thalamus and striatum (CPu); hippocampal dentate gyrus (DG), hippocampal stratum oriens and CA1 (Oriens CA1); cerebellar white matter (WM) and molecular and granule cell layers (Mol+Gr) in the lateral hemispheres (Lat) and vermis (Verm). (D). ** p<0.01, *** p<0.001, one way ANOVA.

149 matter of the lateral hemispheres. As with aged Cln3-/- mice (Chapter 3), there was a contrast in the expression of GFAP-positive staining between the neocortex and the hippocampus, with significantly lower levels of GFAP immunoreactivity in DG and OrCA1 of P8 mutant mice when compared to control littermates.

Qualitative assessment of F4/80-positive microglia: F4/80 staining was evenly distributed within each region of the CNS in both mutant and control littermates. Higher magnification of the neocortex revealed F4/80-positive microglia, with small soma and numerous thin ramified processes extending into the neuropil in animals of either genotype (data not shown). However, there were notable exceptions to this trend, for example, in V1 of controls there was an infrequent occurrence of F4/80- positive microglia that were partially transformed into a brain macrophage-like morphology. These occasional microglia in V1 had slightly larger soma and shorter thickened processes, but their occurrence was similar between individual laminae (Figures 5.6A, C). In Cln3-/- mice, V1 had a more pronounced staining of F4/80- positive microglia, with a subset of microglia displaying partial morphological transformation to a brain macrophage-like morphology. This morphological appearance occurred with a far greater incidence than in controls (Figures 5.6B, D), with microglia exhibiting larger soma and numerous short thickened processes that were more intensely stained in mutant mice (Figure 5.6D).

Subcortical structures showed a prominent reduction in the intensity of F4/80 immunoreactivity in mutant vs. control littermates (Figures 5.6E, F). Although relatively weak staining of F4/80-positive microglia was evident in Cln3-/-, there was still occasional evidence for partial transformation into a brain-like macrophage morphology. In the thalamus, this was observed within the laterodorsal (LD) nucleus (Figure 5.6F), but in the hippocampus and striatum, such transformations were not restricted to specific subfields. In marked contrast, partial transformation of microglia in controls was observed extensively throughout subcortical structures (Figure 5.6E).

Quantitative analysis of F4/80 immunoreactivity: Thresholding image analysis revealed a complex series of changes in the expression of F4/80 immunoreactivity, which were consistent with qualitative observations. Neocortically, V1 exhibited a significant increase of F4/80 immunoreactivity that was approximately two fold

150

Figure 5.6 Microglial responses in Cln3-/- mice. Immunohistochemical staining for F4/80 revealed regionally variable changes in microglia in P8 Cln3-/- mice compared with littermate controls (+/+).There was little morphological evidence for microglial activation, with the exception of the primary visual cortex (V1) where F4/80-positive microglia displayed more pronounced staining and had numerous short, ramified processes in Cln3-/- (B, D) vs. controls (A, C). In marked contrast, F4/80 positive staining within the thalamus (Thal) was less pronounced in Cln3-/- mice (F) and microglia rarely adopted a macrophage-like morphology in animals of either genotype (E, F). Thalamic nuclei include VPM/VPL, LD, VM, posterior (Po), centrolateral (CL) and mediodorsal (MD).

151

Figure 5.7 Quantitative assessment of microglial responses in Cln3-/- mice. (A) Thresholding image analysis revealed the significantly increased expression of GFAP in the primary visual (V1), but not in motor (M1) or the somatosensory barrel field (S1BF) cortex of P8 Cln3-/- compared with littermate controls (+/+). (B) In contrast, subfields of the hippocampus and subcortex exhibited significantly reduced levels of F4/80 in mutant mice. (C) Cerebellar subfields were highly variable, exhibiting a significant increase of F4/80-positive staining in white matter of the lateral hemispheres (WM Lat) and grey matter of the vermis (Mol+Gr Verm), but a significant lower expression within grey matter of the lateral hemispheres (Mol+Gr Lat). Areas surveyed include neocortical M1, S1BF and V1; subcortical thalamus and striatum (CPu), hippocampal dentate gyrus (DG), hippocampal stratum oriens and CA1 (Oriens CA1); cerebellar molecular and granule cell layers (Mol+Gr) in the lateral hemispheres (Lat). *** p<0.001, one way ANOVA.

152 greater in Cln3-/- mice vs. age matched controls (Figure 5.7A). Negligible changes in expression patterns were noted in other neocortical subfields, including M1 and S1BF. Conversely, within subcortical structures of mutant mice there was a significantly lower level of F4/80-positive staining in the striatum, thalamus and within hippocampal DG and OrCA1 (Figure 5.7B). Of these immunoreactive changes, microglia in the hippocampal DG underwent the most noticeable change with a ~6 fold decrease of F4/80 immunoreactivity when compared to age matched controls. There were contrasting patterns of F4/80 immunoreactivity within the cerebellum between animals of either genotype (Figure 5.7C). These included a significant increase and reduction of F4/80 staining in the white and grey matter of the lateral hemispheres respectively, in mutant mice. There was also significantly increased F4/80 staining in the grey of within the vermis of mutant mice, but only negligibly changed F4/80 staining in the white matter.

5.6 Discussion

In this chapter, the evaluation of neuronal and glial phenotypes in P8 mice revealed unexpectedly early effects upon the JNCL CNS. A detailed understanding of these events may reveal them to be a particularly early neuropathological phenotype, or alternatively to reflect a developmental abnormality. In either instance, such data may prove invaluable for the strategic timing and placement of potential therapeutic interventions in this well established mouse model of JNCL.

Surprisingly, microscopic analysis of Cln3-/- brains revealed a reduced volume of the cortical mantle and hippocampus in Cln3-deficient mice, but with relative sparing of other regions in the CNS when compared to control littermates. Closer examination of the neocortex of mutant mice revealed thinning of subfields that was greatest rostrally in M1 and became thicker in V1. These complex changes extended to individual laminae, where deeper laminae (V and VI) provided the greatest contribution to overall thickness changes within neocortical subfields. Although changes in volume and thickness have been extensively explored in the CNS of Cln3-/- mice, it still remains unclear what this range of changes actually means. One can postulate that

153 these phenotypic abnormalities may be due to the loss of neurons. Alternatively this neocortical thinning may represent a delay in the growth of these specific parts of the brain. At a cellular level within the neocortex, this phenotype may have many underlying causes. For example, corticogenesis may itself be dysfunctional, involving either abnormal generation of neurons, improper migration of neocortical neurons, or defective neuronal connectivity. It will be important to explore each of these possibilities in future studies.

We have already shown that older Cln3-/- mice display a selective loss of neuronal populations from different compartments within the brain (Chapter 3). Moreover, neuronal loss mostly corresponded to a reduced thickness of that region. In this present study, we focussed upon lamina V of M1 and components of the thalamocortical system which appeared thinner in P8 Cln3-/- mice and exhibited selective cell loss in older mutants (Chapter 3). Analysis of one week old mutants did not reveal the loss of any analysed neuronal population. Furthermore, the size of these neurons did not vary between animals of either genotype. These data suggest that neuronal populations are correctly targeted to their corresponding laminae during development in mutant mice, in a similar manner to control mice. As such, the changes in laminar thickness may depend on other factors, such as the packing of neurons within individual laminae or effects upon dendritic arborisation, each of which warrant further investigation.

We have already established that there is a defined series of reactive changes in the CNS of aged Cln3-/- mice (Chapter 3). Until recently, it has been unclear whether such reactive events are an early event during JNCL pathogenesis, or only occur in the later stages of disease progression. Data from a sheep model of NCL have suggested that marked astrocytosis and microglial activation are present as early as 12 days of age (Oswald et al., 2005), and indeed may start prenatally (Kay et al., unpublished observations). Data from this chapter provide the first evidence for glial responses in Cln3-/- mice during the early postnatal period. GFAP-positive staining of astrocytes was consistently more pronounced in superficial laminae and more dispersed in deeper laminae of the neocortex in Cln3-/- vs. littermate controls. However, the cues regulating this laminar specific distribution of astrocytes within the neocortex remain unclear.

154

Radial glia were also present in the neocortex and were more abundant in mutant than in control mice. Since GFAP-positive radial glia are differentiating either into astrocytes or neurons during this period of development (Voigt, 1989; Gotz et al., 2002; Tramontin et al., 2003), the relative abundance of these glial precursors in mutant mice may suggest a delay in their differentiation. Indeed, as regional volume was reduced in the same structure, but unaffected in older mice, this provides further evidence for a developmental delay within the neocortex, although it is unclear what is causing these pathological abnormalities to occur.

Similar to patterns of immunoreactivity in the neocortex, cerebellar expression of GFAP was more prominent in mutant mice, although the distribution of astrocytes tended to be generalised within individual layers. In marked contrast, GFAP-positive staining was reduced in the hippocampus of mutant mice, but unchanged in subcortical structures. These data are consistent with the localized astrocytic responses we have described within the aging brain (Chapter 3), and further suggest that cues mediating astrocytosis have subsided in the hippocampus.

F4/80-positive staining followed similar expression patterns to GFAP in parts of the neocortex and in the hippocampus of mutant mice. As astrocytic and microglial responses mirrored each other within these structures, it may be that similar molecular cues are in operation to govern the expression of GFAP and F4/80 in these parts of the brain. Cues may even be mediated by astrocytes or microglia themselves and thus be capable of influencing the expression of F4/80 and GFAP respectively. These data are also consistent with astrocytes protecting neurons from excess levels of glutamate that are present within the JNCL CNS (Chattopadhyay et al., 2002a). Indeed, as no neuronal loss was reported at 8 days of age, but glial responses were still evident, this may suggest that astrocytosis is sufficient to offer neuroprotection in the mutant CNS.

Similar to hippocampal staining, the expression of F4/80 was reduced subcortically, but was highly variable in the cerebellum of mutants. Since glial markers within these structures followed markedly different staining patterns from one another, it is likely that astrocytes and microglia are controlled by distinct cues outside of the cortex. It

155 may even be a combination of several different cues that govern glial responses to a variable extent depending on the environment.

As with older mice (Chapter 3), the extent of microglial transformation varied, depending on the brain region and genotype of the mouse. In marked contrast, GFAP- positive astrocytes in Cln3-/- mice failed to display the significant morphological changes that are typically associated with reactive astrocytosis (Graeber and Kreutzberg, 1986). The absence of transformation in regions of neuronal loss, microglia activation, or upregulation of GFAP may further suggest that astrocytes themselves are being targeted in the disease. Indeed, in regions where microglial activation was most apparent, an upregulation of GFAP-positive astrocytes may therefore represent an attempt to compensate and maintain a homeostatic balance.

In contrast to patterns of astrocytosis, morphological evidence for activation of microglia correlated with an increased expression of F4/80 in the neocortex of mutant mice. Consistent with these findings, transformation was infrequent in subcortical structures, where F4/80 immunoreactivity was weaker. Therefore, similar cues are likely to govern both expression and morphology of F4/80-positive microglia in the mutant CNS.

Changes in the CNS of P8 Cln3-/- mice may either be indicative of an early neuropathological phenotype, or developmental abnormalities. In this study, detailed analysis of glial and neuronal phenotypes show that glial changes occur prior to neuronal loss. Indeed, neuronal loss is only evident in older mutants (Chapter 3), but glial abnormalities are present as early as 8 days of age. Nevertheless, early changes in glial populations do not accurately predict subsequent neurodegeneration in Cln3-/- mice. As such, the role of glia in JNCL pathogenesis is likely to be complex and will require further investigation.

156

CHAPTER 6

Neuropathological Changes in Early Postnatal Homozygous Cln3∆ex7/8 Knock-in Mice

157 6.1 Introduction

Chapter 4 of this thesis describes neuropathological changes in 12 month old homozygous Cln3∆ex7/8 mice, an age which represents an advanced stage of disease pathogenesis. Having looked at one end of the disease spectrum, a wealth of information can be obtained by examining the phenotype of younger mice to determine when the neuropathological phenotype of murine JNCL is first apparent. At present, it is unknown whether there is a period of normal development prior to neuropathological changes that are subsequently evident in JNCL. As such, early evaluation of disease phenotype will help us to predict whether abnormalities are evident during development, and if so, which are the first neuropathological events to occur.

In order determine whether early changes occur in the CNS of homozygous Cln3∆ex7/8 mice, mutants were assessed for neurodegenerative and reactive phenotypes at one week of age (P7). This analysis was guided by data from aged Cln3∆ex7/8 mice (Chapter 4; Pontikis et al., 2005), as well as other mouse models of NCL (Chapter 3; Mitchison et al., 1999; Pontikis et al., 2004; Bible et al., 2004). A combination of unbiased, design-based stereological methodology and quantitative thresholding immunohistochemistry were used to investigate the phenotype of these mutant mice.

6.2 Assessment of regional atrophy in P7 homozygous Cln3∆ex7/8 mice

Macroscopic examination of P7 homozygous Cln3∆ex7/8 brains did not reveal an overt phenotype, when compared to age matched control mice (data not shown). Consistent with these observations, the brains of homozygous Cln3∆ex7/8 mice were only negligibly lighter than age matched controls and this reduction did not reach statistical significance (control 286±29mg, n=7; Cln3∆ex7/8 280±27mg, n=5, p=0.737). To investigate more subtle effects upon regional atrophy, the Cavalieri method was used to obtain unbiased estimates of the volume of the cortical mantle, hippocampus, striatum, thalamus and cerebellum in Nissl stained sections (Figure 6.1). Despite

158

Figure 6.1 Unbiased Cavalieri estimates of regional volume in homozygous Cln3∆ex7/8 mice. Unbiased Cavalieri estimates of regional volume revealed no significant regional atrophy in Cln3∆ex7/8 homozygotes vs. littermate controls (+/+) at 7 days of age. Regions examined include cortical mantle (Neocortex); hippocampus (Hipp); thalamus (Thal); hypothalamus (Hypoth); striatum (CPu); whole cerebellum (Cb total); cerebellar vermis (Cb vermis) and lateral hemisphere (Cb lat).

159 small reductions in the volume of the neocortex and cerebellum, there was no significant atrophy in any analysed CNS region in P7 homozygous Cln3∆ex7/8 mice vs. age matched controls.

6.3 Regional effects upon neocortical thinning and lamination in P7 homozygous Cln3∆ex7/8 mice

Since regional measurements of volume cannot discriminate between changes within individual neocortical subfields, a series of neocortical thickness measurements were made in the primary motor (M1), somatosensory barrel field (S1BF), primary visual (V1) and lateral entorhinal (Lent) cortex (Figure 6.2A, Table 6.1). Homozygous Cln3∆ex7/8 mice exhibited contrasting effects upon neocortical thickness in different subfields, with significantly greater neocortical thickness rostrally in M1 and S1BF, no change in thickness in V1 and significantly reduced neocortical thickness caudally in Lent.

To determine whether these changes were due to lamina-specific events, individual laminae were measured in each subfield (Figures 6.2B-E, Table 6.1). Deeper laminae (V and VI) followed the changes in overall neocortical thickness S1BF (Figure 6.2B) and M1 (Figure 6.2E). In contrast, other laminae within S1BF were unchanged, whereas M1 exhibited significant thickening of laminae II and III, and thinning of lamina I. A complex array of thickness changes was evident in V1, with laminae II, III and IV significantly increased in thickness, in contrast to lamina V, which displayed significant atrophy. Taken together these changes in individual laminar thickness in V1 resulted in no overall changed in the thickness of this subfield (Figure 6.2C). Within Lent, lamina II exhibited significant thinning, although a significant thickening was noted in laminae I and IV (Figure 6.2D), with remaining laminae negligibly affected.

160

Figure 6.2 Subregion and laminar-specific changes in neocortical thickness in homozygous Cln3∆ex7/8 mice. (A) Neocortical thickness measurements. Compared with littermate controls (+/+), the primary motor (M1) and primary somatosensory barrel field cortex (S1BF) exhibited a significant increase in overall thickness, the lateral entorhinal cortex (Lent) was significantly thinner and the primary visual cortex (V1) was unchanged in Cln3∆ex7/8 homozygotes. (B-E) Laminar thickness measurements in these neocortical regions revealed a complex series of changes in thickness of individual laminae in P7 homozygous Cln3∆ex7/8 mice compared with littermate controls. * p<0.05, ** p<0.01, *** p<0.001, one way ANOVA. These data are also summarized in Table 6.1.

161

I II/III III IV V VI Total Thickness

M1 ↓ ↑ na na ↑ ↑ ↑

S1BF NS NS na NS ↑ ↑ ↑

V1 NS ↑ na ↑ ↓ NS NS

Lent ↑ ↓ NS ↑ NS NS ↓

Table 6.1 Tabular depiction of significant changes in individual laminar thickness in homozygous Cln3∆ex7/8 mice. Measurements taken in the primary motor (M1), primary somatosensory barrel field (S1BF), primary visual (V1) and lateral entorhinal cortex (Lent) of P7 homozygous Cln3∆ex7/8 mice vs. littermate controls (+/+).↑ = significantly thicker in Cln3∆ex7/8; ↓ = significantly thinner in Cln3∆ex7/8; na = not applicable, M1 has no lamina IV. A combined measurement of laminae II and III was made in all regions except Lent where these laminae were measured separately.

162 6.4 Regional and laminar effects upon neuronal number in P7 homozygous Cln3∆ex7/8 mice

In order to determine whether these changes in individual laminar thickness were due to effects upon neuronal number and/or neuronal size, we obtained optical fractionator estimates of neuronal number and nucleator estimates of neuronal volume in Nissl stained sections (Figures 6.3A, B). These analyses focussed upon selective laminae of Lent and M1 (Lamina II and V respectively), as these are two neocortical regions which subserve contrasting functions, and displayed opposing patterns of atrophy in P7 homozygous Cln3∆ex7/8 mice (Figure 6.2). Selective components of the thalamocortical system were also investigated (Laminae IV and V of S1BF; ventral posterior thalamic nucleus [VPM/VPL]), as these structures exhibited regional atrophy and selective neuron loss in older homozygous Cln3∆ex7/8 mice (Chapter 4; Pontikis et al., 2005).

There was a significant loss of lamina II granule neurons within Lent of homozygous Cln3∆ex7/8 mice, which are responsible for cortico-cortical commissural and associative connections. However, no other neuronal populations examined were significantly affected in P7 homozygous Cln3∆ex7/8 mice when compared to age matched controls (Figure 6.3A). Furthermore, nucleator estimates of neuronal volume revealed negligible changes in volume that were not significant in any neocortical lamina or thalamic nucleus (Figure 6.3B).

6.5 Early astrocytic and microglial responses in P7 homozygous Cln3∆ex7/8 mice

We have previously described a profound glial response in aged homozygous Cln3∆ex7/8 mice (Chapter 4; Pontikis et al., 2005). It is unclear whether such events also occur early during JNCL pathogenesis. To examine evidence for early glial responses in homozygous Cln3∆ex7/8 mice we first surveyed the distribution of cells immunoreactive for the astrocytic markers GFAP and S100β, in addition to the microglial marker F4/80. Subsequently, the expression of these glial markers were

163

Figure 6.3 Unbiased optical fractionator estimates of neuronal number in homozygous Cln3∆ex7/8 mice. (A) There was a significant loss of granule neurons in lamina II of the lateral entorhinal cortex (Lent), but other regions of the CNS exhibited negligible changes of neurons in P7 homozygous Cln3∆ex7/8 mice compared with littermate controls (+/+). (B) Nucleator estimates of neuronal volume revealed no significant atrophy or hypertrophy in any laminae of Cln3∆ex7/8 homozygotes. Regions examined include lamina II of Lent, lamina V of the primary motor cortex (M1), laminae IV and V of the primary somatosensory barrel field cortex (S1BF), and the ventral posterior thalamic nucleus (VPM/VPL). ** p<0.01, one way ANOVA.

164 quantified using thresholding image analysis in neocortical subregions M1, S1BF and V1; subcortical structures, thalamus and striatum; and hippocampal subfields dentate gyrus (DG) and stratum oriens with CA1 (OrCA1). As detailed previously (Petito et al., 1990; Wu and Schwartz, 1998; Pekny and Nilsson, 2005), reactive astrocytosis may represent hypertrophy of astrocytes (thickening of processes and changes in cell body volume), which may be independent of cell proliferation. Therefore, we also obtained optical fractionator and nucleator estimates the number and size of astrocytes immunoreactive for S100β, a calcium binding protein that is expressed predominantly in astrocytes (Boyes et al., 1986).

Qualitative assessment of GFAP and S100β-positive astrocytes: In the neocortex of P7 control mice, GFAP-positive astrocytes were present predominantly in lamina VI and lamina I adjacent to the pial surface, although occasional scattered immunoreactive astrocytes were observed in intermediate laminae (Figure 6.4A). Age matched homozygous Cln3∆ex7/8 mice exhibited similar, if not a more pronounced astrocytosis in superficial laminae (Figure 6.4B). In deeper laminae, dispersed clusters of GFAP-positive astrocytes were evident in lamina V, most notably in M1 and V1 of these mutants. In contrast, GFAP-positive astrocytes were seldom seen in laminae II and III in any neocortical region of Cln3∆ex7/8 homozygotes. Morphologically, GFAP- positive astrocytes were typically protoplasmic in the neocortical grey matter in animals of either genotype. However, there was a greater incidence of fibrous, more intensively GFAP-positive astrocytes in homozygous Cln3∆ex7/8 mice vs. control littermates. In most neocortical subfields, GFAP-positive staining revealed small soma and moderate branching of intensely stained processes in Cln3∆ex7/8 homozygotes (Figure 6.4B), whereas age matched controls displayed subtler branching of processes and paler staining (Figure 6.4A). In marked contrast to the localized patterns of GFAP immunoreactivity, S100β immunoreactivity within the neocortex was distributed evenly across laminae within numerous intensely stained cell soma and did not differ obviously between animals of either genotype (Figures 6.4C, D).

Radial glia were also present in the neocortex, with characteristically long radial GFAP-positive fibres spanning the width of the neocortex from the corpus callosum

165

Figure 6.4 Astrocytic responses in homozygous Cln3∆ex7/8 mice. (A, B) Immunohistochemical staining for GFAP revealed a more pronounced astrocytosis with more abundant radial glia in the neocortex of P7 Cln3∆ex7/8 homozygotes (B) compared with age matched controls (+/+, A). Within the primary motor cortex (M1), additional staining of GFAP-positive astrocytes was present in laminae I and V of mutant mice, (B). In contrast, there were no obvious differences in the distribution or number of S100β-positive astrocytes in M1 of animals of either genotype (C, D). In the thalamus (Thal), astrocytosis was typically present within the ventral posterior (VPM/VPL), laterodorsal (LD) and ventromedial (VM) thalamic nuclei, and was more pronounced in homozygous Cln3∆ex7/8 mice (F) vs. control littermates (E). Additional sparse staining was exhibited throughout the thalamus in mutant mice that was absent from controls. Thalamic nuclei include VPM/VPL, LD, VM, posterior (Po), centrolateral (CL) and mediodorsal (MD). I-VI = individual lamina; ◄ = division of laminae

166 to the pial surface. Similar to early postnatal Cln3-/- mice (Chapter 5), soma of these radial glial cells could not be clearly identified due to the abundance of GFAP- positive astrocytic processes within the neocortex. Nevertheless, GFAP-positive radial glial fibres were more abundant and more intensely stained in P7 homozygous Cln3∆ex7/8 mice (Figure 6.4B), than in age matched controls (Figure 6.4A). Such findings may be indicative of abnormal differentiation of radial glia within the neocortex of mutant mice.

Subcortically, GFAP immunoreactivity was less pronounced and although widespread, was restricted to a range of individual nuclei. In the striatum, astroctytosis was more prominent immediately below the corpus callosum, but did not form clusters of activated astrocytes in either homozygous Cln3∆ex7/8 mice or littermate controls (data not shown). Although the distribution of GFAP immunoreactivity was similar for animals of either genotype, control mice displayed more pronounced staining in the striatum. Astrocytes within the thalamus was present predominantly in clusters of GFAP-positive cells in animals of either genotype (Figures 6.4E, F), with additional sparse staining in the thalamus of mutant mice (Figure 6.4F). Astrocytosis was typically observed in the ventral posterior, laterodorsal and ventromedial thalamic nuclei and was generally more pronounced in homozygous Cln3∆ex7/8 mice (Figure 6.4F) than in control littermates (Figure 6.4E). The distribution of GFAP-positive astrocytes within the hippocampal formation was widespread, although immunoreactivity was again more prominent in mutant mice (data not shown). Greater magnification revealed that hippocampal astrocytes were typically fibrous, with thicker processes and slightly larger soma in homozygous Cln3∆ex7/8 mice vs. littermate controls. In contrast, GFAP-positive astrocytes in the remaining subcortical regions were morphologically similar in animals of either genotype (data not shown).

Quantitative analysis of astrocytosis: GFAP immunoreactivity was pronounced in almost all cortical and subcortical structures of both homozygous Cln3∆ex7/8 and control mice (Figure 6.5A). In the neocortex of mutant mice, M1 and V1 exhibited significantly increased staining of GFAP when compared to control littermates, but only negligibly changed GFAP immunoreactivity in S1BF. Significantly increased staining for GFAP was also present in the stratum oriens and CA1 of the hippocampus

167

Figure 6.5 Quantitative assessment of regional astrocytosis in homozygous Cln3∆ex7/8 mice at 7 days of age. (A) Thresholding image analysis typically revealed a significantly increased expression of GFAP in cortical and subcortical regions of Cln3∆ex7/8 homozygotes compared with littermate controls (+/+). (B-C) The notable exceptions were the primary somatosensory barrel field cortex (S1BF) and hippocampal dentate gyrus (DG), which were unchanged; and the striatum, which exhibited significantly reduced levels of GFAP in homozygous Cln3∆ex7/8 mice vs. control littermates. (C) In contrast to frontal expression, the lateral hemispheres of the cerebellum displayed significantly reduced levels of GFAP in mutant mice compared with controls. Areas surveyed include primary motor (M1), S1BF and visual cortex (V1); thalamus (Thal); hippocampal DG, hippocampal stratum oriens and CA1 (OrCA1); cerebellar white matter (WM) and molecular and granule cell layers (Mol+Gr) in the lateral hemispheres (Lat) and vermis (Verm). (D) Optical fractionator estimates revealed no significant change in the number of S100β-positive astrocytes in either M1 or V1 of homozygous Cln3∆ex7/8 mice, compared with littermate controls. ** p<0.01, *** p<0.001, one way ANOVA.

168 of homozygous Cln3∆ex7/8 mice. In marked contrast, GFAP immunoreactivity was significantly reduced in white and grey matter of the cerebellar lateral hemispheres of mutant mice vs. control littermates, but was unchanged in the vermis (Figure 6.5B). Subcortical structures displayed markedly different degrees of GFAP immunoreactivity, with a significant increase in GFAP staining in the striatum of mutant mice, but a reduction in GFAP staining in the thalamus of mutant vs. control mice (Figure 6.5A).

To determine whether these changes in GFAP immunoreactivity reflected changes in the number of astrocytes, we next obtained unbiased stereological estimates of the number of S100β-positive astrocytes (Figure 6.5C). Optical fractionator data were collected within M1 and V1, as these cortical subfields displayed significant changes in GFAP immunoreactivity (Figure 6.5A). However, similar to older Cln3∆ex7/8 mice (Chapter 4), this analysis revealed no significant change in the number of astrocytes within either region of homozygous Cln3∆ex7/8 mice compared with control littermates (Figure 6.5C).

Qualitative assessment of F4/80-positive microglia: F4/80-positive microglia were widely distributed in animals of either genotype, although expression of this antigen differed markedly between neocortical subfields (Figures 6.6A-F). The neocortex of Cln3∆ex7/8 homozygotes displayed a more pronounced F4/80 immunoreactivity in most subfields (Figures 6.6B, D) when compared to control mice (Figures 6A, C). Indeed, microglia from mutants were generally more intensely stained, had thickened and less ramified processes and more prominent cell soma, but full macrophage-like morphology was not evident in these mice (Figure 6.6D). In contrast, control littermates exhibited paler F4/80 immunoreactivity within the cell soma, which were generally smaller with thin branched processes extending into the neuropil (Figure 6.6C).

More intense F4/80 staining of microglia was also present within the hippocampus of mutant mice vs. control littermates, with morphological changes within the hippocampus similar to those described in the neocortex (data not shown). Subcortically, the relative level of F4/80-positive staining was regionally variable in

169

Figure 6.6 Microglial responses in homozygous Cln3∆ex7/8 mice. Immunohistochemical staining for F4/80 revealed a graded activation of microglia in P7 homozygous Cln3∆ex7/8 mice compared with littermate controls (+/+). In the primary motor cortex (M1), F4/80-positive microglia were more darkly stained, had thickened ramified processes and more prominent cell soma in Cln3∆ex7/8 homozygotes (B, D) when compared to littermate controls (A, C). This difference between genotypes was more pronounced in the thalamus (Thal), with greater evidence for partial activation in Cln3∆ex7/8 mice (F) vs. age matched controls (E). Thalamic nuclei include ventral posterior (VPM/VPL), laterodorsal (LD), posterior (Po), centrolateral (CL), mediodorsal (MD) and ventromedial (VM).

170 Cln3∆ex7/8 homozygotes vs. littermate controls. In the striatum, control brains displayed intensive staining of microglia that occasionally exhibited partial transformation into a macrophage-like morphology. Mutant mice had paler F4/80 staining, but there was a greater incidence of F4/80 positive microglia displaying morphological evidence of microglial activation (data not shown). The expression of F4/80 immunoreactivity was widespread within the thalamus, although F4/80-positive microglia were less abundant in VPM/VPL of Cln3∆ex7/8 homozygotes compared to other thalamic nuclei (Figure 6.6F). Nevertheless, collective F4/80 staining in the thalamus of mutant mice was more intense than in other brain regions and compared to age matched controls. Furthermore, Cln3∆ex7/8 homozygotes displayed a similar, but more frequent incidence of microglial transformation to that observed in the neocortex. In contrast, control mice exhibited weak F4/80 staining of microglia, which exhibited smaller soma, thinner processes and little morphological evidence for microglial activation (Figure 6.6E).

Quantitative analysis of F4/80-positive microglia: Quantitative analysis of F4/80 immunoreactivity revealed a similar distribution of staining compared to GFAP immunoreactivity (Figures 6.7A, B). The neocortex, M1 and V1 of homozygous Cln3∆ex7/8 exhibited significantly increased expression of F4/80 vs. control littermates, with minimal changes in F4/80 immunoreactivity in S1BF (Figure 6.7A). Similarly, F4/80 staining in the hippocampus of mutant mice was more pronounced in the stratum oriens and CA1, but was unchanged in the dentate gyrus when compared to age matched controls. Subcortical expression of F4/80 was highly variable between different CNS regions, with a significantly increased staining of F4/80 in the thalamus and a reduction in F4/80 immunoreactivity in the striatum of homozygous Cln3∆ex7/8 mice vs. littermate controls. In the cerebellum, staining of F4/80 contrasted patterns of GFAP immunoreactivity (Figure 6.7B), with significantly reduced F4/80 immunoreactivity the in white and grey matter of the cerebellar vermis of mutant mice, but no difference in F4/80 staining in the lateral hemispheres of mutant vs. control littermates.

171

Figure 6.7 Immunohistochemical analysis of microglial responses in P7 homozygous Cln3∆ex7/8 mice. F4/80 immunoreactivity was significantly elevated in almost all analysed brain regions. The notable exceptions were the primary somatosensory barrel field cortex (S1BF), hippocampal dentate gyrus (DG) and lateral hemispheres of the cerebellum, which were unchanged. In contrast, the striatum displayed significantly reduced levels of F4/80 in homozygous Cln3∆ex7/8 mice vs. control littermates. Areas surveyed include primary motor (M1), S1BF and primary visual cortex (V1); thalamus (Thal); hippocampal DG, hippocampal stratum oriens and CA1 (OrCA1); cerebellar white matter (WM) and molecular and granule cell layers (Mol+Gr) in the lateral hemispheres (Lat) and vermis (Verm).* p<0.05, ** p<0.01, *** p<0.001, one way ANOVA.

172 6.6 Discussion

Early postnatal evaluation of neuropathological phenotypes has provided valuable information about the timing of neurodegenerative and glial changes in Cln3∆ex7/8 homozygotes. Indeed, changes in neocortical thickness, selective patterns of neuronal loss and pronounced glial responses throughout the CNS may be indicative of developmental abnormalities or the early appearance of a neuropathological phenotype. Furthermore, the early postnatal gliosis in neocortical S1BF and thalamic VPM/VPL clearly precedes the neuronal loss that is evident in aged homozygous Cln3∆ex7/8 mice (Chapter 4).

Gross and regional examination of brains did not reveal any atrophy, although similar to older mice (Chapter 4), changes in neocortical thickness were observed in almost all neocortical subfields with only V1 unaffected. Furthermore, changes in neocortical thickness predicted the regional pattern of subsequent neocortical atrophy observed in aged homozygous Cln3∆ex7/8 mice (Chapter 4). Measurements of laminar thickness in early postnatal mutants revealed a complex series of changes with individual laminae contributing to variable extents to the overall changes in thickness of each neocortical subfield. More precisely, deeper laminae (V and V1) were primarily responsible for the increased thickness of M1 and S1BF in Cln3∆ex7/8 mice, but were unchanged in Lent, which was atrophied in these mutant mice. Remaining laminae underwent highly variable changes and contributed to overall neocortical thickness by different extents.

We have already shown that in older mice there was a selective loss of neuronal populations from different parts of the brain (Chapter 4). However, such changes in neuron number did not correspond directly with variations in neocortical thickness. To determine whether this was also true for P7 mutants, a stereological survey of neuronal number was conducted in selected laminae. Cln3∆ex7/8 homozygotes displayed a loss of lamina II granule neurons in Lent, which is consistent with an overall thinning of this neocortical subfield when compared to control brains. Conversely, no other subfields displayed neuronal loss or even changes in neuronal volume, but thickness changes were still observed within these regions of the

173 neocortex. Thus, changes in neocortical thickness are multifactorial and may be in part due to changes in neuronal number, as well as other compounding factors, such as altered packing between cells or effects on dendritic arborisation. As such, it would be highly informative to investigate these parameters using a combination of stereology, dye impregnation or intracellular filling techniques and thresholding image analysis.

Since neuronal loss is mostly absent from younger mutants, neuronal populations appear to be correctly targeted to their corresponding compartments during development, in a similar manner to control mice. Subsequent neurodegenerative changes occur to different extents in selective neuronal populations in a temporally regulated manner. However, neuropathological cues which control the spatiotemporal progression of this disease phenotype at this time remain unknown. Perhaps selective neuronal populations are more vulnerable to changes in their immediate environment, such as an excitatory/inhibitory imbalance (Dodd, 2002), or changes in glial populations resulting from the local release of cytokines from injured neurons (Klein et al., 1997).

We have shown that aged homozygous Cln3∆ex7/8 mice exhibit pronounced astrocytosis, but that microglial responses are far less pronounced (Chapter 4). Consistent with these findings, data presented in this study suggest there to be a more prominent GFAP-positive astrocytic response in P7 Cln3∆ex7/8 homozygotes vs. littermate controls. Nevertheless, there was little evidence of astrocytic proliferation or hyperplasia, with no difference in the number of S100β-positive astrocytes between mutants and control mice. Thus, reduced GFAP staining is likely to represent a downregulation of this astrocytic marker, rather than a loss of astrocytes.

Radial glia were present in animals of both genotypes, but were more abundant and intensively GFAP-immunoreactive within the neocortex of mutant mice. As detailed previously, during this early period of postnatal development radial glia are differentiating either into astrocytes or neurons (Chapter 5), therefore a greater abundance of these precursor cells suggests that there may be a delay in the differentiation of radial glia within the neocortex, although this hypothesis currently remains unproven. Possible causes of such an abnormal phenotype may arise from

174 deficiencies in molecular cues that govern the maturation of radial glia. Previous findings suggest that expression of a radial glial cell identity in mammalian forebrain is determined by the availability of diffusible inducing signals from embryonic cortical cells (Hunter and Hatten, 1994). Furthermore, these signals may also transform mature astrocytes back to a radial glial phenotype (Hunter and Hatten, 1994), thus multiple possibilities may account for the more pronounced expression of GFAP-positive radial glia in P7 Cln3∆ex7/8 homozygotes.

The upregulation of GFAP staining was exclusively present in the cerebrum, with this marker reduced in the cerebellum, which incidentally is the last structure to develop within the CNS (Wang and Zoghbi, 2001). Thus, cues responsible for mediating astrocytosis or astrocyte differentiation may be different and potentially compromised in the developing or recently developed brain. Alternatively, astrocytes may themselves be targets of the disease, as has been suggested in P8 Cln3-/- mice (Chapter 5).

Although widespread, astrocytosis was not evident in all regions within the cerebrum of mutant mice. Neocortically, M1 and V1 displayed an increased expression of GFAP-immunopositive astrocytes, whereas S1BF was unaffected. Contrasting changes in GFAP expression were also observed in hippocampal and subcortical structures. Again, it is likely that cues mediating astrocytosis are affected to different extents in selective regions of the CNS.

Curiously, staining of GFAP-positive astrocytes in the neocortex was consistently more pronounced in superficial (I) and deeper neocortical laminae (VI) with dispersed clusters of astrocytes noted in lamina V of mutant mice, that were absent from control littermates. Such patterns of reactive astrogliosis may be subtle, but are consistent with observations in aged Cln3∆ex7/8 homozygotes (Chapter 4) and may be indicative of a glial response to early postnatal damage to neocortical neurons. Indeed, the precise mechanisms of neuronal injury resulting from the presence of an abnormal Cln3 protein remain unknown, but are likely to initiate during early postnatal development and progress during the course of disease. Since regional expression of GFAP mirrored patterns of F4/80 immunoreactivity, it is possible that similar cues are controlling astrocytes and microglia. As such, the ability of astrocytes to clear

175 extracellular glutamate (Rothstein et al., 1996; Duan et al., 1999) may offer neuroprotection from activated microglia in the developing brain. Alternatively microglial activation may occur in response to unknown pathological insults, with subsequent triggering of astrocytosis from cues secreted by microglia.

The extent of glial transformation in mutant mice was not only distinct from that observed in control littermates, but also varied between CNS regions. In the neocortex, there was morphological evidence for activation of astrocytes and microglia, which was consistent with the increased expression of GFAP and F4/80 in the neocortex. Indeed, microglia become activated in response to pathological changes in the CNS (Kreutzberg, 1996), then migrate towards damaged cells (Heppner et al., 1998) before secreting a number of neurotrophic factors, or neurotoxic factors, depending on the extent of neuronal injury (Kreutzberg, 1996; Heppner et al., 1998; Koutsilieri et al., 1999; Liu and Hong, 2003). Astrocytes classically mediate responses to neuronal injury (Raivich et al., 1999), such as glutamate induced excitotoxicity from activated microglia. Thus, selective patterns of astrocytosis could potentially be explained by the interplay between these two glial populations. Conversely, the lack of morphological transformation of glial cells in other regions of the CNS may further suggest that glia themselves are targeted as part of the disease process, or simply that the cues needed to activate these cell types are not present.

Although reactive gliosis is prominent within selective structures of the CNS, neuronal loss was largely absent in P7 homozygous Cln3∆ex7/8 mice. Instead, neurodegeneration is typically a later event during disease progression and has a complex relationship with glial changes that is largely unclear. Perhaps the neuropathological effects of activated microglia are only partially buffered by astrocytosis in mutant mice; therefore, chronic exposure to activated microglia may cumulatively contribute to a late neurodegenerative phenotype. Alternatively, microglia may be attempting to assist in neuronal recovery, as has been suggested in previous literature (Kreutzberg, 1996). Indeed, microglia may only play a minor contributory role to overall neuronal loss, a process which may be initiated by other unknown mechanisms.

176

CHAPTER 7

General Discussion

177 7.1 Introduction

The studies described in this thesis investigate the sequence of pathological changes during disease progression, and have begun to define the nature and timing of pathological effects upon the CNS in two major mouse models of JNCL. With very little known about the pathogenesis of JNCL, the development of Cln3-/- and homozygous Cln3∆ex7/8 mice can provide invaluable insights into the progressive nature of pathological changes within the CNS of JNCL patients. Furthermore, such data may enable the strategic timing and placement of potential therapeutic interventions in these established mouse models of JNCL.

An important goal of this thesis was to investigate when pathological events first occur within the JNCL CNS. These studies were prompted by the discovery that glial activation is a relatively early event, occurring several months before Cln3-/- mice become symptomatic (Pontikis et al., 2004). These data also raised the question whether the CNS develops normally in this disorder, and this issue was addressed using unbiased stereological methodology to characterize the extent of neuronal loss and glial activation during early postnatal development in Cln3-/- and homozygous Cln3∆ex7/8 mice.

During the period of early postnatal development, although neuronal loss was negligible in either mouse model, there were marked changes in the expression of glial markers. It is not clear whether these changes reflect abnormal glial differentiation, or their activation at an early stage of pathogenesis. Nevertheless, these changes were localized and differed in extent between CNS regions, but rarely predicted subsequent cell loss in older mice. As the disease progressed, both mouse models exhibited variable thinning across subfields of the neocortex, which was due to a range of changes in individual laminar thickness and selective effects upon neuronal survival. Further analysis revealed a profound loss of neurons within the thalamocortical system of both mouse models. These changes were accompanied by reactive changes, which were more pronounced in homozygous Cln3∆ex7/8 mice and included localized clusters of microglia and widespread astrocytosis.

178 These data provide evidence for reactive gliosis and selective thalamic neuron loss early in the pathogenesis of JNCL. These glial responses persist throughout the progression of disease and are accompanied by neuronal loss that becomes more widespread with increased age. However, it remains unclear whether these glial responses, represent a regenerative or degenerative response to mutations in the Cln3 protein.

7.2 Mouse models of JNCL

Cln3-/- mice are missing an integral coding region of the Cln3 gene, which prevents the production of functional Cln3 protein. As such these mice serve as a model of JNCL and, similar to individuals with JNCL, these mice display an accumulation of autofluorescent storage material (Figure 3.1; Mitchison et al., 1999), with a selective loss of neuronal populations, including interneurons and pyramidal cells (Figures 3.5, 3.6; Mitchison et al., 1999; Pontikis et al., 2004; Tyynelä et al., 2004). Indeed, Cln3-/- mice display many of the features of human JNCL, including visual, motor and seizure phenotypes. Seizures are a common clinical phenotype of JNCL and may be associated to some extent with this loss of GABAergic interneurons (Goebel and Wisniewski, 2004). However, this loss of interneurons is not pronounced in mutant mice and only becomes significant at an advanced age (Figure 3.7; Pontikis et al., 2004). This loss of interneurons results in an altered threshold for seizure generation, but not the production of spontaneous seizure activity (Kriscenski-Perry et al., 2002). Retinal pathology is also common to both Cln3-/- mice and human JNCL, although this is subtle in mutant mice (Seigel et al., 2002; Goebel and Wisniewski, 2004). In contrast, patients with JNCL show prominent retinal degeneration and visual failure (Goebel and Wisniewski, 2004). Recent data from Cln3-/- mice suggests that the altered visual function displayed by these mice is the result of selective loss of neurons in the dorsal lateral geniculate nucleus rather than effects upon the retina (Weimer et al., 2005a)

Similarities between human JNCL and the phenotype of Cln3-/- mice also extend to neuroimmune responses. Indeed, neuronal loss in the CNS of human JNCL and Cln3-/-

179 mice is accompanied by a pronounced glial response with regionally specific activation of microglia and astrocytes (Chapter 3; Pontikis et al., 2004; Tyynelä et al., 2004). Nevertheless, the glial responses displayed by Cln3-/- mice are not as pronounced as those evident in human JNCL autopsy material (Tyynelä et al., 2004), with little evidence of astrocytic hypertrophy or transformation of microglia to brain macrophage-like morphology (Chapter 3; Pontikis et al., 2004). One feature that is characteristic of both murine and human JNCL is the presence of an autoimmune response that involves multiple brain-directed autoantigens (Lim et al., 2005b), including autoantibodies to glutamic acid decarboxylase (GAD65) (Chattopadhyay et al., 2002a; Chattopadhyay et al., 2002b).

Bearing the major deletion that is present in over 85% of CLN3 alleles in JNCL (International Batten Disease Consortium, 1995), Cln3∆ex7/8 mice represent another model of this disorder, that is genetically accurate to the human disease (Cotman et al., 2002). Certainly, the homozygous Cln3∆ex7/8 mice used in this study exhibit an early onset of neurologic deficits (Cotman et al., 2002) that are more pronounced than the relatively mild retinal and neurological phenotypes of Cln3-/- mice (Katz et al., 1999; Mitchison et al., 1999; Seigel et al., 2002). This neuronal loss was accompanied by severe reactive gliosis, which was more conspicuous than that seen in Cln3-/- mice (Cotman et al., 2002; Pontikis et al., 2005). Conversely, seizures were not prominent in homozygous Cln3∆ex7/8 mice (Cotman et al., 2002) and GABAergic interneurons were largely spared (Figure 3.7; Pontikis et al., 2005). Thus, both animals recapitulate numerous clinico-pathologic features of the human disease, but have subtly different phenotypes. However, any comparisons between these mice are necessarily complicated by the different strain backgrounds these models have been raised on (Mitchison et al., 1999 and Cotman et al., 2002).

180 7.3 Importance of strain backgrounds

Analyses of single nucleotide polymorphisms have recently been used to characterize genetic variation between inbreed strains of mouse, and have revealed the extent of variability to be considerable (Wade et al., 2002). Indeed, differences in genetic background between strains have been shown to affect subsets of genes (Sandberg et al., 2000; Pavlidis and Noble, 2001), which may account for phenotypic differences between mouse models on different strain backgrounds. However, the subset of genes that are significantly affected are small (Sandberg et al., 2000; Pavlidis et al., 2001), and only ~1% of expressed genes differ between two separate mouse strains (Sandberg et al., 2000).

The NCL mouse models generated so far have been on a variety of mixed, 129Sv or C57Bl/6 strain backgrounds (Mitchison et al., 1999; Gupta et al., 2001; Bible et al., 2004; Kopra et al., 2004; Pontikis et al., 2004; Sleat et al., 2004; Jalanko et al., 2005; Pontikis et al., 2005). These strains are known to exhibit marked polymorphisms in many of their genetic markers (Simpson et al., 1997). Furthermore, behavioral patterns between these strains are quite distinct, exhibiting marked differences in tests of learning, memory, anxiety, pain responsivity, olfactory discrimination and sensitivity to psychotropic drugs (Homanics et al., 1999; Murphy et al., 2001; Crabbe et al., 2003; Lee et al., 2003). It is worth noting that the extent of phenotypic change due to a mutation may vary according to the individual strain background (Voikar et al., 2001; Lipp and Wolfer, 2003). However, it has been suggested that the number of genes affected by genetic background is considerably lower than the number of differentially expressed genes caused by a genetic defect (Turk et al., 2004).

Possible background influences of modifier genes on disease causing mutations have been well illustrated in the mnd mouse model of CLN8, which displayed a milder disease phenotype on a mixed C57B1/6J background before being transferred to an inbred AKR/J strain (Messer et al., 1987; Messer et al., 1995; Messer et al., 1999). Conversely, PPT1-/- mice exhibited similar neuropathological phenotypes such as severe atrophy, with no

181 reduction in cerebellar volume on both mixed 129SV/C57 and congenic C57B1/6J backgrounds (Maddox et al., 2005). Thus, it will be essential to re-evaluate phenotypes of both Cln3-/- and homozygous Cln3∆ex7/8 mice once these models of JNCL are all available upon a common C57B1/6J strain background. At present, efforts are underway to transfer all mouse models of NCL onto the same C57B1/6J congenic background (www.nclmodels.org). Such unified strategies will enable more direct comparisons of the phenotypes of different mouse models of NCL and will greatly aid future work within this field.

7.4 Accumulation of autofluorescent storage material in mouse models of JNCL

The accumulation of autofluorescent storage material is a consistent neuropathological feature of all forms of human NCL (Elleder et al., 1999; Goebel and Wisniewski, 2004b), as well as various animal models of the disease (Mitchison et al., 2004). Consistent with previous findings (Mitchison et al., 1999), autofluorescent storage material was evident within the CNS of severely affected Cln3-/- mice, but there was also an accumulation of age-related lipopigments in similarly aged control mice (Figure 3.1). It has previously been assumed that that an abundance of lipopigments within a cell would potentially impede normal function, cause cells to become hypertrophic and ultimately result in their destruction. Nevertheless, is unclear whether the extent of storage deposition is related to severity of atrophy or neuronal loss. Indeed, studies have shown a disconnect between the extent of storage deposition and neurodegenerative changes observed in the NCLs (Hofmann and Peltonen, 2001; Tyynelä et al., 2004). This disconnect has been demonstrated most clearly in OCL6 sheep, with the widespread and progressive accumulation of storage material being in stark contrast to the localized nature of neurodegenerative and reactive events (Oswald et al., 2005). Furthermore, this lack of association has now been recapitulated in Cln3-/- mice, which display a widespread accumulation of autofluorescent storage material, but a selective loss of neuronal populations within specific regions (Chapter 3). A widespread accumulation of storage

182 material was also exhibited throughout homozygous Cln3∆ex7/8 brains and again, this did not correlate with patterns of neuronal loss (Chapter 4). Although data may suggest that these two events are independent of one another (Hofmann and Peltonen, 2001; Tyynelä et al., 2004), the possibility remains that certain neuron populations are more vulnerable to the accumulation of storage material.

7.5 Regionally specific changes in mouse models of JNCL a) Regionally specific changes in the Cln3-/- mouse model of JNCL

Brain atrophy is a feature that is common to various mouse models of NCL, including PPT1-/- (Gupta et al., 2002; Bible et al., 2004), TPP-I-/- (Sleat et al., 2004), Cln6-deficient mice (nclf, Lam et al., 1999), Cln8-deficient mice (mnd, Cooper et al., 1999) and cathepsin D-deficient mice (Haapanen et al., 2005). These histological data are supported by recent MRI analysis of these mouse models (Haapanen et al., 2005), which has revealed neocortical and thalamic atrophy to be a consistent feature of these mutant brains. Moderate cerebral atrophy and a slight reduction of thalamic signal intensity have been observed in the brains of aged JNCL patients (Autti et al., 1996). Recent analysis of Cln3-/- mice showed regional volume changes in the adult brain to be subtle (Figure 3.2; Pontikis et al., 2004), but closer examination of the neocortex revealed thinning of specific neocortical subfields. More precisely, this neocortical thinning was apparent in the primary motor (M1) and somatosensory barrel field cortex (S1BF) at 18 months of age, but was absent from 5 and 14 month old mice (Figure 3.3). Since 18-20 months of age represents terminal stages of the disease in Cln3-/- mice (Mitchison et al., 2004), these atrophic changes are not only mild, but are a relatively late event in JNCL pathogenesis. Furthermore, since this atrophy is more pronounced in M1 and S1BF, but the primary visual (V1) and lateral entorhinal cortex (Lent) appear thicker (Figure 3.3), the effects of Cln3 mutation upon neocortical thickness appear to differ markedly along the rostrocaudal axis. Intruigingly, a similar phenotype was evident in Cln3-/- mice at one week of age (Figure 5.2), accurately predicting the later phenotype of these mice. The

183 underlying molecular basis for these effects remains unclear, but a variety of transcription factors are recognized to underlie the patterning of the forebrain and arealization of the neocortex (Rubenstein and Beachy, 1998; Bishop et al., 2002). It will be informative to survey the expression of these and other transcription factors within the developing Cln3- /- brain.

b) Regionally specific changes in the early postnatal Cln3-/- mouse model of JNCL

Despite the absence of notable atrophy in 5 and 14 month old Cln3-/- mice (Figure 3.3), the Cavalieri method revealed a reduction of volume in the cortical mantle and hippocampal formation of P8 mutant mice (Figure 5.1). Thickness changes were also prominent in neocortical subfields, again with greatest thinning rostrally in the brain (Figure 5.2). It will be important to determine the cellular and molecular basis for this reduced neocortical volume and distinguish whether these changes represent an ongoing degenerative process or a delay in neocortical maturation which appears to be most pronounced rostrally. It is also possible that corticogenesis may be dysfunctional in Cln3- /- mice, which may involve an abnormal generation of neurons, improper migration of neocortical neurons or defective neuronal connectivity. Nevertheless, stereological estimates of neocortical neuron number suggest that neuronal survival is not significantly affected in the early postnatal Cln3-/- brain (Figure 5.3). Indeed, our data argue for only subtle effects on neurons in the Cln3-deficient neocortex during the early postnatal period, instead revealing more pronounced effects on the expression of astrocytic and microglial markers at this age (Chapter 5). The molecular events that underlie this early phenotype of these mice remain unknown. However, one clue may come from transcription factor profiling of Cln3-/- mice, which revealed specifically upregulated expression of Notch2 (Weimer et al., 2005c). Notch signaling is required throughout development to regulate the spatial patterning, timing and outcomes of many different cell fate decisions (Bray, 1998; Artavanis-Tsakonas et al., 1999). Thus, changes in Notch signaling or other transcription factors may have deleterious effects on the developing

184 brain. The complex developmental role of transcription factors may potentially extend to selective effects upon dendritic arborisation. Indeed, as dendrites form ~95% of the total volume occupied by neurons (Kincaid et al., 1998), if the dendritic tree does not form properly or there is excessive pruning back of dendrites, this may result in an apparent atrophy of the neocortex. In which case, Golgi staining or intracellular filling techniques would be useful for the assessment of defective neuronal connectivity.

c) Regionally specific atrophy in the homozygous Cln3∆ex7/8 mouse model of JNCL

Consistent with the phenotype of Cln3-/- mice (Chapter 3), regional atrophy in the brains of 12 month old homozygous Cln3∆ex7/8 mice was very mild, only reaching significance in the thalamus (Figure 4.2; Pontikis et al., 2005). Indeed, since approximately 20% of these mice do not reach 12 months of age (Cotman et al., 2002), by performing our analysis at this age, we may have selected mice that present with a less pronounced or delayed neurodegenerative phenotype (Pontikis et al., 2005). MRI studies of JNCL patients have revealed a slight reduction of signal intensity in the thalamus during the early symptomatic stages of the disease (Autti et al., 1996). However, unlike homozygous Cln3∆ex7/8 mice, thalamic atrophy has not yet been described in human JNCL. Nevertheless, such changes may highlight early pathological events in the thalamus which are common to both murine and human JNCL, although the precise nature of these events remains unclear.

Since subtle effects on neocortical thickness may be overlooked when solely assessing regional volume, data from aged Cln3∆ex7/8 homozygotes emphasize usefulness of these measurements to reveal highly localized and sub-region specific effects upon the cortical mantle (Figure 4.3). As proposed for Cln3-/- mice, such changes may be attributed to problems with arealization of the neocortex, abnormal generation of neurons, improper migration of neocortical neurons, or defective neuronal connectivity, all of which warrant further investigation.

185 d) Regionally specific changes in the early postnatal homozygous Cln3∆ex7/8 mouse model of JNCL

Regional atrophy was absent from homozygous Cln3∆ex7/8 mice at 7 days of age (Figure 6.1), which initially suggests a normal period of development before atrophic changes subsequently become apparent later in disease progression. Nevertheless, closer examination of neocortical subfields in one week old Cln3∆ex7/8 mice revealed changes in neocortical thickness that predict the regional pattern of subsequent neocortical atrophy observed in aged homozygous Cln3∆ex7/8 mice (Figures 6.2, 4.3). It is unclear whether the early thinning of Lent in homozygous Cln3∆ex7/8 mice is due to a delay in its maturation, but this was the only subfield to display a significant loss of neurons in these mice at 1 week of age (Figure 6.3). In order to determine the true nature of these differences between mutants and age matched controls, it will be necessary examine homozygous Cln3∆ex7/8 mice prior to P7.

7.6 Lamina specific changes in mouse models of JNCL

Closer examination of neocortical subfields revealed selective effects on neocortical thinning that extended to changes in individual laminae (Chapters 3-6). Of these events, lamina V typically provided the greatest contribution to changes in neocortical subfield thickness in both Cln3-/- and homozygous Cln3∆ex7/8 mice. However, deeper lamina VI was also a major determinant of changes in neocortical subfield thickness in early postnatal mutants. Since these phenotypic abnormalities were consistently prominent in lamina V, these data suggest that lamina V pyramidal neurons are selectively vulnerable to the downstream effects of a dysfunctional Cln3 protein, as was confirmed by stereological estimates of neuronal number in aged Cln3-/- mice. However, the heterogeneous nature of neurons in lamina VI necessarily complicates the assignment of vulnerability within this layer. Although these data argue for lamina specific changes, there was also a more diffuse and widespread effect upon the neocortex in Cln3-/- and homozygous Cln3∆ex7/8 mice, but as discussed previously, the cellular and molecular basis

186 for these effects is presently unknown. Nevertheless, the stereological assessment of neuronal number and volume has been highly informative in revealing these lamina specific changes. Indeed, these data suggest that changes in laminar thickness are not always consistent with patterns of neuronal loss. Moreover, neuronal volume was consistently comparable between animals of either genotype. Thus, changes in laminar thickness are likely to be complex, and may involve multiple pathological mechanisms which may potentially contribute to these phenotypic abnormalities to different extents.

7.7 Selective loss of neocortical neurons in mouse models of JNCL a) Selective loss of neocortical neurons in Cln3-/- mice

Our data from aged Cln3-/- and homozygous Cln3∆ex7/8 mice revealed a range of selective effects upon neocortical thickness that varied between different subfields (Figures 3.3, 4.3). In 18 month old Cln3-/- mice, these effects extended to individual laminae (Figure 3.4) and, as highlighted by data from the motor and somatosensory cortex (Figure 3.5), to neuronal number within these individual laminae. Neuronal loss was most prominent within lamina V of M1, which is the main motor output to motor neurons in the spinal cord. Previous studies in human JNCL and Cln3-/- mice have localized motor deficits to the cerebellum and midbrain (Raininko et al., 1990; Ruottinen et al., 1997; Weimer et al., 2005c; Weimer et al., 2005d), but as yet have failed to implicate the motor cortex. Indeed, in Cln3-/- mice there was a selective loss of Purkinje cells and their efferent target neurons within the deep cerebellar nuclei, and as such is a likely candidate for problems with motor timing and coordination (Weimer et al., 2005d). Extrapyramidal dysfunction can also be attributed to selective affects to striatal circuitary within the basal ganglia, including oxidative damage and progressive loss of neurons that has been demonstrated in these mice (Weimer et al., 2005b). However, as all motor commands eventually converge on motor neurons, before exiting the spinal cord (Ghez and Krakaeur, 2000), it is not unreasonable to postulate that the loss of lamina V pyramidal neurons in M1 may exacerbate motor dysfunction described in Cln3-/- mice. Alternatively, other deficits may

187 arise as a result of hypointensive signaling from lamina V of M1. Neuronal loss was also noted for lamina V pyramidal neurons of S1BF (Figure 3.5). Since these neurons are responsible for providing somatosensory feedback to the thalamus (Lopez-Bendito and Molnar, 2003), the transfer of information in the thalamocortical system and the adjustment of long term sensory sensitivity may be compromised. At present, the clinical consequences of these events are unclear, but will be readdressed at a later point in this thesis.

Patterns of neuronal loss were altogether absent from S1BF in younger mutants (Figure 5.3). Thus, taken together with findings from M1, the loss of lamina V pyramidal neurons is clearly a relatively late event in JNCL pathogenesis, but the precise nature of associated pathological cues that influence these events also remains unclear. Since lamina V pyramidal neurons were consistently targeted in the neocortex of mutant mice, these neurons may indeed be selectively vulnerable to downstream affects of a dysfunctional Cln3 protein. Similarly, large pyramidal neurons of lamina V are targeted in Alzheimer disease, and indeed are the most severely affected cells in the neocortex in this condition (Jessell and Sanes, 2000). Although the intralysosomal accumulation of autofluorescent storage material in JNCL is notably different from the extracellular build- up of amyloid plaques in Alzheimer disease, it is possible that large pyramidal neurons may simply be more sensitive to an accumulation of abnormally processed or undegraded protein than other neuronal populations. It is clear that the precise reasons for this selective vulnerability are lacking, but important clues may come from expression studies performed on laser capture dissected pyramidal neurons and other more resistant neuron populations.

b) Selective loss of neocortical neurons in the homozygous Cln3∆ex7/8 mouse model of JNCL

Lamina V neurons were largely spared in aged Cln3∆ex7/8 homozygotes, which may reflect the different strain background, the residual function of mutated Cln3 protein or

188 alternatively, this phenotype may take longer to develop in these mice. Nevertheless, other neocortical neuron populations were targeted in this mouse model, as illustrated in the somatosensory cortex (Figure 4.4). The most pronounced loss within this neocortical region was of laminae II and III neurons, which supply commissural and association projections to other areas of neocortex (Amaral, 2000). The loss of these neurons may potentially have significant consequences for coordinating neuronal activity between hemispheres and neocortical regions. In this respect, it will be important to determine the functional correlates of these pathological changes in the neocortex of homozygous Cln3∆ex7/8 mice. Lamina IV granule neurons were also lost in aged mutants, and since this is where most of the thalamic input terminates (Lopez-Bendito and Molnar, 2003), transfer of somatosensory information from the thalamus is likely to be compromised. Consistent with these findings, progressive changes in somatosensory evoked potentials have been reported in JNCL patients (Lauronen et al., 1997; Lauronen et al., 1999), but again the precise clinical consequences of these events remain unclear.

Patterns of neuronal loss observed in the neocortex of aged Cln3∆ex7/8 homozygotes were altogether absent from P7 mutant mice (Figure 6.3). This age related loss of neurons suggests that such neurodegenerative changes are a later event in disease pathogenesis. As such, the early identification of neuropathological cues may provide a more targeted approach to therapy in JNCL. In contrast, lamina II of Lent displayed a prominent loss of neurons, even at one week of age (Figure 6.3). Since lamina II of the entorhinal cortex projects to the molecular layer of the dentate gyrus and this is the major route between the neocortex and the hippocampus (Witter and Amaral, 1991), such patterns of neuronal loss may compromise the afferent input to the hippocampus. Indeed, the hippocampus is crucially involved in learning and memory (Kim and Diamond, 2002), which are properties known to be impaired in human JNCL (Stengel, 1826; Zeman et al., 1970).

189 7.8 Selective vulnerability of interneurons in mouse models of JNCL

Having already described the vulnerability of excitatory pyramidal and granule cell populations in mouse models of JNCL, inhibitory interneuron populations should also be considered. Interneurons receive a network of inhibitory and excitatory connections, where the relative contributions of each stimulus vary according to neuronal subtype and the region of the brain (Markram et al., 1998; Wang et al., 1999). Interneurons also have multiple connections with excitatory neurons, including pyramidal cells in the neocortex (Amaral, 2000). Thus, impaired function of either excitatory or inhibitory neuronal populations could result in an inhibitory/excitatory imbalance of inputs to either GABAergic interneurons or excitatory neurons. Whichever side of the equation is affected (inhibitory or excitatory), depending on the extent of the imbalance, such changes may potentially lead to the loss of selective neuronal populations. Indeed, excitatory neurotransmission involves the binding of presynaptic excitatory amino acids (i.e. glutamate or aspartate) to their receptors on post-synaptic neurons, which leads to depolarization and an increase in physiological activity (Dodd, 2002). Conversely, inhibitory transmission involves the interaction of GABA and its receptor, which functions to moderate excitation and thereby moderate excessive depolarization (Dodd, 2002). Since the overstimulation of excitatory receptors (NMDA and non-NMDA) is associated with an increase in intracellular ions, this physiological imbalance may lead to osmotic swelling and cell rupture (Choi, 1988; Zeevalk et al., 1995), or alternatively an excitotoxic cascade may be mediated via Ca2+-dependant cellular processes (Choi, 1988).

Consistent with human JNCL (Braak and Goebel, 1978; Braak and Goebel, 1979; Tyynelä et al., 2004) and animal models of NCL (Cooper et al., 1999; Bible et al., 2004; Oswald et al., 2005), our findings in aged Cln3-/- mice show a selective loss of GABAergic interneurons in specific neocortical and hippocampal subfields (Figure 3.7; Pontikis et al., 2004). The extent of this interneuron loss was dependent not only on location, but on the phenotypic marker expressed within these neuron populations. When compared to other mouse models of NCL, interneuron loss in Cln3-/- mice was only subtle (e.g. Cooper et al., 1999; Bible et al., 2004). Nevertheless, our findings are

190 consistent with observations in human JNCL, which also revealed relatively subtle effects on interneuron populations when compared to the other forms of NCL (Tyynelä et al., 2004). The complex patterns of cell loss that were observed in the neocortex and hippocampus may reflect the heterogeneous nature of interneuron populations within these different CNS regions (Freund and Buzaki, 1996; Markram et al., 2004). Indeed, these neurons can be divided into various subtypes based on their differences in electrophysiology, morphology and connectivity, although there is considerable overlap between these properties (Freund and Buzaki, 1996; Markram et al., 2004). Thus, the relationship between these functional properties and the lack of functional Cln3 protein may impact interneuron survival in specific regions of the brain. Site-specific cues may also influence the fate of interneurons, rather than depending solely upon phenotypic identity. Indeed, the loss of excitatory neurons may be more prominent in particular parts of the brain, leading to a site-specific imbalance between excitatory and inhibitory input.

Other clues about why interneuron populations may be selectively vulnerable come from the recent identification of autoantibodies against glutamic acid decarboxylase (GAD65) in JNCL patients (Chattopadhyay et al., 2002b; Chattopadhyay et al., 2002b) and Cln3-/- mice (Chattopadhyay et al., 2002a). This autoantibody is capable of inhibiting the activity of GAD65, which is an essential enzyme for the synthesis of GABA from glutamate (Chattopadhyay et al., 2002a). The resultant accumulation of excitatory transmitter, glutamate may potentially be toxic to surrounding neurons. Furthermore, the depletion of inhibitory transmitter, GABA may also have deleterious effects on membrane depolarization and again lead to loss or dysfunction of these neuron populations. Nevertheless, autoantibodies to GAD65 are reportedly absent from all other forms of NCL (Chattopadhyay et al., 2002b), despite the severe loss of GABAergic interneurons seen in these disorders (Cooper et al., 1999; Bible et al., 2004; Tyynelä et al., 2004; Oswald et al., 2005). Taken together, these data suggest that GAD65 autoantibodies may play only a minor contributory role, if any, in the loss of these interneuron populations. Indeed, since GABAergic interneurons are proposed to have far greater metabolic rates, oxygen demand and firing rates than other neurons, these properties may instead account for their vulnerability in the NCLs (Houser et al., 1984; Walkley et al., 1993).

191

In contrast to patterns of neuronal loss in human JNCL and in Cln3-/- mice, interneuron populations in Cln3∆ex7/8 homozygotes were not significantly affected (Figure 4.5). Instead, more subtle neurodegenerative changes were observed in various neocortical and hippocampal subfields. The lack of significant interneuron loss may be characteristic of these mice and reflect the different strain background, or alternatively, this phenotype may take longer to develop.

7.9 Loss of thalamic neurons in mouse models of JNCL

Our analysis of regional volume was informative revealing thalamic atrophy in homozygous Cln3∆ex7/8 mice (Figure 4.2), a phenotype that is also apparent in PPT1- deficient mice (Bible et al., 2004). Although the thalamus typically displays a reduced MRI signal intensity in symptomatic INCL and JNCL patients (Autti et al., 1997; Vanhanen et al., 2004), the cellular basis of this hypointensity has remained unknown. Indeed, in contrast to the wealth of information available for the neocortex and hippocampus (Braak and Goebel, 1978; Braak and Goebel, 1979; Haltia, 2001; Haltia et al., 2003; Tyynelä et al., 2004), neuropathological data of neuronal loss within subcortical structures in JNCL has been extremely limited (Braak and Braak, 1987; Braak et al., 1979). As such, our data from both Cln3-/- and homozygous Cln3∆ex7/8 mice are significant in providing the first direct evidence for the loss of thalamic relay neurons in this disorder. It is unclear whether thalamic neurons are themselves inherently vulnerable in JNCL, or degenerate in response to unknown pathological events. However, analysis of different ages in mutant mice has provided some insight into the sequence of neurodegenerative events in the thalamocortical system (Chapters 3-6; Figure 7.1).

192 a) Loss of thalamic neurons in the Cln3-/- mouse model of JNCL

In Cln3-/- mice, the loss of thalamic relay neurons from VPM/VPL was a relatively early event that was evident by 5 months of age (Figure 3.6), but was not apparent at one week of age (Figure 5.3). In contrast, neurons of the somatosensory cortex were only lost at the terminal stages of the disease and this was confined to lamina V pyramidal neurons (Figure 3.5). As such, these neurodegenerative events within the thalamus clearly precede the loss of neocortical neurons, but the precise cause of neuronal loss in the thalamocortical system is unclear. Nevertheless, despite the preservation of lamina IV granule neurons until late in the disease, it is possible that these cells may be functionally abnormal in younger animals and not capable of providing adequate trophic support to thalamic relay neurons resulting in their depletion. Indeed, neurons have been shown to respond to trophic factors during early stages of development, and then become more dependent on them for survival as the embryo matures (Rodriguez-Tebar et al., 1989). Timing of trophic dependency has also been shown to coincide with the timing of target innervation (Johnson et al., 1986; Cohen-Cory and Fraser, 1994). In order to determine whether trophic support is indeed dysfunctional in Cln3-/- mice, expression studies could be performed on laser capture dissected neurons from lamina IV of S1BF. Such studies would be highly informative and potentially offer a new insight into the mechanisms of neuronal loss in JNCL.

The loss of thalamic relay neurons in Cln3-/- mice was not confined to the somatosensory system, but also extended to neurons within the dorsal lateral geniculate nucleus (LGNd) that relay information to V1 (Livingstone and Hubel, 1988; Sincich and Horton, 2004). These neuropathological changes were present in aged mutant mice (Figure 3.5), but were absent at 14 months of age (Figure 3.6), revealing that the timing of thalamic neuron loss may differ between sensory modalities. Furthermore, this loss of visual relay neurons in LGNd of Cln3-/- mice occurred in the absence of significant effects upon their neocortical target neurons in lamina IV of V1 (Figure 3.5). It has long been accepted that damage to V1 results in death of projection neurons in LGNd (Lashley, 1941; Matthews, 1973; Agarwala and Kalil, 1998a) and that these effects may be attributable to loss of

193 194

Figure 7.1 Diagrammatic representation of neuronal loss in the thalamocortical system. Thalamic neurons in the ventral posterior nucleus (VPM/VPL, red) send projections to lamina IV granule neurons and receive afferent input from lamina V pyramidal neurons (blue) of somatosensory cortex. Laminae II and III neurons supply commissural and association projections to other areas of neocortex. (A) The relative timing of neurodegenerative events are now established in Cln3-/-mice. (B) Selective neuronal loss also occurred in this system in homozygous Cln3∆ex7/8 mice, although timing of these events remains to be determined. GABAergic interneurons are shown in green and roman numerals (I- VI) indicate individual laminae.

trophic support after target removal (Eagleson et al., 1990; Agarwala and Kalil, 1998b). Nevertheless, our present findings reveal that thalamic neuron loss in both somatosensory and visual nuclei occurs before loss of their target neurons in lamina IV. Similarly, neurodegenerative changes within LGNd of Cln3-/- mice have been shown to precede cell loss in the retina, but during this period of disease progression there was also a reduction of nerve conduction in the optic nerve (Weimer et al., 2005a). Thus, neuronal loss within the visual system is likely to initiate in LGNd, but other functional abnormalities may be present elsewhere within the visual system prior to this loss of thalamic relay neurons.

b) Loss of thalamic neurons in the homozygous Cln3∆ex7/8 mouse model of JNCL

In homozygous Cln3∆ex7/8 mice, the loss of thalamic relay neurons from VPM/VPL was evident late during pathogenesis of disease (Figure 4.4), but it is currently unclear when this phenotype first becomes apparent, or what mechanisms trigger the loss of VPM/VPL neurons. Somatosensory lamina IV neurons were also lost in aged Cln3∆ex7/8 mice (Figure 4.4), but were unaffected at one week of age (Figure 6.3). Since lamina IV granule neurons from S1BF receive sensory information from the thalamus, loss of this neuron population may have deleterious effects upon the survival of their afferent neuron populations within the thalamus. Equally, the selective loss of these somatosensory relay neurons could possibly occur after a loss or impairment of afferent sensory input from the spinal cord. However, this explanation seems implausible since no overt clinical effects upon the peripheral nervous system have been reported in JNCL. Alternatively, the loss of VPM/VPL neurons may simply reflect an inherent vulnerability of this neuron population, for reasons that remain unclear at present. To distinguish between these mechanistic possibilities, it will be important to determine the precise sequence of events within the thalamocortical system of Cln3∆ex7/8 mice. It remains to be seen whether the loss of thalamic neurons in these mice follows a similar time course to that seen in Cln3 null mutants. However, these experiments must await the availability of both mouse models on the same strain background.

195 In other neurodegenerative disorders including Alzheimer and Parkinson disease, the extensive loss of neurons has been associated with a decrease in synaptic contacts or other dendritic abnormalities (Catala et al., 1988; Masliah et al., 1989; Moolman et al., 2004). The loss of various neuron populations is now well described in the NCLs (Braak and Goebel, 1978; Braak and Goebel, 1979; Bible et al., 2004; Pontikis et al., 2004; Mitchison et al., 2004; Tyynelä et al., 2004; Oswald et al., 2005; Pontikis et al., 2005), but dendritic or synaptic defects are rarely reported and are markedly different to that observed in other lysosomal storage disorders (Walkley, 1998). Indeed, the storage of GM2 gangliosidoses in Tay Sachs disease leads to the growth of an abnormal dendritic membrane, whereas the formation of meganeurites and axonal swellings are noted in INCL, but these do not contain new dendritic outgrowth (Purpura, 1976). In PPT1-/- mice these abnormalities extend to the altered expression of presynaptic markers synaptophysin and α-synuclein, which is observed in neocortical regions associated with decreased neuronal number (Cooper, 2005). Although, such abnormalities have not been described in human JNCL or mouse models of the disease, it will be highly informative to investigate synaptic morphology and function in the thalamocortical system where neuronal loss is prominent and an early event in disease pathogenesis.

The clinical consequences of neurodegenerative events within the thalamocortical system remains unclear, although progressive changes in somatosensory evoked potentials are consistently reported in multiple forms of NCL (Tackmann and Kuhlendahl, 1979; Vercruyssen et al., 1982; Vanhanen et al., 1997; Lauronen et al., 1997), including JNCL (Lauronen et al., 1997 and Lauronen et al., 1999). Indeed, it has been suggested that progressive thalamic dysfunction may also contribute to the sleep disturbances that are characteristic of the NCLs (Vanhanen et al., 1997). It will be highly informative to determine whether the loss of thalamic neurons and/or their neocortical targets/feedback neurons contributes to similar phenotypes in Cln3-/- and homozygous Cln3∆ex7/8 mice. Visual processing may also be affected by pathological changes within the thalamocortical system. Although previous studies suggest the retina to be the primary sight of insult to the visual system in the NCLs (Goebel et al., 1974; Goebel, 1992; Bensaoula et al., 2000), determining whether degenerative changes in the LGNd of Cln3-

196 /- mice also extends to the human condition, could provide further evidence that the primary sight of insult lies outside the retina. Indeed, these findings would have important implications for targeting therapies centrally rather than to the retina in JNCL.

Problems with motor coordination have already been reported by 12 months of age in Cln3∆ex7/8 homozygotes (Cotman et al., 2002), but this current study demonstrated no overt effects within the cerebellum of these mice. However, recent findings in Cln3-/- mice suggest that cell proliferation and migration are altered within the cerebellum early during neural development (Weimer et al., 2005c). Indeed, there may be subtle effects on other neuronal populations, but our data suggest that neurological problems in Cln3-/- and homozygous Cln3∆ex7/8 mice may involve altered sensory feedback through the thalamus. In this context, it will be important to correlate sensory behaviors and other forms of neurological impairment with pathological changes in the thalamocortical system during disease progression.

7.10 Glial responses in mouse models of JNCL

Pronounced gliosis is a consistent feature of autopsy material from individuals with JNCL (Braak and Goebel, 1978, Braak and Goebel, 1979; Tyynelä et al., 2004) and other forms of NCL (Haltia et al., 1973a, Haltia et al., 1973b, Herva et al., 2000, Tyynelä et al., 1997; Tyynelä et al., 2004). Similar reactive events have also been demonstrated in mouse models of NCL including CLN8 (Cooper et al., 1999), and infantile NCL (Bible et al., 2004; Jalanko et al., 2005). Consistent with these findings, glial responses were prominent in 12 month old homozygous Cln3∆ex7/8 mice (Pontikis et al., 2005), but a more subtle glial phenotype was evident in Cln3-/- mice (Pontikis et al., 2004).

197 7.10.1 Astrocytic responses in mouse models of JNCL

Astrocytes are highly complex and have a diverse spectrum of proposed roles that are essential in the normal and diseased CNS. Such roles include regulating fluid homeostasis (Simard and Nedergaard, 2004) and controlling local concentrations of glutamate, thereby influencing neuronal activity and synaptic efficacy (Oliet et al., 2001; Piet et al., 2002). Astrocytes are classically known to mediate responses to neuronal injury (Raivich et al., 1999). Astrocytosis in a pathological state may thus represent an attempt to maintain a homeostatic balance within the CNS. Indeed, these reactive changes are capable of restricting glutamate induced excitotoxicity from activated microglia so that optimal neuronal responses can be maintained. Clearance of extracellular glutamate occurs via high affinity glutamate transporters that are enriched in astrocytic processes (Rothstein et al., 1996; Duan et al., 1999). Alternatively, astrocytes can release glutamate from internal stores in response to increased levels of Ca2+, thereby modulating synaptic transmission (Cornel-Bell et al., 1990; Haydon et al., 2001).

a) Astrocytic responses in the Cln3-/- mouse model of JNCL

Aged Cln3-/- mice exhibited a more pronounced astrocytosis than littermate controls, but, the morphological appearance of astrocytes was remarkably similar in animals of either genotype (Figures 3.8, 3.9). Since astrocytosis in neocortical M1 and subcortical thalamus was accompanied by prominent neuronal loss, yet limited evidence of morphological transformation, astrocytes may be incapable of normal responses. Thus, an increased level of astrocytosis may represent an attempt to compensate for compromised astrocytic function. In contrast to other brain regions, GFAP expression in the hippocampus was markedly reduced in mutant mice, which may indicate that cues mediating astrocytosis have subsided within the hippocampus. To determine whether this reduction of GFAP-positive astrocytes extends to the total number of astrocytes, S100β staining could be assessed in a similar manner to that in homozygous Cln3∆ex7/8 mice (Pontikis et al., 2005).

198 Interestingly, altered expression of GFAP was also evident in Cln3-/- mice as early as one week of age (Figures 5.4, 5.5). At this time patterns of connectivity are being fine tuned (Sur and Leamey, 2001) and radial glia are differentiating either into astrocytes or neurons (Voigt, 1989; Gotz et al., 2002; Tramontin et al., 2003). As such, the relative abundance of GFAP-positive radial glial cells in the neocortex of P8 Cln3-/- mice may represent a delay in the differentiation of these cells. Alternatively, abnormal molecular cues in mutant mice may lead to an increased number of radial glia to form from precursor cells within the neocortex. Since radial glia can differentiate into neurons and are pivotal for the tangential migration of neurons in the developing neocortex (Weissman et al., 2003), the fact that there is a normal number of neurons in P8 Cln3-/- mice suggests that differentiation of radial glia into neurons is normal. Furthermore, these neocortical neurons also appear to be correctly targeted to their corresponding laminae.

The possible causes for the relative abundance of radial glia in the developing neocortex of Cln3-/- mice may arise from deficiencies in molecular cues that govern the maturation of radial glia. Indeed, radial glial cell identity in the mammalian forebrain is determined by the availability of diffusible inducing signals from embryonic neocortical cells (Hunter and Hatten, 1994). Furthermore, these signals may also transform mature astrocytes back to a radial glial phenotype (Hunter and Hatten, 1994). Thus multiple possibilities may account for the increased abundance of GFAP-positive radial glia in P8 Cln3-/- mice. To gain further insights into these processes it will be informative to study whether markers expressed by radial glia and astrocytes at different stages of development (e.g. GLAST, BLBP and RC1) are regulated normally in Cln3 deficient mice.

The neocortical distribution of GFAP-positive astrocytes was more widespread in superficial laminae, but was largely confined to the deeper laminae of P8 mutant mice (Figure 5.4). However, the cues regulating this lamina specific distribution of astrocytes within the neocortex remain unclear. Nevertheless, as astrocytes are classically known to mediate responses to neuronal injury (Raivich et al., 1999), such patterns of astrocytosis may suggest early damage within deeper and superficial laminae of these mice.

199 Consistent with the localized astrocytic responses found within the aging brain (Figures 3.8, 3.9), GFAP expression in the hippocampus was reduced in Cln3-deficient mice and further suggests that cues mediating astrocytosis have subsided within the hippocampus, or that astrocytes are being targeted by the disease process (Figures 5.4, 5.5). Perivascular GFAP staining was also less apparent in the hippocampus of Cln3-/- brains. Such staining patterns may be of particular interest since astrocytes have been implicated in the maintenance of the blood brain barrier (BBB, Simard and Nedergaard, 2004), which is reportedly breached in Cln3-/- mice (Lim et al., 2005a). Indeed, a size selective breach in BBB integrity has been demonstrated by intravenous labeling studies and the uptake of intraperitoneally injected human IgGs into the CNS of Cln3-/- mice. Moreover, an increased immunoglobulin G (IgG) deposition is evident in the CNS of Cln3-/- mice and individuals with end stage JNCL (Lim et al., 2005a). Thus, taken together, the reduction of perivascular GFAP in these mice may be indicative of an improperly formed BBB, but again this suggestion awaits experimental verification.

b) Astrocytic responses in the homozygous Cln3∆ex7/8 mouse model of JNCL

Our data from aged Cln3∆ex7/8 homozygotes provide a distinctly different picture of reactive changes in these mice, with widespread hypertrophy of astrocytes across the cortical mantle and variable responses subcortically (Figures 4.6, 4.7). Curiously, there was little evidence of astrocytic proliferation or hyperplasia in homozygous Cln3∆ex7/8 mice, with no difference in the number of S100β-positive astrocytes between animals of either genotype (Figures 4.6, 4.7). These S100β data also suggest that reduced GFAP immunoreactivity evident in the cerebellum and hippocampus of homozygous Cln3∆ex7/8 mice appears to represent downregulation of this marker rather than the targeting of astrocytes that has been proposed in Cln3-/- (Pontikis et al., 2004).

Previous findings suggest that all astrocytes are S100β-positive, but the inactive protoplasmic variety is mainly GFAP-negative (Didier et al., 1986). Following neuronal

200 injury, reactive changes are typically observed in protoplasmic astrocytes, such as the rapid increase in synthesis of GFAP (Graeber and Kreutzberg, 1986; Laskawi and Wolff, 1996). This process is followed by increased appearance of glial filaments and astrocytic processes, which become more numerous and thicker as astrocytes transform into the reactive fibrous type (Graeber and Kreutzberg, 1986). Since the absence of interleukin 6 (IL6) has been shown to interfere with the appearance of fibrous astrocytes (Klein et al., 1997) and neurons are capable of synthesizing chemokines (Loddick and Rothwell, 2002), morphological transformation of these reactive cells may be governed by cytokines secreted from injured neurons. Thus, partial transformation of astrocytes in either mouse model of JNCL may in fact suggest that neuronal injury may be subtle, or that injured neurons are unable to secrete sufficient levels of cytokines for the full morphological transformation of astrocytes.

Although the localized events that trigger astrocytosis in homozygous Cln3∆ex7/8 mice are complex, these may be initiated by neuronal damage due to accumulated storage material. However, in contrast to the widespread accumulation of storage material (Figure 4.1), astrocytic activation within the thalamus was restricted to individual nuclei, with particularly prominent GFAP staining within the VPM/VPL which also exhibited neuronal loss (Figure 4.6). These reactive changes may represent a localized protective response to the effects of the Cln3 mutation, as has been suggested in human JNCL (Tyynelä et al., 2004). In this respect, despite the analysis of both young and aged mutants, investigating the precise spatiotemporal relationship between astrocytosis and neuronal loss at intermediate stages of disease progression is likely to be highly informative. As described above for Cln3-/- mice, in early postnatal Cln3∆ex7/8 homozygotes, patterns of connectivity are being fine tuned (Sur and Leamey, 2001) and radial glia are differentiating into astrocytes or neurons (Voigt, 1989; Gotz et al., 2002; Tramontin et al., 2003). Since GFAP-positive radial glia were apparently more abundant in Cln3∆ex7/8 homozygotes than in controls (Figure 6.4), this may suggest that triggers of radial glial differentiation are abnormal or inadequate, or that radial glia are also targeted by the disease process. At present, the cellular basis of timing and progressive maturation of

201 radial glia is still largely undefined. However, these developmental processes are likely to involve complex interactions between environmental signals, cell-cell interactions and transcriptional regulatory events (Mehler, 2002).

During this period of early postnatal development, the morphological appearance of GFAP-positive astrocytes was subtly different between homozygous Cln3∆ex7/8 mice and controls. However, regional expression of GFAP was highly variable, with an elevation in the cerebrum and a reduction in the cerebellum (Figures 6.4, 6.5). Compared to the neocortex, the cerebellum is still actively developing in P7 mice (Wang and Zoghbi, 2001) and astrocytes have been implicated in neurite outgrowth within this region (Hatten et al., 1984). As such, a reduction of GFAP in the cerebellum may suggest that astrocytes are functionally abnormal and may potentially impair neurite growth during development. In contrast, because cerebral patterns of astrocytosis were pronounced and mirrored that of F4/80-positive microglia (Chapter 6), it is possible that similar cues are controlling both cell types. As such, the ability of astrocytes to clear extracellular glutamate (Rothstein et al., 1996; Duan et al., 1999) may offer neuroprotection from this excitatory neurotransmitter which is released from activated microglia (Koutsilieri et al., 1999), or increased as part of the autoimmune response (Chattopadhyay et al., 2002a). However, it is worth noting that similar to older mice, there was little evidence of astrocytic proliferation or hyperplasia at this age, with no difference in the number of S100β- positive astrocytes between animals of either genotype (Figure 6.5). Although it is unlikely that astrocytes are recruited into areas of neuronal injury, as cells are already in close contact with neurons (Haydon, 2001), by changing morphology this may offer some degree of protection.

7.10.2 Microglial responses in mouse models of JNCL

The functional role of microglia is proposed to range from homeostatic regulation to the repair of damaged cells (Kreutzberg, 1996). Microglia are typically found in a resting state, but rapidly become activated in response to even minor pathological changes within

202 the CNS (Kreutzberg, 1996). The activation of microglia is a progressive and graded phenomenon that takes place in different ‘stages’ from extensively ramified cells to full blown brain macrophages (Raivich et al., 1999) and as such has been considered a sensitive marker of local neuronal damage (Streit, 2000, Streit, 2002). Once activated, microglia may mainly act as scavenger cells (Kreutzberg, 1996; Hirt et al., 2000), although they have also been implicated in tissue repair and regeneration of neurons (Kreutzberg, 1996; Batchelor et al., 2002). Indeed, activated microglia increase in mobility as they migrate towards damaged cells (Heppner et al., 1998), secrete neurotrophic factors, cytokines, glutamate, reactive oxygen and nitrogen species, and when repair is no longer feasible assist in the clearance of dying cells (Kreutzberg, 1996; Heppner et al., 1998; Koutsilieri et al., 1999; Hirt et al., 2000; Liu and Hong, 2003).

a) Microglial responses in the Cln3-/- mouse model of JNCL

Dissimilar to patterns of astrocytosis, aged Cln3-/- mice displayed a prominent and widespread distribution of F4/80-positive microglia in structures throughout the brain. Indeed, F4/80 can stain both activated and nonactivated microglia but its expression becomes increased after activation (Chen et al., 2005). Consistent with the expression of F4/80, microglia showed signs of partial activation in Cln3-/- mice, but rarely reached their full extent to reveal brain macrophage morphology, not even in areas where neuronal loss and/or astrocytosis were pronounced (Chapter 3; Pontikis et al., 2004). Taken together with variable and sometimes opposing expression patterns of glial markers in the brain (Chapter 3), these findings may be indicative of either multiple or separate cues controlling astrocytosis and microglial activation. Indeed, the absence of IL6 in genetically IL6-deficient mice has been shown to interfere with the number of activated GFAP-positive astrocytes, but microglial activation and proliferation was only moderately affected in these mice (Klein et al., 1997). Therefore, such cues are likely to differ in the extent of their influence on astrocytosis and microglial activation in a regionally dependant manner. In order to explore this regional variation, microarray data could be obtained from regions with variable patterns of glial activation. Subsequent

203 assessment of gene expression could identify potential candidate genes and lead to the strategic generation of null mutants for these potential signaling cues, and help determine their precise influence on glial activation.

The elevated expression of F4/80 coincided with neuronal loss in the neocortex and thalamus of aged mutants (Chapter 3). Since neuronal loss leads to a further transformation of microglia into phagocytotic cells (Streit et al., 1988; Möller et al., 1996), these findings are consistent with an early, but progressive loss of thalamic neurons. Although the loss of neocortical neurons is a later event in JNCL pathogenesis, this is likely to be part of a vicious cycle of neurodegenerative events. F4/80-positive staining was also prominent within the hippocampus (Figure 3.11) and coincided with interneuron loss within selective subfields (Figure 3.7). Since astrocytosis was reduced within the hippocampus, this may have potentially left these regions vulnerable to excitotoxic insult. Alternatively, activated microglia themselves may have exerted a positive influence on neurons which may have been injured via unknown pathological mechanisms.

Early activation of microglia has been reported in a sheep model of NCL (Oswald et al., 2005). These changes appear to start prenatally in the white matter and then extend to the neocortical grey matter shortly after birth (Kay et al., unpublished observations; Oswald et al., 2005). Similarly, microglial responses were pronounced in Cln3-/- mice during the early postnatal period, but these pathological changes were mostly evident in the grey matter and there was little morphological evidence for activation of these cells (Figures 5.6, 5.7). Nevertheless, subtle changes in microglial populations preceded patterns of neuronal loss in all evaluated regions of the CNS (Pontikis et al., 2004). Thus any connection between these two pathological events is likely to be complex and may be difficult to determine.

Curiously, F4/80 was significantly down regulated in the hippocampus and subcortical regions of P8 Cln3-/- mice when compared to controls (Figure 5.6, 5.7). The pronounced down regulation of this microglial marker may support the concept of microglia being

204 targets of the disease (Pontikis et al., 2004), and may indicate that such changes occur early during postnatal development and proceed into adulthood. Since patterns of F4/80 expression largely followed GFAP staining in the neocortex and hippocampus, but were variable elsewhere in the brain, it is again likely that molecular cues may differ in the extent of their influence on astrocytosis and microglial activation in a regionally dependant manner.

b) Microglial responses in the homozygous Cln3∆ex7/8 mouse model of JNCL

In contrast to the prominent astrocytic responses in aged Cln3∆ex7/8 homozygotes, microglial responses in these mice were not as pronounced (Figure 4.8). Furthermore, microglial activation rarely reached its full extent to reveal brain macrophage morphology, even in areas where a profound astrocytic response was evident. These data suggest that whatever underlying molecular cues trigger astrocytosis in homozygous Cln3∆ex7/8 mice, these are not sufficient to promote full morphological transformation of microglia. Alternatively, microglia may themselves be targeted by the effects of the Cln3 mutation, as has been suggested in Cln3-/- mice (Pontikis et al., 2004).

Neuron-glial interactions are also complex, since microglial responses were prominent in the neocortex and less marked in subcortical structures (Figure 4.8), but neuronal loss was still reported in both of these regions (Figure 4.4). As such microglial activation and neuronal loss appear to be independent events within this mouse model. Indeed, other triggers of neuronal loss may be acting separately or even in tandem with glial responses to result in this neurodegenerative phenotype. It is unclear at present what these triggers may be, but injured neurons have been shown to rapidly change their gene expression and stimulate nearby microglia and astrocytes for a supportive response (Kreutzburg, 1995; Raivich et al., 1995).

205 Early evidence of microglial activation has also been observed in Cln3∆ex7/8 homozygotes and is consistent with reactive changes reported in the early postnatal period in the sheep model of NCL (Oswald et al., 2005). However, as in older mice, the microglial activation in P7 Cln3∆ex7/8 homozygotes rarely displayed transformation of these cells to their full extent to reveal brain macrophage-like morphology (Figure 6.6). Since the patterns of F4/80 expression followed comparable trends to astrocytosis in the cerebrum, these data suggest that similar molecular cues may govern the activation of microglia and astrocytes outside of the cerebellum of these mice. Alternatively, injured neurons may activate microglia to secrete cytokines that may in turn influence the morphological transformation and activation of astrocytes.

It is unclear whether the early postnatal reduction in the number of granule neurons within the entorhinal cortex of Cln3∆ex7/8 homozygotes was due to cell death within this neuron population or their improper migration into this lamina. In either instance, because microglia displayed negligible morphological transformation within this neocortical subfield, the underlying mechanisms are likely to occur independently of microglial activation. However, it is also possible that prenatal activation of microglia may have led to this particular phenotype. Indeed, most remaining brain regions where neuronal loss was not yet evident, exhibited prominent microglial responses (Chapter 6). Thus similar to Cln3-/- mice, glial changes in Cln3∆ex7/8 mice again preceded neuronal loss (Pontikis et al., 2004), but such reactive changes were not always an accurate predictor of a neuron loss.

It is generally accepted that prolonged exposure to activated microglia may lead to neurodegeneration, although this is not always the case and such responses may also be regenerative to damaged neurons (Gebicke-Haerter, 2001). The precise causes of neuronal loss in JNCL remain unclear, but during neuronal injury activated microglia are also capable of displacing synaptic input (i.e. synaptic stripping, Graeber and Kreutzburg, 1988), before increasing the expression of receptors for microglial mitogens (i.e. microglial growth factors, Raivich et al., 1991; Raivich et al., 1998) with subsequent proliferation (Graeber et al., 1998). If additional damage is avoided and neurons are

206 allowed to recover, then microglia gradually decrease in number, lose activation markers and revert to a resting state (Raivich et al., 1999).

Our current understanding of the connection between microglial activation, astrocytosis and the neurodegenerative changes in JNCL remains unclear, and we can only postulate whether these events represent a degenerative response, regenerative response, or perhaps a combination of the two. To help resolve this issue, mice deficient in microglia or astrocytes could potentially be crossed with a given mouse model of JNCL. Since astrocytes are essential for the maintenance of the BBB (Simard and Nedergaard, 2004) and microglia are believed to assist in the clearance of dying cells during brain remodelling (Liu and Hong, 2003), glial-deficient mice may be likely to die at birth. However, an alternative strategy would be to generate conditional knockouts of glial phenotypes, thus enabling the appropriate genes to be inactivated postnatally, with subsequent analysis of pathological progression.

7.11 Conclusions

The assessment of progressive neuropathological changes in two distinct mouse models of JNCL has revealed a pronounced neurodegenerative phenotype at terminal stages of the disease, which was more subtle during the early postnatal period of development. Such changes included the widespread deposition of autofluorescent storage material, regional atrophy, neuronal loss and glial activation, all of which are characteristic of human JNCL (Goebel and Wisniewski, 2004; Tyynelä et al., 2004). Of these changes, the thalamocortical system was consistently targeted by the disease, but the precise clinical consequences of such events remain unclear.

Taken together with data from other mouse models of NCL (Cooper et al., 1999; Lam et al., 1999; Mitchison et al., 1999; Bible et al., 2004), sheep with NCL (Oswald et al., 2001; Oswald et al., 2005; Kay et al., unpublished observations) and human NCL (Tyynelä et al., 2004), our data suggest that highly regionalized interactions occur

207 between neurons and glia at different stages of disease progression. Whether these events contribute directly to subsequent neurodegeneration has not been demonstrated, but the suggestion that neuron–glial interactions differ markedly between CNS regions may provide important clues to the pathogenesis of these disorders.

208 References

Agarwala S, Kalil RE (1998a). Axotomy-induced neuronal death and reactive astrogliosis in the lateral geniculate nucleus following a lesion of the visual cortex in the rat. J Comp Neurol. 392, 252–263

Agarwala S, Kalil RE (1998b). Long-term protection of axotomized neurons in the dorsal lateral geniculate nucleus in the rat following a single administration of basic fibroblast growth factor or ciliary neurotrophic factor. J Comp Neurol. 392, 264–272

Alberts B, Bray D, Lewis J, Raff M, Roberts K, Watson (1994).Vesicular Traffic in the Secretory and Endocytic Pathways. In: Molecular Biology of The Cell, 3rd ed. Garland Publishing Press, pp. 599-651

Amaral DG (2000). The anatomical organization of the central nervous system. In: Kandel ER, Schwartz JH, Jessel TM (Eds.) Principles of neural science, 4th ed. McGraw- Hill, pp. 317-366

Antunes S.M.G. (1995). The optical fractionator. Zoological Studies. 34, 180-183

Aronson NN Jr, Kuranda MJ (1989). Lysosomal degradation of Asn-linked glycoproteins. FASEB J. 3, 2615-2622

Artavanis-Tsakonas S, Rand MD, Lake RJ (1999). Notch signaling: cell fate control and signal integration in development. Science. 284, 770–776

Autti T, Raininko R, Santavuori P, Vanhanen SL, Poutanen VP, Haltia M (1997). MRI of neuronal ceroid lipofuscinosis: II. Postmortem MRI and histopathological study of the brain in 16 cases of neuronal ceroid lipofuscinosis of juvenile or late infantile type. Neuroradiology. 39, 371–377

209 Autti T, Raininko R, Vanhanen SL, Santavuori P (1996). MRI of neuronal ceroid lipofuscinosis. I. Cranial MRI of 30 patients with juvenile neuronal ceroid lipofuscinosis. Neuroradiology. 38, 476-482

Bach G, Zeigler M, Schaap T, Kohn G (1979). Mucolipidosis type IV: ganglioside sialidase deficiency. Biochem Biophys Res Commun. 90, 1341-1347

Bäckman M, Tyynelä J, Härkönen T, Knip M, Peltola J, Åberg L (2005). GAD-antibodies in Juvenile Neuronal Ceroid Lipofuscinosis; Immunomodulation trial. Abstract. NCL 2005 O45

Batchelor PE, Porritt MJ, Martinello P, Parish CL, Liberatore GT, Donnan GA, Howells DW (2002). Macrophages and Microglia Produce Local Trophic Gradients That Stimulate Axonal Sprouting Toward but Not beyond the Wound Edge. Mol Cell Neurosci. 21, 436-453

Batten FE (1903). Cerebral degeneration with symmetrical changes in the maculae in two members of a family. Trans Ophthalmol Soc UK. 23, 386-390

Bauchot R (1967). Modifications of brain weight in the course of fixation. J Hirnforsch. 9, 253-283

Bellizzi JJ 3rd, Widom J, Kemp C, Lu JY, Das AK, Hofmann SL, Clardy J (2000). The crystal structure of palmitoyl protein thioesterase 1 and the molecular basis of infantile neuronal ceroid lipofuscinosis. Proc Natl Acad Sci U S A. 97, 4573-4578

Benedict JW, Weimer JM, Pearce DA (2005). Interacting partners of CLN3 assist in the understanding of Batten disease. Abstract. NCL 2005 O41

Bensaoula T, Shibuya H, Katz ML, Smith JE, Johnson GS, John SK, Milam AH (2000). Histopathologic and immunocytochemical analysis of the retina and ocular tissues in Batten disease. Ophthalmology. 107, 1746-1753

210 Bernar J, Tietze F, Kohn LD, Bernardini I, Harper GS, Grollman EF, Gahl WA (1986). Characteristics of a lysosomal membrane transport system for tyrosine and other neutral amino acids in rat thyroid cells. J Biol Chem. 261, 17107-17112

Bible E, Gupta P, Hofmann SL, Cooper JD (2004). Regional and cellular neuropathology in the palmitoyl protein thioesterase-1 null mutant mouse model of infantile neuronal ceroid lipofuscinosis. Neurobiol Dis. 16, 346-359

Bishop KM, Rubenstein JL, O'Leary DD (2002). Distinct actions of Emx1, Emx2, and Pax6 in regulating the specification of areas in the developing neocortex. J Neurosci. 22, 7627-7638

Blinzinger K, Kreutzberg G (1968). Displacement of synaptic terminals from regenerating motorneurons by microglial cells. Z Zellforsch Mikrosk Anat. 85, 145-157

Bolivar VJ, Scott Ganus J, Messer A (2002). The development of behavioral abnormalities in the motor neuron degeneration (mnd) mouse. Brain Res. 937, 74-82

Boyce S, Webb JK, Carlson E, Rupniak NM, Hill RG, Martin JE (1999). Onset and progression of motor deficits in motor neuron degeneration (mnd) mice are unaltered by the glycine/NMDA receptor antagonist L-701,324 or the MAO-B inhibitor R(-)-deprenyl. Exp Neurol. 155, 49-58

Boyes BE, Kim SU, Lee V, Sung SC (1986). Immunohistochemical co-localization of S- 100b and the glial fibrillary acidic protein in rat brain. Neuroscience. 17, 857-865

Braak H, Braak E (1987). Projection neurons of basolateral amygdaloid nuclei develop meganeurites in juvenile and adult human neuronal ceroid lipofuscinosis. Clin Neuropathol. 6, 116–119

Braak H, Goebel HH (1978). Loss of pigment-laden stellate cells: a severe alteration of the isocortex in juvenile neuronal ceroid-lipofuscinosis. Acta Neuropathol (Berl). 42, 53- 57

211 Braak H, Goebel HH (1979). Pigmentoarchitectonic pathology of the isocortex in juvenile neuronal ceroid-lipofuscinosis: axonal enlargements in layer IIIab and cell loss in layer V. Acta Neuropathol (Berl). 46, 79-83

Brady RO (1982). Inherited metabolic storage disorders. Annu Rev Neurosci. 5, 33-56

Brady RO, Quarles RH (1973). The enzymology of myelination. Mol Cell Biochem. 2, 23-29

Braulke T, Tippmer S, Neher E, von Figura K (1989). Regulation of the mannose 6- phosphate/IGF II receptor expression at the cell surface by mannose 6-phosphate, insulin like growth factors and epidermal growth factor. EMBO J. 8, 681-686

Bray S (1998). Notch signalling in Drosophila: three ways to use a pathway. Semin Cell Dev Biol. 9, 591–597

Bronson RT, Donahue LR, Johnson KR, Tanner A, Lane PW, Faust JR (1998). Neuronal ceroid lipofuscinosis (nclf), a new disorder of the mouse linked to chromosome 9. Am J Med Genet. 77, 289-297

Bronson RT, Donahue LR, Johnson KR, Tanner A, Lane PW, Faust JR (1998). Neuronal ceroid lipofuscinosis (nclf), a new disorder of the mouse linked to chromosome 9. Am J Med Genet. 77, 289-297

Bronson RT, Lake BD, Cook S, Taylor S, Davisson MT (1993). Motor neuron degeneration of mice is a model of neuronal ceroid lipofuscinosis (Batten's disease). Ann Neurol. 33, 381-385

Broom MF, Zhou C (2001). Fine mapping of ovine ceroid lipofuscinosis confirms orthology with CLN6. Eur J Paediatr Neurol. 5, 33-35

Brown WJ, Goodhouse J, Farquhar MG (1986). Mannose-6-phosphate receptors for lysosomal enzymes cycle between the Golgi complex and endosomes. J Cell Biol. 103, 1235-1247

212 Camp LA, Verkruyse LA, Afendis SJ, Slaughter CA, Hofmann SL (1994). Molecular cloning and expression of palmitoyl-protein thioesterase. J Biol Chem. 269, 23212-23219

Carlsson SR, Fukuda M (1992). The lysosomal membrane glycoprotein lamp-1 is transported to lysosomes by two alternative pathways. Arch Biochem Biophys. 296, 630- 639

Casey PJ (1995). Protein lipidation in cell signaling. Science. 268, 221-225

Catala I, Ferrer I, Galofre E, Fabregues I (1988). Decreased numbers of dendritic spines on cortical pyramidal neurons in dementia. A quantitative Golgi study on biopsy samples. Hum Neurobiol. 6, 255-259

Cavalieri B (1665). Geometrica indivisibilibus continuorum. Bononiae: Typis Clementis Ferronij (reprinted 1966 as Geometria Degli Indivisibili. Torino: Unione Tipografico- Editrice Torinese).

Chattopadhyay S, Ito M, Cooper JD, Brooks AI, Curran TM, Powers JM, Pearce DA. (2002a). An autoantibody inhibitory to glutamic acid decarboxylase in the neurodegenerative disorder Batten disease. Hum Mol Genet. 11, 1421-1431

Chattopadhyay S, Kriscenski-Perry E, Wenger DA, Pearce DA (2002b). An autoantibody to GAD65 in sera of patients with juvenile neuronal ceroid lipofuscinoses. Neurology. 59, 1816-1817

Chattopadhyay S, Pearce DA (2000). Neural and extraneural expression of the neuronal ceroid lipofuscinoses genes CLN1, CLN2, and CLN3: functional implications for CLN3. Mol Genet Metab. 71, 207-211

Chen Z, Duan RS, Quezada HC, Mix E, Nennesmo I, Adem A, Winblad B, Zhu J (2005). Increased microglial activation and astrogliosis after intranasal administration of kainic acid in C57BL/6 mice. J Neurobiol. 62, 207-218

Choi DW (1988). Calcium-mediated neurotoxicity: relationship to specific channel types and role in ischemic damage. Trends Neurosci. 11, 465-469

213 Cohen-Cory S, Fraser SE (1994). BDNF in the development of the visual system of Xenopus. Neuron. 12, 747-761

Constantopoulos G, Dekaban AS (1978). Neurochemistry of the mucopolysaccharidoses: brain lipids and lysosomal enzymes in patients with four types of mucopolysaccharidosis and in normal controls. J Neurochem. 30, 965-973

Constantopoulos G, Iqbal K, Dekaban AS (1980). Mucopolysaccharidosis types IH, IS, II, and IIIA: glycosaminoglycans and lipids of isolated brain cells and other fractions from autopsied tissues. J Neurochem. 34, 1399-1411

Conzelmann E, Sandhoff K (1978). AB variant of infantile GM2 gangliosidosis: deficiency of a factor necessary for stimulation of hexosaminidase A-catalyzed degradation of ganglioside GM2 and glycolipid GA2. Proc Natl Acad Sci U S A. 75, 3979-3983

Cooper (2005). Neuropathological landmarks of disease progression in the NCL CNS. Abstract. NCL 2005 O29

Cooper JD, Messer A, Feng AK, Chua-Couzens J, Mobley WC (1999). Apparent loss and hypertrophy of interneurons in a mouse model of neuronal ceroid lipofuscinosis: evidence for partial response to insulin-like growth factor-1 treatment. J Neurosci. 19, 2556-2567

Cooper JD, Payne JN, Horobin RW (1988). Accurate counting of neurons in frozen sections: some necessary precautions. J Anat. 157, 13-21

Cooper JD. (2003). Progress towards understanding the neurobiology of Batten disease or neuronal ceroid lipofuscinosis. Curr. Opin. Neurol. 16, 121-128

Cornell-Bell AH, Finkbeiner SM, Cooper MS, Smith SJ (1990). Glutamate induces calcium waves in cultured astrocytes: long-range glial signaling. Science. 247, 470-473

Cotman SL, Vrbanac V, Lebel LA, Lee RL, Johnson KA, Donahue LR, Teed AM, Antonellis K, Bronson RT, Lerner TJ, MacDonald ME (2002). Cln3(Deltaex7/8) knock-

214 in mice with the common JNCL mutation exhibit progressive neurologic disease that begins before birth. Hum Mol Genet. 11, 2709-2721

Crabbe JC, Metten P, Yu CH, Schlumbohm JP, Cameron AJ, Wahlsten D (2003). Genotypic differences in ethanol sensitivity in two tests of motor incoordination, J Appl Physiol. 95, 1338–1351

Das AK, Becerra CH, Yi W, Lu JY, Siakotos AN, Wisniewski KE, Hofmann SL (1998). Molecular genetics of palmitoyl-protein thioesterase deficiency in the U.S. J Clin Invest. 102, 361-370

Das AK, Lu J, Hofmann SL. (2001). Biochemical analysis of mutations in palmitoyl- protein-thiosesterase causing infantile and late-onset forms of neuronal ceroid lipofuscinoses. Hum Mol Gen. 10, 1431-1439

Davis KE, Straff DJ, Weinstein EA, Bannerman PG, Correale DM, Rothstein JD, Robinson MB (1998). Multiple signaling pathways regulate cell surface expression and activity of the excitatory amino acid carrier 1 subtype of Glu transporter in C6 glioma. J Neurosci. 18, 2475-2485

Didier M, Harandi M, Aguera M, Bancel B, Tardy M, Fages C, Calas A, Stagaard M, Mollgard K, Belin MF (1986). Differential immunocytochemical staining for glial fibrillary acidic (GFA) protein, S-100 protein and glutamine synthetase in the rat subcommissural organ, nonspecialized ventricular ependyma and adjacent neuropil. Differential immunocytochemical staining for glial fibrillary acidic (GFA) protein, S-100 protein and glutamine synthetase in the rat subcommissural organ, nonspecialized ventricular ependyma and adjacent neuropil. Cell Tissue Res. 245, 343-351

Dietzel I, Heinemann U, Lux HD (1989). Relations between slow extracellular potential changes, glial potassium buffering, and electrolyte and cellular volume changes during neuronal hyperactivity in cat brain. Glia. 2, 25-44

Dodd PR (2002). Excited to death: different ways to lose your neurones. Biogerontology. 3, 51-6.

215 Dorfman A, Matalon R (1976). The mucopolysaccharidoses (a review). Proc Natl Acad Sci U S A. 73, 630-637

D'Orlando C, Celio MR, Schwaller B (2002). Calretinin and calbindin D-28k, but not parvalbumin protect against glutamate-induced delayed excitotoxicity in transfected N18- RE 105 neuroblastoma-retina hybrid cells. Brain Res. 945, 181-190

Dorph-Petersen KA, Nyengaard JR, Gundersen HJ (2001). Tissue shrinkage and unbiased stereological estimation of particle number and size. J Microsc. 204, 232-246

Duan S, Anderson CM, Stein BA, Swanson RA (1999). Glutamate induces rapid upregulation of astrocyte glutamate transport and cell-surface expression of GLAST. J Neurosci. 19, 10193-10200

Eagleson KL, Haun F, Cunningham TJ (1990). Different populations of dorsal lateral geniculate nucleus neurons have concentration-specific requirements for a cortically derived neuron survival factor. Exp Neurol. 110, 284–290

Elleder M, Lake BD, Goebel HH, Rapola J, Haltia M, Carpenter S (1999). Definitions of the Ultrastructural Patterns found in NCL, pages 5-15; in The Neuronal Ceroid Lipofuscinoses (Batten Disease). Eds. Goebel HH, Mole SE, Lake BD. Publishers; McGraw-Hill Companies Inc. IOS Press

Elshatory Y, Brooks AI, Chattopadhyay S, Curran TM, Gupta P, Ramalingam V, Hofmann SL, Pearce DA. (2003). Early changes in gene expression in two models of Batten disease. FEBS Lett. 538, 207-212

Erickson AH, Blobel G (1983). Carboxyl-terminal proteolytic processing during biosynthesis of the lysosomal enzymes beta-glucuronidase and cathepsin D. Biochemistry. 22, 5201-5205

Eskelinen EL, Illert AL, Tanaka Y, Schwarzmann G, Blanz J, Von Figura K, Saftig P (2002). Role of LAMP-2 in lysosome biogenesis and autophagy. Mol Biol Cell. 13, 3355- 68

216 Eskelinen EL, Tanaka Y, Saftig P (2003). At the acidic edge: emerging functions for lysosomal membrane proteins. Trends Cell Biol. 13, 137-145

Ezaki J, Takeda-Ezaki M, Koike M, Ohsawa Y, Taka H, Mineki R, Murayama K, Uchiyama Y, Ueno T, Kominami E (2003). Characterization of Cln3p, the gene product responsible for juvenile neuronal ceroid lipofuscinosis, as a lysosomal integral membrane glycoprotein. J Neurochem. 87, 1296-1308

Ezaki J, Tanida I, Kanehagi N, Kominami E (1999). A lysosomal proteinase, the late infantile neuronal ceroid lipofuscinosis gene (CLN2) product, is essential for degradation of a hydrophobic protein, the subunit c of ATP synthase. J Neurochem. 72, 2573-2582

Flugel A, Hager G, Horvat A, Spitzer C, Singer GM, Graeber MB, Kreutzberg GW, Schwaiger FW (2001). Neuronal MCP-1 expression in response to remote nerve injury. J Cereb Blood Flow Metab. 21, 69-76

Freund TF, Buzsaki G (1996). Interneurons of the hippocampus. Hippocampus. 6, 347- 470

Futerman AH, van Meer G (2004). The cell biology of lysosomal storage disorders. Nat Rev Mol Cell Biol. 5, 554-565

Gahl WA, Bashan N, Tietze F, Bernardini I, Schulman JD (1982a). Cystine transport is defective in isolated leukocyte lysosomes from patients with cystinosis. Science. 217, 1263-1265

Gahl WA, Tietze F, Bashan N, Steinherz R, Schulman JD (1982b). Defective cystine exodus from isolated lysosome-rich fractions of cystinotic leucocytes. J Biol Chem. 257, 9570-9575

Gao H, Boustany RM, Espinola JA, Cotman SL, Srinidhi L, Antonellis KA, Gillis T, Qin X, Liu S, Donahue LR, Bronson RT, Faust JR, Stout D, Haines JL, Lerner TJ, MacDonald ME (2002). Mutations in a novel CLN6-encoded transmembrane protein cause variant neuronal ceroid lipofuscinosis in man and mouse. Am J Hum Genet. 70, 324-335

217 Gardiner RM (2002). Clinical features and molecular genetic basis of the neuronal ceroid lipofuscinoses. Adv Neurol. 89, 211-215

Gebicke-Haerter PJ (2001). Microglia in neurodegeneration: molecular aspects. Microsc Res Tech. 54, 47-58

Gehrmann J, Matsumoto Y, Kreutzberg GW (1995). Microglia: intrinsic immuneffector cell of the brain. Brain Res Brain Res Rev. 20, 269-287

Geuze HJ, Slot JW, Strous GJ, Hasilik A, Von Figura K (1984). Ultrastructural localization of the mannose 6-phosphate receptor in rat liver. J Cell Biol. 98, 2047-2054

Ghez C, Krakauer J (2000). The organization of movement. In: Kandel ER, Schwartz JH, Jessel TM (Eds.) Principles of neural science, 4th ed. McGraw-Hill, pp. 653-673

Glickman J, Croen K, Kelly S, Al-Awqati Q (1983). Golgi membranes contain an electrogenic H+ pump in parallel to a chloride conductance. J Cell Biol. 97, 1303-1308

Goebel HH (1992). Retina in various animal models of neuronal ceroid-lipofuscinosis. Am J Med Genet. 42, 605-608

Goebel HH (1995). The neuronal ceroid-lipofuscinoses. J Child Neurol. 10, 424-437

Goebel HH (2004). Symposium: the neuronal ceroid-lipofuscinoses (NCL)-a group of lysosomal diseases come of age. Brain Pathol. 14, 59-60

Goebel HH, Fix JD, Zeman W (1974). The fine structure of the retina in neuronal ceroid- lipofuscinosis. Am J Ophthalmol. 77, 25-39

Goebel HH, Schochet SS, Jaynes M, Bruck W, Kohlschutter A, Hentati F (1999). Progress in neuropathology of the neuronal ceroid lipofuscinoses. Mol Genet Metab. 66, 367-372

Goebel HH, Wisniewski KE (2004). Current state of clinical and morphological features in human NCL. Brain Pathol. 14, 61-69

218 Golabek AA, Kaczmarski W, Kida E, Kaczmarski A, Michalewski MP, Wisniewski KE (1999). Expression studies of CLN3 protein (battenin) in fusion with the green fluorescent protein in mammalian cells in vitro. Mol Genet Metab. 66, 277-282

Golabek AA, Kida E, Walus M, Kaczmarski W, Michalewski M, Wisniewski KE (2000). CLN3 protein regulates lysosomal pH and alters intracellular processing of Alzheimer's amyloid-beta protein precursor and cathepsin D in human cells. Mol Genet Metab. 70, 203-213

Golabek AA, Walus M, Wisniewski KE, Kida E (2005). Glycosaminoglycans modulate activation, activity, and stability of tripeptidyl-peptidase I in vitro and in vivo. J Biol Chem. 280, 7550-7561

Gotz M, Hartfuss E, Malatesta P (2002). Radial glial cells as neuronal precursors: a new perspective on the correlation of morphology and lineage restriction in the developing cerebral cortex of mice. Brain Res Bull. 57, 777-788

Graeber MB, Kreutzberg GW (1986). Astrocytes increase in glial fibrillary acidic protein during retrograde changes of facial motor neurons. J Neurocytol. 15, 363-373

Graeber MB, Kreutzberg GW (1988). Delayed astrocyte reaction following facial nerve axotomy. Journal of Neurocytology. 17, 209-220

Graeber MB, Moran LB (2002). Mechanisms of cell death in neurodegenerative diseases: fashion, fiction, and facts. Brain Pathol. 12, 385-390

Graeber MB, Streit WJ, Kreutzberg GW (1988a). Axotomy of the rat facial nerve leads to increased CR3 complement receptor expression by activated microglial cells. J Neurosci Res. 21, 18-24

Graeber MB, Tetzlaff W, Streit WJ, Kreutzberg GW (1988b). Microglial cells but not astrocytes undergo mitosis following rat facial nerve axotomy. Neuroscience Letters. 85, 317-321

219 Gravel RA, Kaback MM, Proia RL, Sandoff K, Suzuki K, Suzuki K (2001). The GM2 Gangliosidoses, pages 3827-3876; in The Metabolic Bases of Inherited Disease (8th edition), vol III. Eds. Scriver CR, Beaudet AL, Sly WS, Valle D. Publishers; McGraw- Hill Companies Inc.

Green SA, Zimmer KP, Griffiths G, Mellman I (1987). Kinetics of intracellular transport and sorting of lysosomal membrane and plasma membrane proteins. J Cell Biol. 105, 1227-1240

Griffey, M., Bible, E., Vogler, C., Levy, B., Gupta, P., Cooper, J.D., Sands, M (2003). Adeno-associated virus 2 -mediated gene therapy decreases autofluorescent storage material and increases brain mass in a murine model of Infantile Neuronal Ceroid Lipofuscinosis (INCL). Neurobiol Dis. 16, 360-369

Griffin JL, Muller D, Woograsingh R, Jowatt V, Hindmarsh A, Nicholson JK, Martin JE (2002). Vitamin E deficiency and metabolic deficits in neuronal ceroid lipofuscinosis described by bioinformatics. Physiol Genomics. 11, 195-203

Griffiths G, Gruenberg J (1991). The arguments for pre-existing early and late endosomes. Trends Cell Biol. 1, 5-9

Griffiths G, Hoflack B, Simons K, Mellman I, Kornfeld S (1988). The mannose 6- phosphate receptor and the biogenesis of lysosomes. Cell. 52, 329-341

Grube D (2004). Constants and variables in immunohistochemistry. Arch. Histology and Cytology. 67, 115-134

Guillery RW (1997). Quantification without pontification: choosing a method for counting objects in sectioned tissues. J Comp Neurol. 386, 2-7

Gundersen HJ (1986). Stereology of arbitrary particles. A review of unbiased number and size estimators and the presentation of some new ones, in memory of William R. Thompson. J Microsc. 143, 3-45

Gundersen HJ (1988). The nucleator. J Microsc. 151, 3-21

220 Gundersen HJ, Jensen EB (1985). Stereological estimation of the volume-weighted mean volume of arbitrary particles observed on random sections. J Microsc. 138, 127-142

Gundersen HJ, Jensen EB (1987). The efficiency of systematic sampling in stereology and its prediction. J Microsc. 147, 229-263

Guntinas-Lichius O, Neiss WF, Gunkel A, Stennert E (1994). Differences in glial, synaptic and motoneuron responses in the facial nucleus of the rat brainstem following facial nerve resection and nerve suture reanastomosis. Eur Arch Otorhinolaryngol. 251, 410-417

Gupta P, Soyombo AA, Atashband A, Wisniewski KE, Shelton JM, Richardson JA, Hammer RE, Hofmann SL (2001). Disruption of PPT1 or PPT2 causes neuronal ceroid lipofuscinosis in knockout mice. Proc Natl Acad Sci U S A. 98, 13566-13571

Haapanen A, AboRamadan U, von Schantz C, Fuhrmann J, Jentsch TJ, Braulke T, Saftig P, Jalanko A, Autti T, Tyynelä J (2005). Magnet resonance imaging of NCL- and NCL related mouse models. Abstract. NCL 2005 O33

Hall NA, Lake BD, Palmer DN, Jolly RD, Patrick AD (1989). Glycoconjugates in storage cytosomes from ceroid-lipofuscinosis (Batten's disease) and in lipofuscin from old-age brain. Adv Exp Med Biol. 266, 225-241

Hallermayer K, Harmening C, Hamprecht B (1981). Cellular localization and regulation of glutamine synthetase in primary cultures of brain cells from newborn mice. J Neurochem. 37, 43-52

Haltia M (2003). The neuronal ceroid-lipofuscinoses. J Neuropathol Exp Neurol. 62, 1-13

Haltia M, Herva R, Suopanki J, Baumann M, Tyynelä J (2001). Hippocampal lesions in the neuronal ceroid lipofuscinoses. Eur J Paediatr Neurol. 5, 209–211

Haltia M, Rapola J, Santavuori P (1973a). Infantile type of so-called neuronal ceroid- lipofuscinosis. Histological and electron microscopic studies. Acta Neuropathol (Berl). 26, 157–170

221 Haltia M, Rapola J, Santavuori P, Keranen A (1973b). Infantile type of so-called neuronal ceroid-lipofuscinosis. 2. Morphological and biochemical studies. J Neurol Sci. 18, 269- 285

Hansen SN, Alexander NA, von Schantz C, Kopra O, Jalanko A, Cooper JD (2005). Neuropathological and reactive changes in aged Cln5 deficient mice. Abstract. NCL 2005 P9

Harikumar P, Reeves JP (1983). The lysosomal proton pump is electrogenic. J Biol Chem. 258, 10403-10410

Harper PA, Walker KH, Healy PJ, Hartley WJ, Gibson AJ, Smith JS (1988). Neurovisceral ceroid-lipofuscinosis in blind Devon cattle. Acta Neuropathol (Berl). 75, 632-636

Harrison JK, Jiang Y, Chen S, Xia Y, Maciejewski D, McNamara RK, Streit WJ, Salafranca MN, Adhikari S, Thompson DA, Botti P, Bacon KB, Feng L (1998). Role for neuronally derived fractalkine in mediating interactions between neurons and CX3CR1-

Harter C, Mellman I (1992). Transport of the lysosomal membrane glycoprotein lgp120 (lgp-A) to lysosomes does not require appearance on the plasma membrane. J Cell Biol. 117, 311-325

Hasilik A, Neufeld EF (1980). Biosynthesis of lysosomal enzymes in fibroblasts. Phosphorylation of mannose residues. J Biol Chem. 255, 4946-4950

Haskell RE, Carr CJ, Pearce DA, Bennett MJ, Davidson BL (2000). Batten disease: evaluation of CLN3 mutations on protein localization and function. Hum Mol Genet. 9, 735-744

Haskell RE, Derksen TA, Davidson BL (1999). Intracellular trafficking of the JNCL protein CLN3. Mol Genet Metab. 66, 253-260

222 Hatten ME, Mason CA, Liem RK, Edmondson JC, Bovolenta P, Shelanski ML (1984). Neuron-astroglial interactions in vitro and their implications for repair of CNS injury. Cent Nerv Syst Trauma. 1, 15-27

Haydon PG (2001). GLIA: listening and talking to the synapse. Nat Rev Neurosci. 2, 185- 193

Heinonen O, Salonen T, Jalanko A, Peltonen L, Copp A (2000). CLN-1 and CLN-5, genes for infantile and variant late infantile neuronal ceroid lipofuscinoses, are expressed in the embryonic human brain. J Comp Neurol. 426, 406-412

Hellsten E, Vesa J, Jalanko A, Peltonen L (1997). From locus to cellular disturbances: positional cloning of the infantile neuronal ceroid lipofuscinosis gene. Neuropediatrics. 28, 9-11

Heppner FL, Skutella T, Hailer NP, Haas D, Nitsch R (1998). Activated microglial cells migrate towards sites of excitotoxic neuronal injury inside organotypic hippocampal slice cultures. Eur J Neurosci. 10, 3284-3290

Herva R, Tyynela J, Hirvasniemi A, Syrjakallio-Ylitalo M, Haltia M (2000). Northern epilepsy: a novel form of neuronal ceroid-lipofuscinosis. Brain Pathol. 10, 215-222

Hirt UA, Gantner F, Leist M (2000). Phagocytosis of nonapoptotic cells dying by caspase-independent mechanisms. J Immunol. 64, 6520-6529

Hof PR, Young WG, Bloom FE, Belichenko PV, Celio MR (2000). Comparative Cytoarcchitectonic Atlas of the C57BL/6 and 129/Sv Mouse Brains. Elsevier, Amsterdam

Hoflack B, Kornfeld S (1985). Purification and characterization of a cation-dependent mannose 6-phosphate receptor from murine P388D1 macrophages and bovine liver. J Biol Chem. 260, 12008-12014

Hofmann SL, Peltonen L (2001). The neuronal ceroid lipofuscinosis, pages 3877-3894; in The metabolic & molecular basis of inherited disease (8th edition), vol III. Eds. Scriver CR, Beaudet AL, Sly WS, Valle D. Publishers; McGraw-Hill Companies Inc.

223 Holmberg V, Jalanko A, Isosomppi J, Fabritius AL, Peltonen L, Kopra O (2004). The mouse ortholog of the neuronal ceroid lipofuscinosis CLN5 gene encodes a soluble lysosomal glycoprotein expressed in the developing brain. Neurobiol Dis. 16, 29-40

Holopainen JM, Saarikoski J, Kinnunen PK, Jarvela I (2001). Elevated lysosomal pH in neuronal ceroid lipofuscinoses (NCLs). Eur J Biochem. 268, 5851-5856

Homanics GE, Quinlan JL, Firestone LL (1999). Pharmacologic and behavioral responses of inbred C57BL/6J and strain 129/SvJ mouse lines. Pharmacol Biochem Behav. 63, 21– 26

Houser CR, Barber RP, Vaughn JE (1984). Immunocytochemical localization of glutamic acid decarboxylase in the dorsal lateral vestibular nucleus: evidence for an intrinsic and extrinsic GABAergic innervation. Neurosci Lett. 47, 213-220

Howard CV, Reed MG (1998a). 4 Estimation of component volume and volume fraction. In: Howard CV, Reed MG (Eds.) Unbiased Stereology: Three-Dimentional Measurement in Microscopy. Bios Scientific Publishers, pp 55-68

Howard CV, Reed MG (1998b). 5 Number estimation. In: Howard CV, Reed MG (Eds.) Unbiased Stereology: Three-Dimentional Measurement in Microscopy. Bios Scientific Publishers, pp 69-106

Howard CV, Reed MG (1998c). 9 Partical sizing. In: Howard CV, Reed MG (Eds.) Unbiased Stereology: Three-Dimentional Measurement in Microscopy. Bios Scientific Publishers, pp 139-150

Hunter KE, Hatten ME. Radial glial cell transformation to astrocytes is bidirectional: regulation by a diffusible factor in embryonic forebrain (1995). Proc Natl Acad Sci U S A. 92, 2061-2065

Hurley SD, Coleman PD (2003). Facial nerve axotomy in aged and young adult rats: analysis of the glial response. Neurobiol Aging. 24, 511-518

224 International Batten Disease Consortium (1995). Isolation of a novel gene underlying Batten disease, CLN3. Cell. 82, 949-957

Isosomppi J, Vesa J, Jalanko A, Peltonen L (2002). Lysosomal localization of the neuronal ceroid lipofuscinosis CLN5 protein. Hum Mol Genet. 11, 885-891

Jalanko A, Vesa J, Manninen T, von Schantz C, Minye H, Fabritius AL, Salonen T, Rapola J, Gentile M, Kopra O, Peltonen L (2005). Mice with Ppt1(Deltaex4) mutation replicate the INCL phenotype and show an inflammation-associated loss of interneurons. Neurobiol Dis. 18, 226-241

Janes RW, Munroe PB, Mitchison HM, Gardiner RM, Mole SE, Wallace BA (1996). A model for Batten disease protein CLN3: functional implications from homology and mutations. FEBS Lett. 399, 75-77

Jarvela I, Lehtovirta M, Tikkanen R, Kyttala A, Jalanko A (1999). Defective intracellular transport of CLN3 is the molecular basis of Batten disease (JNCL). Hum Mol Genet. 8, 1091-1098

Jarvela I, Sainio M, Rantamaki T, Olkkonen VM, Carpen O, Peltonen L, Jalanko A (1998). Biosynthesis and intracellular targeting of the CLN3 protein defective in Batten disease. Hum Mol Genet. 7, 85-90

Jessell MT, Sanes JR (2000). The development of the nervous system. In: Kandel ER, Schwartz JH, Jessel TM (Eds.) Principles of neural science, 4th ed. McGraw-Hill, pp. 1019-1039

Jolly RD, West DM (1976). Blindness in South Hampshire sheep: a neuronal ceroidlipofuscinosis. N Z Vet J. 24, 123

Kaczmarski W, Wisniewski KE, Golabek A, Kaczmarski A, Kida E, Michalewski M (1999). Studies of membrane association of CLN3 protein. Mol Genet Metab. 66, 261- 264

225 Kalla R, Liu Z, Xu S, Koppius A, Imai Y, Kloss CU, Kohsaka S, Gschwendtner A, Moller JC, Werner A, Raivich G (2001). Microglia and the early phase of immune surveillance in the axotomized facial motor nucleus: impaired microglial activation and lymphocyte recruitment but no effect on neuronal survival or axonal regeneration in macrophage-colony stimulating factor-deficient mice. J Comp Neurol. 436,182-201

Kandel ER (1991) Nerve cell and behaviour. In: Kandel ER, Schwartz JH, Jessel TM (Eds.) Principles of neural science. New York: Elsevier Science Publishing, pp 18-32

Katay L, Latzkovits L, Fonagy A, Janka Z, Lajtha A (1998). Effects of arginine vasopressin and atriopeptin on chloride uptake in cultured astroglia. Neurochem Res. 23, 831-836

Katz ML, Gao CL, Prabhakaram M, Shibuya H, Liu PC, Johnson GS (1997). Immunochemical localization of the Batten disease (CLN3) protein in retina. Invest Ophthalmol Vis Sci. 38, 2375-2386

Katz ML, Johnson GS (2001). Mouse gene knockout models for the CLN2 and CLN3 forms of ceroid lipofuscinosis. Eur J Paediatr Neurol. 5, 109-114

Katz ML, Khan S, Awano T, Shahid SA, Siakotos AN, Johnson GS (2005). A mutation in the CLN8 gene in English Setter dogs with neuronal ceroid-lipofuscinosis. Biochem Biophys Res Commun. 327, 541-547

Katz ML, Shibuya H, Liu PC, Kaur S, Gao CL, Johnson GS (1999). A mouse gene knockout model for juvenile ceroid-lipofuscinosis (Batten disease). J Neurosci Res. 57, 551-556

Kay GW, Palmer DN, Rezaie P, Cooper JD (2005). Prenatal activation of non-neuronal cells within the CNS of ovine neuronal ceroid lipofuscinosis (CLN6). In Press

Kida E, Golabek AA, Walus M, Wujek P, Kaczmarski W, Wisniewski KE (2001). Distribution of tripeptidyl peptidase I in human tissues under normal and pathological conditions. J Neuropathol Exp Neurol. 60, 280-292

226 Kida E, Kaczmarski W, Golabek AA, Kaczmarski A, Michalewski M, Wisniewski KE (1999). Analysis of intracellular distribution and trafficking of the CLN3 protein in fusion with the green fluorescent protein in vitro. Mol Genet Metab. 66, 265-271

Kim JJ, Diamond DM (2002). The stressed hippocampus, synaptic plasticity and lost memories. Nat Rev Neurosci. 3, 453-462

Kim Y, Ramirez-Montealegre D, Pearce DA (2003). A role in vacuolar arginine transport for yeast Btn1p and for human CLN3, the protein defective in Batten disease. Proc Natl Acad Sci U S A. 100, 15458-15462

Kincaid AE, Zheng T, Wilson CJ (1998). Connectivity and convergence of single corticostriatal axons. J Neurosci. 18, 4722-4731

Kirkwood BR. And Sterne AC (2003). Comparisons of means from several groups: analysis of variance, pages 80-86; in essential Medical Statistics (2nd edition). Eds. Goodgame F and Moore K. Publishers; Blackwell Science Ltd.

Klein MA, Moller JC, Jones LL, Bluethmann H, Kreutzberg GW, Raivich G (1997). Impaired neuroglial activation in interleukin-6 deficient mice. Glia. 19, 227-233

Kohlschutter A, Laabs R, Albani M (1988). Juvenile neuronal ceroid lipofuscinosis (JNCL): quantitative description of its clinical variability. Acta Paediatr Scand. 77, 867- 872

Koike M, Shibata M, Ohsawa Y, Nakanishi H, Koga T, Kametaka S, Waguri S, Momoi T, Kominami E, Peters C, Figura K, Saftig P, Uchiyama Y (2003). Involvement of two different cell death pathways in retinal atrophy of cathepsin D-deficient mice. Mol Cell Neurosci. 22, 146-161

Koppang N (1988). The English setter with ceroid-lipofuscinosis: a suitable model for the juvenile type of ceroid-lipofuscinosis in humans. Am J Med Genet. 5, 117-125

Kopra O, Vesa J., von Schantz C, Manninen T, Minye H, Fabritius AL, Rapola J, van Diggelen OP, Saarela J, Jalanko A, Peltonen L (2004).A mouse model for Finnish variant

227 late infantile neuronal ceroid lipofuscinosis, CLN5, reveals neuropathology associated with early aging. Hum Mol Gen. 13, 2893–2906

Kornfeld R, Bao M, Brewer K, Noll C, Canfield W (1999). Molecular cloning and functional expression of two splice forms of human N-acetylglucosamine-1- phosphodiester alpha-N-acetylglucosaminidase. J Biol Chem. 274, 32778-32785

Kornfeld S, Sly WS (2001). I-Cell Disease and Pseudo-Hurler Polydystrophy: Disorders of Lysosomal Enzyme Phosphorlation and Localization, pages 3469-3483; in The Metabolic Bases of Inherited Disease (8th edition), vol III. Eds. Scriver CR, Beaudet AL, Sly WS, Valle D. Publishers; McGraw-Hill Companies Inc.

Koutmani Y, Hurel C, Patsavoudi E, Hack M, Gotz M, Thomaidou D, Matsas R (2004). BM88 is an early marker of proliferating precursor cells that will differentiate into the neuronal lineage. Eur J Neurosci. 20, 2509-2523

Koutsilieri E, Scheller C, Tribl F, Riederer P (2002). Degeneration of neuronal cells due to oxidative stress-microglial contribution. Parkinsonism Relat Disord. 8, 401-406

Koutsilieri E, Sopper S, Heinemann T, Scheller C, Lan J, Stahl-Hennig C, ter Meulen V, Riederer P, Gerlach M (1999). Involvement of microglia in cerebrospinal fluid glutamate increase in SIV-infected rhesus monkeys (Macaca mulatta). AIDS Res Hum Retroviruses. 15, 471-477

Kracun I, Rosner H, Cosovic C, Stavljenic A (1984). Topographical atlas of the gangliosides of the adult human brain. J Neurochem. 43, 979-989

Kraehenbuhl JP, Jamieson JD (1974). Localization of intracellular antigens by immunoelectron microscopy. Int Rev Exp Pathol. 13, 1-53

Kramer H, Phistry M (1996). Mutations in the Drosophila hook gene inhibit endocytosis of the boss transmembrane ligand into multivesicular bodies. J Cell Biol. 133, 1205-1215

Kreutzberg GW (1995). Reaction of the neuronal cell body to axonal damage. In: Waxman SG, Kocsis JD, Stys PK (Eds.) The Axon. Oxford Univ, Press, pp. 355–374

228 Kreutzberg GW. Microglia: a sensor for pathological events in the CNS (1996). Trends Neurosci. 19, 312-318

Kriscenski-Perry E, Applegate CD, Serour A, Mhyre TR, Leonardo CC, Pearce DA (2002). Altered flurothyl seizure induction latency, phenotype, and subsequent mortality in a mouse model of Juvenile Neuronal Ceroid Lipofuscinosis/Batten Disease. Epilepsia 43, 1137-1140

Kurachi Y, Oka A, Itoh M, Mizuguchi M, Hayashi M, Takashima S (2001). Distribution and development of CLN2 protein, the late-infantile neuronal ceroid lipofuscinosis gene product. Acta Neuropathol (Berl). 102, 20-26

Kwon JM, Marshall FJ, deBlieck EA, Mink JW, Dure LS, Adams H, Messing S, Rothberg PG, Levy E, McDonough T, DeYoung J, Wang M, Pearce DA (2005). Unified Batten Disease Rating Scale: Description, Reliability, and Use in Monitoring Disease Progression. Abstract. NCL 2005 O47

Kyttala A, Ihrke G, Vesa J, Schell MJ, Luzio JP (2004). Two motifs target Batten disease protein CLN3 to lysosomes in transfected nonneuronal and neuronal cells. Mol Biol Cell. 15, 1313-1323

Kyttala A, Yliannala K, Schu P, Jalanko A, Luzio JP (2005). Batten disease: evaluation of CLN3 mutations on protein localization and function. J Biol Chem. 280, 10277-10283

Lane SC, Jolly RD, Schmechel DE, Alroy J, Boustany RM (1996). Apoptosis as the mechanism of neurodegeneration in Batten's disease. J Neurochem. 67, 677-683

Lashley KS (1941). Thalamocortical connection in the rat’s brain. J Comp Neurol. 75, 67–121

Laskawi R, Wolff JR (1996). Changes in glial fibrillary acidic protein immunoreactivity in the rat facial nucleus following various types of nerve lesions. Eur Arch Otorhinolaryngol. 253, 475-480

229 Lau MM, Neufeld EF (1989). A frameshift mutation in a patient with Tay-Sachs disease causes premature termination and defective intracellular transport of the alpha-subunit of beta-hexosaminidase. J Biol Chem. 264, 21376-21380

Lauronen L, Heikkila E, Autti T, Sainio K, Huttunen J, Aronen HJ, Korvenoja A, Ilmoniemi RJ, Santavuori P (1997). Somatosensory evoked magnetic fields from primary sensorimotor cortex in juvenile neuronal ceroid lipofuscinosis. J Child Neurol. 12, 355– 360

Lauronen L, Munroe PB, Jarvela I, Autti T, Mitchison HM, O'Rawe AM, Gardiner RM, Mole SE, Puranen J, Hakkinen AM, Kirveskari E, Santavuori P (1999). Delayed classic and protracted phenotypes of compound heterozygous juvenile neuronal ceroid lipofuscinosis. Neurology 52, 360–365

Ledee RW, RK, Rappor, MM, Suzuki K (1984). Ganglioside Structure, Function and Biochemical Potential. New York, Plenum Publishing

Lee AW, Emsley JG, Brown RE, Hagg T (2003). Marked differences in olfactory sensitivity and apparent speed of forebrain neuroblast migration in three inbred strains of mice. Neuroscience. 118, 263–270

Lee RL, Johnson KR, Lerner TJ (1996). Isolation and chromosomal mapping of a mouse homolog of the Batten disease gene CLN3. Genomics. 35, 617-619

Lemansky P, Bishop DF, Desnick RJ, Hasilik A, von Figura K (1987). Synthesis and processing of alpha-galactosidase A in human fibroblasts. Evidence for different mutations in Fabry disease. J Biol Chem. 262, 2062-2065

Lim MJ, Alexander NA, Benedict JW, Chattophadhyay S, Shemilt SJA, Tyynelä J, Guérin CJ, Cooper JD, Pearce DA (2005a). IgG entry deposition within the CNS as components of the autoimmune response in Batten Disease. Abstract. NCL 2005 O31

Lim MJ, Beake J, Bible E, Curran TM, Ramirez-Montealegre D, Pearce DA, Cooper JD (2005b). Distinct patterns of serum immunoreactivity in Juvenile Neuronal Ceroid Lipofuscinosis. Abstract. NCL 2005 P20

230 Lin L, Sohar I, Lackland H, Lobel P (2001). The human CLN2 protein/tripeptidyl- peptidase I is a serine protease that autoactivates at acidic pH. J Biol Chem. 276, 2249- 2255

Lipp HP, Wolfer DP (2003). Genetical background problems in the analysis of cognitive and neuronal changes in genetically modified animals. Clin Neurosci Res. 3, 223–231

Liu B, Hong JS (2003). Role of microglia in inflammation-mediated neurodegenerative diseases: mechanisms and strategies for therapeutic intervention. J Pharmacol Exp Ther. 304, 1-7

Liu CG, Sleat DE, Donnelly RJ, Lobel P (1998). Structural organization and sequence of CLN2, the defective gene in classical late infantile neuronal ceroid lipofuscinosis. Genomics. 50, 206-212

Livingstone M, Hubel D (1988). Segregation of form, color, movement, and depth: anatomy, physiology, and perception. Science. 240, 740-749

Lloyd (1996). Subcellular Biochemistry. In: Lloyd JB, Mason RW (Eds.) Bilogy of the Lysosome, vol 27. Plenum Press, New York, pp 361-386

Loddick S, Rothwell N (2002). Cytokines in neurodegerative conditions. In: Loddick S, Rothwell N (Eds.) Immune and inflammatory responses in the nervous system, 2nd ed. Oxford Univ, Press, pp. 90–105

Lonka L, Kyttala A, Ranta S, Jalanko A, Lehesjoki AE (2000). The neuronal ceroid lipofuscinosis CLN8 membrane protein is a resident of the endoplasmic reticulum. Hum Mol Genet. 9, 1691-1697

Lonka L, Aalto A, Kopra O, Kuronen M, Kokaia Z, Saarma M, Lehesjoki AE (2005). The neuronal ceroid lipofuscinosis Cln8 gene expression is developmentally regulated in mouse brain and up-regulated in the hippocampal kindling model of epilepsy. BMC Neurosci. 6, 27-38

231 Lopez-Bendito G, Molnar Z (2003). Thalamocortical development: how are we going to get there? Nat Rev Neurosci. 4, 276-289

Luiro K, Kopra O, Lehtovirta M, Jalanko A (2001). CLN3 protein is targeted to neuronal synapses but excluded from synaptic vesicles: new clues to Batten disease. Hum Mol Genet. 10, 2123-2131

Luiro K, Yliannala K, Ahtiainen L, Maunu H, Jarvela I, Kyttala A, Jalanko A (2004). Interconnections of CLN3, Hook1 and Rab proteins link Batten disease to defects in the endocytic pathway. Hum Mol Genet. 13, 3017-3027

Maddox L, Bible E, Griffey M, Sands MS, Cooper JD (2005). Neuropathological and behavioural landmarks of disease progression in C57B1/6 congenic PPT1 null mutant mice. Abstract. NCL 2005 O30

Mao Q, Foster BJ, Xia H, Davidson BL (2003a). Membrane topology of CLN3, the protein underlying Batten disease. FEBS Lett. 541, 40-46

Mao Q, Xia H, Davidson BL (2003b). Intracellular trafficking of CLN3, the protein underlying the childhood neurodegenerative disease, Batten disease. FEBS Lett. 555, 351-357

March PA, Wurzelmann S, Walkley SU (1995). Morphological alterations in neocortical and cerebellar GABAergic neurons in a canine model of juvenile Batten disease. Am J Med Genet. 57, 204-212

Margraf LR, Boriack RL, Routheut AA, Cuppen I, Alhilali L, Bennett CJ, Bennett MJ (1999). Tissue expression and subcellular localization of CLN3, the Batten disease protein. Mol Genet Metab. 66, 283-289

Markram H, Toledo-Rodriguez M, Wang Y, Gupta A, Silberberg G, Wu C (2004). Interneurons of the neocortical inhibitory system. Nat Rev Neurosci. 5, 793-807

Markram H, Wang Y, Tsodyks M (1998). Differential signaling via the same axon of neocortical pyramidal neurons. Proc Natl Acad Sci U S A. 95, 5323-5328

232 Marsh M, Schmid S, Kern H, Harms E, Male P, Mellman I, Helenius A (1987). Rapid analytical and preparative isolation of functional endosomes by free flow electrophoresis. J Cell Biol. 104, 875-886

Martin JJ, Gottlob I, Goebel HH, Mole SE (1999). CLN4 Adult NCL, pages 77-90; in The Neuronal Ceroid Lipofuscinoses (Batten Disease). Eds. Goebel HH, Mole SE, Lake BD. Publishers; McGraw-Hill Companies Inc. IOS Press

Martinus RD, Harper PA, Jolly RD, Bayliss SL, Midwinter GG, Shaw GJ, Palmer DN (1991). Bovine ceroid-lipofuscinosis (Batten's disease): the major component stored is the DCCD-reactive proteolipid, subunit C, of mitochondrial ATP synthase. Vet Res Commun. 15, 85-94

Masliah E, Terry RD, DeTeresa RM, Hansen LA (1989). Immunohistochemical quantification of the synapse-related protein synaptophysin in Alzheimer disease. Neurosci Lett. 103, 234-239

Matthews MA (1973). Death of the central neuron: an electronmicroscopic study of thalamic retrograde degeneration following cortical ablation. J Neurocytol. 2, 265–288

McGeer PL, McGeer EG (1998). Glial cell reactions in neurodegenerative diseases: pathophysiology and therapeutic interventions. Alzheimer Dis Assoc Disord. 12, 1-6

Mehler MF (2002). Mechanisms regulating lineage diversity during mammalian cerebral cortical neurogenesis and gliogenesis. Results Probl Cell Differ. 39, 27-52

Mellman I, Fuchs R, Helenius A (1986). Acidification of the endocytic and exocytic pathways. Annu Rev Biochem. 55, 663-700

Melville SA, Wilson CL, Chaing CS, Studdert VP, Lingaas F, Wilton AN (2005). A Mutation in canine CLN5 causes Neuronal Ceroid Lipofuscinosis in Border Collie dogs. Abstract. NCL 2005 P24

Messer A, Flaherty L (1986). Autosomal dominance in a late-onset motor neuron disease in the mouse. J Neurogenet. 3, 345-355

233 Messer A, Manley K, Plummer JA. An early-onset congenic strain of the motor neuron degeneration (mnd) mouse (1999). Mol Genet Metab. 66, 393-397

Messer A, Plummer J, MacMillen MC, Frankel WN. Messer A, Plummer J, MacMillen MC, Frankel WN (1995). Genetics of primary and timing effects in the mnd mouse. Am J Med Genet. 57, 361-364

Messer A, Strominger NL, Mazurkiewicz JE (1987). Histopathology of the late-onset motor neuron degeneration (Mnd) mutant in the mouse. J Neurogenet. 4, 201-213

Michalski JC, Klein A (1999). Glycoprotein lysosomal storage disorders: alpha- and beta- mannosidosis, fucosidosis and alpha-N-acetylgalactosaminidase deficiency. Biochim Biophys Acta. 1455, 69-84

Michalewski MP, Kaczmarski W, Golabek AA, Kida E, Kaczmarski A, Wisniewski KE. 1999. Posttranslational modification of CLN3 protein and its possible functional implication. Mol Genet Metab 66:272–276.

Miller, K. (2000). Immunocytochemical Techniques. In: Bancroft JD, Gamble M (Eds.) Therory and practice of histological techniques, 5th ed. Churchill Livingstone, pp 421- 464

Mitchell WA, Wheeler RB, Sharp JD, Bate SL, Gardiner RM, Ranta US, Lonka L, Williams RE, Lehesjoki AE, Mole SE (2001). Turkish variant late infantile neuronal ceroid lipofuscinosis (CLN7) may be allelic to CLN8. Eur J Paediatr Neurol. 5, 21-27

Mitchison HM, Bernard DJ, Greene ND, Cooper JD, Junaid MA, Pullarkat RK, de Vos N, Breuning MH, Owens JW, Mobley WC, Gardiner RM, Lake BD, Taschner PE, Nussbaum RL (1999). Targeted disruption of the Cln3 gene provides a mouse model for Batten disease. Neurobiol Dis. 6, 321-334

Mitchison HM, Lim MJ, Cooper JD (2004). Selectivity and types of cell death in the neuronal ceroid lipofuscinoses. Brain Pathol. 14, 86-96

234 Mitchison HM, Taschner PE, Kremmidiotis G, Callen DF, Doggett NA, Lerner TJ, Janes RB, Wallace BA, Munroe PB, O'Rawe AM, Gardiner RM, Mole SE (1997). Structure of the CLN3 gene and predicted structure, location and function of CLN3 protein. Neuropediatrics. 28, 12-14

Mole SE (1998). Batten disease: four genes and still counting. Neurobiol Dis. 5, 287-303

Mole SE (2004). The genetic spectrum of human neuronal ceroid-lipofuscinoses. Brain Pathol. 14, 70-76

Mole SE, Michaux G, Codlin S, Wheeler RB, Sharp JD, Cutler DF (2004). CLN6, which is associated with a lysosomal storage disease, is an endoplasmic reticulum protein. Exp Cell Res. 298, 399-406

Mole SE, Mitchison HM, Munroe PB (1999). Molecular basis of the neuronal ceroid lipofuscinoses: mutations in CLN1, CLN2, CLN3, and CLN5. Hum Mutat. 14, 199-215

Möller JC, Klein MA, Haas S, Jones LL, Kreutzberg GW, Raivich G (1996). Regulation of thrombospondin in the regenerating mouse facial nucleus. Glia, 17, 121–132

Moolman DL, Vitolo OV, Vonsattel JP, Shelanski ML (2004). Dendrite and dendritic spine alterations in Alzheimer models. J Neurocytol. 33, 377-387

Moran LB, Graeber MB (2004). The facial nerve axotomy model. Brain Res Brain Res Rev. 44, 154-178

Mumby SM (1997). Reversible palmitoylation of signaling proteins. Curr Opin Cell Biol. 9, 148-154

Murphy NP, Lam HA, Maidment NT (2001). A comparison of morphine-induced locomotor activity and mesolimbic dopamine release in C57BL6, 129Sv and DBA2 mice. J Neurochem. 79, 626–635

Murphy RF (1991). Maturation models for endosome and lysosome biogenesis. Trends Cell Biol. 1, 77-82

235 Nakai K, Kanehisa M (1992). A knowledge base for predicting protein localization sites in eukaryotic cells. Genomics. 14, 897-911

Nakanishi H, Zhang J, Koike M, Nishioku T, Okamoto Y, Kominami E, von Figura K, Peters C, Yamamoto K, Saftig P, Uchiyama Y (2001). Involvement of nitric oxide released from microglia-macrophages in pathological changes of cathepsin D-deficient mice. J Neurosci. 21, 7526-7533

Neufeld EF, Lim TW, Shapiro LJ (1975). Inherited disorders of lysosomal metabolism. Annu Rev Biochem. 44, 357-376

Neufeld EF, Muenzer J (2001). The Mucopolysaccharidoses, pages 3421-3452; in The Metabolic Bases of Inherited Disease (8th edition), vol III. Eds. Scriver CR, Beaudet AL, Sly WS, Valle D. Publishers; McGraw-Hill Companies Inc.

Nielsen S, Nagelhus EA, Amiry-Moghaddam M, Bourque C, Agre P, Ottersen OP (1997). Specialized membrane domains for water transport in glial cells: high-resolution immunogold cytochemistry of aquaporin-4 in rat brain. J Neurosci. 17, 171-180

Nijssen PC, Ceuterick C, van Diggelen OP, Elleder M, Martin JJ, Teepen JL, Tyynela J, Roos RA (2003). Autosomal dominant adult neuronal ceroid lipofuscinosis: a novel form of NCL with granular osmiophilic deposits without palmitoyl protein thioesterase 1 deficiency. Brain Pathol. 13, 574-581

Nishino I, Fu J, Tanji K, Yamada T, Shimojo S, Koori T, Mora M, Riggs JE, Oh SJ, Koga Y, Sue CM, Yamamoto A, Murakami N, Shanske S, Byrne E, Bonilla E, Nonaka I, DiMauro S, Hirano M (2000). Primary LAMP-2 deficiency causes X-linked vacuolar cardiomyopathy and myopathy (Danon disease). Nature. 406, 906-910

Nolan CM, Creek KE, Grubb JH, Sly WS (1987). Antibody to the phosphomannosyl receptor inhibits recycling of receptor in fibroblasts. J Cell Biochem. 35, 137-151

Offen D, Elkon H, Melamed E (2000). Apoptosis as a general cell death pathway in neurodegenerative diseases. J Neural Transm. 58, 153-166

236 Ohkuma S, Moriyama Y, Takano T (1982). Identification and characterization of a proton pump on lysosomes by fluorescein-isothiocyanate-dextran fluorescence. Proc Natl Acad Sci U S A. 79, 2758-2762

Oppenheim RW (1989). The neurotrophic theory and naturally occurring motoneuron death. Trends Neurosci. 12, 252-255

Ordaz B, Vaca L, Franco R, Pasantes-Morales H (2004). Volume changes and whole cell membrane currents activated during gradual osmolarity decrease in C6 glioma cells: contribution of two types of K+ channels. Am J Physiol Cell Physiol. 286, 1399-1409

Østergaard JR, Egeblad H, Mølgaard H (2005). The evolution of cardiac involvement in juvenile neuronal ceroid lipofuscinosis. Abstract. NCL 2005. P28

Oswald MJ, Kay GW, Palmer DN (2001). Changes in GABAergic neuron distribution in situ and in neuron cultures in ovine (OCL6) Batten disease. Eur J Paediatr Neurol. 5, 135-142

Oswald MJ, Palmer DN, Kay GW, Shemilt SJA, Rezaie P, Cooper JD (2005). Glial activation spreads from specific cerebral foci and precedes neurodegeneration in presymptomatic ovine neuronal ceroid lipofuscinosis (CLN6). Neurobiol Dis. 20, 49-63

Palmer DN, Fearnley IM, Walker JE, Hall NA, Lake BD, Wolfe LS, Haltia M, Martinus RD, Jolly RD (1992). Mitochondrial ATP synthase subunit c storage in the ceroid- lipofuscinoses (Batten disease). Am J Med Genet. 42, 561-567

Pavlidis P, Noble WS (2001). Analysis of strain and regional variation in gene expression in mouse brain. Genome Biol. 2, RESEARCH0042

Paxinos G, Franklin KBJ (2001). The mouse brain in stereotaxic coordinates, Second Edition. Academic Press, San Diego

Pearce DA, Sherman F (1997). BTN1, a yeast gene corresponding to the human gene responsible for Batten's disease, is not essential for viability, mitochondrial function, or degradation of mitochondrial ATP synthase. Yeast. 13, 691-697

237 Pearce DA, Sherman F (1998). A yeast model for the study of Batten disease. Proc Natl Acad Sci U S A. 95, 6915-6918

Pekny M, Nilsson M. Astrocyte activation and reactive gliosis (2005). Glia. 50, 427-434

Persaud-Sawin DA, VanDongen A, Boustany RM (2002). Motifs within the CLN3 protein: modulation of cell growth rates and apoptosis. Hum Mol Genet. 11, 2129-4212

Petito CK, Morgello S, Felix JC, Lesser ML (1990). The two patterns of reactive astrogliosis in postischemic rat brain. J Cereb Blood Flow Metab. 10, 850−859

Petrie A and Sabin C (2000). Numerical data: more than two groups, pages 55-60; in Medical Statistics at a Glance. Eds. Petrie A and Sabin C. Publishers; Blackwell Science Ltd.

Phillips SN, Benedict JW, Weimer JM, Pearce DA (2005). CLN3, the Protein Associated with Batten Disease: Structure, Function and Localization. J Neuro Res. 79, 573-583

Piet R, Poulain DA, Oliet SH (2002). Modulation of synaptic transmission by astrocytes in the rat supraoptic nucleus. J Physiol Paris. 96, 231-236

Pineda-Trujillo N, Cornejo W, Carrizosa J, Wheeler RB, Munera S, Valencia A, Agudelo-Arango J, Cogollo A, Anderson G, Bedoya G, Mole SE, Ruiz-Linares A (2005). A CLN5 mutation causing an atypical neuronal ceroid lipofuscinosis of juvenile onset. Neurology. 64, 740-742

Pisoni RL, Flickinger KS, Thoene JG, Christensen HN (1987b). Characterization of carrier-mediated transport systems for small neutral amino acids in human fibroblast lysosomes. J Biol Chem. 262, 6010-6017

Pisoni RL, Thoene JG, Lemons RM, Christensen HN (1987a). Important differences in cationic amino acid transport by lysosomal system c and system y+ of the human fibroblast. J Biol Chem. 262, 15011-15018

238 Pohlmann R, Waheed A, Hasilik A, von Figura K (1982). Synthesis of phosphorylated recognition marker in lysosomal enzymes is located in the cis part of Golgi apparatus. J Biol Chem. 257, 5323-5325

Pompl PN, Yemul S, Xiang Z, Ho L, Haroutunian V, Purohit D, Mohs R, Pasinetti GM (2003). Caspase gene expression in the brain as a function of the clinical progression of Alzheimer disease. Arch Neurol. 60, 369-376

Pontikis CC, Cella CV, Parihar N, Lim MJ, Chakrabarti S, Mitchison HM, Mobley WC, Rezaie P, Pearce DA, Cooper JD (2004). Late onset neurodegeneration in the Cln3-/- mouse model of juvenile neuronal ceroid lipofuscinosis is preceded by low level glial activation. Brain Res. 1023, 231-242

Pontikis CC, Cotman SL, Macdonald ME, Cooper JD (2005). Thalamocortical neuron loss and localized astrocytosis in the Cln3 (Deltaex7/8) knock-in mouse model of Batten disease. Neurobiol Dis. [Epub ahead of print]

Pullarkat RK, Morris GN (1997). Farnesylation of Batten disease CLN3 protein. Neuropediatrics. 28, 42-44

Purpura DP, Suzuki K (1976). Distortion of neuronal geometry and formation of aberrant synapses in neuronal storage disease. Brain Res. 116, 1-21

Rabchevsky AG, Streit WJ (1997). Grafting of cultured microglial cells into the lesioned spinal cord of adult rats enhances neurite outgrowth. J Neurosci Res. 47, 34-48

Raininko R, Santavuori P, Heiskala H, Sainio K, Palo J (1990). CT findings in neuronal ceroid lipofuscinoses. Neuropediatrics. 21, 95-101

Raivich G, Bohatschek M, Kloss CU, Werner A, Jones LL, Kreutzberg GW (1999). Neuroglial activation repertoire in the injured brain: graded response, molecular mechanisms and cues to physiological function. Brain Res Rev. 30, 77-105

239 Raivich G, Gehrmann J, Kreutzberg GW (1991). Increase of macrophage colony- stimulating factor and granulocyte–macrophage colony stimulating factor receptors in the regenerating rat facial nucleus. J. Neurosci. Res. 30, 682–686

Raivich G, Haas S, Werner A, Klein MA, Kloss CUA, Kreutzberg GW (1998). Regulation of MCSF receptors on microglia in the normal and injured mouse central nervous system: a quantitative immunofluorescence study using confocal laser microscopy. J Comp Neurol. 395, 342–358

Raivich G, Reddington M, Haas CA, Kreutzberg GW (1995). Peptides in motoneurons. Prog Brain Res. 104, 3–20

Ranta S, Lehesjoki AE (2000). Northern epilepsy, a new member of the NCL family. Neurol Sci. 21, 43-47

Ranta S, Savukoski M, Santavuori P, Haltia M (2001). Studies of homogeneous populations: CLN3 and CLN8. In: Wisniewski KE, Zhong N (Eds.) Batten Disease: Diagnosis, Treatment and Research. Academic Press: San Diego, pp. 123-140

Ranta S, Zhang Y, Ross B, Lonka L, Takkunen E, Messer A, Sharp J, Wheeler R, Kusumi K, Mole S, Liu W, Soares MB, Bonaldo MF, Hirvasniemi A, de la Chapelle A, Gilliam TC, Lehesjoki AE (1999). The neuronal ceroid lipofuscinoses in human EPMR and mnd mutant mice are associated with mutations in CLN8. Nat Genet. 23, 233-236

Rees-Jones R, Al-Awqati Q (1984). Proton-translocating adenosinetriphosphatase in rough and smooth microsomes from rat liver. Biochemistry. 23, 2236-2240

Reijngoud DJ, Tager JM (1977). The permeability properties of the lysosomal membrane. Biochim Biophys Acta. 472, 419-449

Reitman ML, Varki A, Kornfeld S (1981). Fibroblasts from patients with I-cell disease and pseudo-Hurler polydystrophy are deficient in uridine 5'-diphosphate-N- acetylglucosamine: glycoprotein N-acetylglucosaminylphosphotransferase activity. J Clin Invest. 67, 1574-1579

240 Rinne JO, Ruottinen HM, Nagren K, Aberg LE, Santavuori P (2002). Positron emission tomography shows reduced striatal dopamine D1 but not D2 receptors in juvenile neuronal ceroid lipofuscinosis. Neuropediatrics. 33, 138-141

Robinson MB, Dowd LA (1997). Heterogeneity and functional properties of subtypes of sodium-dependent glutamate transporters in the mammalian central nervous system. Adv Pharmacol. 37, 69-115

Rodriguez-Tebar A, Jeffrey PL, Thoenen H, Barde YA (1989). The survival of chick retinal ganglion cells in response to brain derived neurotrophic factor depends on their embryonic age. Dev Biol. 136, 296-303

Rosenfeld MG, Kreibich G, Popov D, Kato K, Sabatini DD (1982). Biosynthesis of lysosomal hydrolases: their synthesis in bound polysomes and the role of co- and post- translational processing in determining their subcellular distribution. J Cell Biol. 93, 135- 143

Rothstein JD, Dykes-Hoberg M, Pardo CA, Bristol LA, Jin L, Kuncl RW, Kanai Y, Hediger MA, Wang Y, Schielke JP, Welty DF (1996). Knockout of glutamate transporters reveals a major role for astroglial transport in excitotoxicity and clearance of glutamate. Neuron. 16, 675-686

Rubenstein JL, Beachy PA (1998). Patterning of the embryonic forebrain. Curr Opin Neurobiol. 8, 18-26

Ruottinen HM, Rinne JO, Haaparanta M, Solin O, Bergman J, Oikonen VJ, Jarvela I, Santavuori P (1997). [18F]fluorodopa PET shows striatal dopaminergic dysfunction in juvenile neuronal ceroid lipofuscinosis. J Neurol Neurosurg Psychiatry. 62, 622-625

Sabatini DD, Adesnik MB (2001). The Biogenesis of Membranes and Organelles, pages 475-484; in The Metabolic Bases of Inherited Disease (8th edition), vol III. Eds. Scriver CR, Beaudet AL, Sly WS, Valle D. Publishers; McGraw-Hill Companies Inc.

241 Sahagian GG, Distler J, Jourdian GW (1981). Characterization of a membrane-associated receptor from bovine liver that binds phosphomannosyl residues of bovine testicular beta- galactosidase. Proc Natl Acad Sci U S A. 78, 4289-4293

Sandberg R, Yasuda R, Pankratz DG, Carter TA, Del Rio JA, Wodicka L, Mayford M, Lockhart DJ, Barlow C (2000). Regional and strain-specific gene expression mapping in the adult mouse brain. Proc Natl Acad Sci U S A. 97, 11038–11043

Sandhoff K, Andreae U, Jatzkewitz H (1968). Deficient hexozaminidase activity in an exceptional case of Tay-Sachs disease with additional storage of kidney globoside in visceral organs. Life Sci. 7, 283-288

Sandhoff K, Harzer K, Wassle W, Jatzkewitz H (1971). Enzyme alterations and lipid storage in three variants of Tay-Sachs disease. J Neurochem. 18, 2469-2489

Savukoski M, Kestila M, Williams R, Jarvela I, Sharp J, Harris J, Santavuori P, Gardiner M, Peltonen L (1994). Defined chromosomal assignment of CLN5 demonstrates that at least four genetic loci are involved in the pathogenesis of human ceroid lipofuscinoses. Am J Hum Genet. 55, 695-701

Savukoski M, Klockars T, Holmberg V, Santavuori P, Lander ES, Peltonen L (1998). CLN5, a novel gene encoding a putative transmembrane protein mutated in Finnish variant late infantile neuronal ceroid lipofuscinosis. Nat Genet. 19, 286-288

Schiefer J, Kampe K, Dodt HU, Zieglgansberger W, Kreutzberg GW (1999). Microglial motility in the rat facial nucleus following peripheral axotomy. J Neurocytol. 28, 439-453

Schmid SL, Fuchs R, Male P, Mellman I (1988). Two distinct subpopulations of endosomes involved in membrane recycling and transport to lysosomes. Cell. 52, 73-83

Schmitz C, Hof PR (2005). Design-based stereology in neuroscience. Neuroscience. 130, 813-831

242 Schriner JE, Yi W, Hofmann SL (1996). cDNA and genomic cloning of human palmitoyl-protein thioesterase (PPT), the enzyme defective in infantile neuronal ceroid lipofuscinosis. Genomics. 34, 317-322

Schulz A, Dhar S, Rylova S, Dbaibo G, Alroy J, Hagel C, Artacho I, Kohlschutter A, Lin S, Boustany RM (2004). Impaired cell adhesion and apoptosis in a novel CLN9 Batten disease variant. Ann Neurol. 56, 342-350

Seigel GM, Lotery A, Kummer A, Bernard DJ, Greene ND, Turmaine M, Derksen T, Nussbaum RL, Davidson B, Wagner J, Mitchison HM (2002). Retinal pathology and function in a Cln3 knockout mouse model of juvenile Neuronal Ceroid Lipofuscinosis (batten disease). Mol Cell Neurosci. 19, 515-527

Sharp JD, Wheeler RB, Parker KA, Gardiner RM, Williams RE, Mole SE (2003). Spectrum of CLN6 mutations in variant late infantile neuronal ceroid lipofuscinosis. Hum Mutat. 22, 35-42

Simard M, Arcuino G, Takano T, Liu QS, Nedergaard M (2003). Signaling at the gliovascular interface. J Neurosci. 23, 9254-9262

Simard M, Nedergaard M (2004). The neurobiology of glia in the context of water and ion homeostasis. Neuroscience. 129, 877-96

Simpson EM, Linder CC, Sargent EE, Davisson MT, Mobraaten LE, Sharp JJ (1997). Genetic variation among 129 substrains and its importance for targeted mutagenesis in mice. Nat Genet. 16, 19–27

Sincich LC, Horton JC (2004). The Circuitry of V1 and V2: Integration of Color, Form, and Motion. Annu Rev Neurosci. [Epub ahead of print]

Sleat DE, Donnelly RJ, Lackland H, Liu CG, Sohar I, Pullarkat RK, Lobel P (1997). Association of mutations in a lysosomal protein with classical late-infantile neuronal ceroid lipofuscinosis. Science. 277, 1802-1805

243 Sleat DE, Wiseman JA, El-Banna M, Kim KH, Mao Q, Price S, Macauley SL, Sidman RL, Shen MM, Zhao Q, Passini MA, Davidson BL, Stewart GR, Lobel P (2004). A mouse model of classical late-infantile neuronal ceroid lipofuscinosis based on targeted disruption of the CLN2 gene results in a loss of tripeptidyl-peptidase I activity and progressive neurodegeneration. J Neurosci. 24, 9117-9126

Sly WS, Fischer HD (1982). The phosphomannosyl recognition system for intracellular and intercellular transport of lysosomal enzymes. J Cell Biochem. 18, 67-85

Sohar I, Sleat DE, Jadot M, Lobel P (1999). Biochemical characterization of a lysosomal protease deficient in classical late infantile neuronal ceroid lipofuscinosis (LINCL) and development of an enzyme-based assay for diagnosis and exclusion of LINCL in human specimens and animal models. J Neurochem. 73, 700-711

Solovyev VV, Salamov AA (1994). Predicting alpha-helix and beta-strand segments of globular proteins. Comput Appl Biosci. 10, 661-669

Stein M, Zijderhand-Bleekemolen JE, Geuze H, Hasilik A, von Figura K (1987). Mr 46,000 mannose 6-phosphate specific receptor: its role in targeting of lysosomal enzymes. EMBO J. 6, 2677-2681

Stengel C (1826). Beretning om et maerkeligt Sygnomstilfaelde hos fire Socdskende I Naerheden af Roeraas. Eyr et medicinsk Tidsrift. 1, 347-352

Streit WJ, Kreutzberg GW (1988). Response of endogenous glial cells to motor neuron degeneration induced by toxic ricin. J Comp Neurol. 268, 248–263

Sugie K, Yamamoto A, Murayama K, Oh SJ, Takahashi M, Mora M, Riggs JE, Colomer J, Iturriaga C, Meloni A, Lamperti C, Saitoh S, Byrne E, DiMauro S, Nonaka I, Hirano M, Nishino I (2002). Clinicopathological features of genetically confirmed Danon disease. Neurology. 58, 1773-1778

Suopanki J, Tyynela J, Baumann M, Haltia M (1999). Palmitoyl-protein thioesterase, an enzyme implicated in neurodegeneration, is localized in neurons and is developmentally regulated in rat brain. Neurosci Lett. 265, 53-56

244 Sur M, Leamey CA (2001). Development and plasticity of cortical areas and networks. Nat Rev Neurosci. 2, 251-262

Suzuki R, Watanabe J, Arata S, Funahashi H, Kikuyama S, Shioda S (2003). A transgenic mouse model for the detailed morphological study of astrocytes. Neurosci Res. 47, 451- 454

Tackmann W, Kuhlendahl D (1979). Evoked potentials in neuronal ceroid lipofuscinosis. Eur Neurol. 18, 234–242

Tager JM, Jonsson LV, Aerts JM, Elferink RP, Schram AW, Erickson AH, Barranger JA (1984). Metabolic consequences of genetic defects in lysosomes. Biochem Soc Trans. 12, 902-905

Tammen I, Cook RW, Nicholas FW, Raadsma HW (2001). Neuronal ceroid lipofuscinosis in Australian Merino sheep: a new animal model. Eur J Paediatr Neurol. 5, 37-41

Tammen I, Houweling PJ, Cavanagh JA, Mitchel NL, Raadsma HW, Kay GW, Palmer DN (2005). Genetic characterisation of ruminant animal models of NCLs. Abstract. NCL 2005 P39

Thomas GH. (2001). Disorders of Gycoprotein Degradation: α-Mannosidosis, β- Mannosidoses, Fucosidosis, and Sialidosis, pages 3507-3533; in The Metabolic Bases of Inherited Disease (8th edition), vol III. Eds. Scriver CR, Beaudet AL, Sly WS, Valle D. Publishers; McGraw-Hill Companies Inc.

Tramontin AD, Garcia-Verdugo JM, Lim DA, Alvarez-Buylla A (2003). Postnatal development of radial glia and the ventricular zone (VZ): a continuum of the neural stem cell compartment. Cereb Cortex. 13, 580-587

Turk R, Hoen PA, Sterrenburg E, de Menezes RX, de Meijer EJ, Boer JM, van Ommen GJ, den Dunnen JT (2004). Gene expression variation between mouse inbred strains. BMC Genomics. 5, 57

245 Tyynelä J, Cooper JD, Khan MN, Shemilts SJ, Haltia M (2004). Hippocampal pathology in the human neuronal ceroid-lipofuscinoses: distinct patterns of storage deposition, neurodegeneration and glial activation. Brain Pathol. 14, 349-357

Tyynelä J, Suopanki J, Santavuori P, Baumann M, Haltia M (1997). Variant late infantile neuronal ceroid-lipofuscinosis: pathology and biochemistry. J Neuropathol Exp Neurol. 56, 369–375

Url A, Bauder B, Thalhammer J, Nowotny N, Kolodziejek J, Herout N, Furst S, Weissenbock H (2001). Equine neuronal ceroid lipofuscinosis. Acta Neuropathol (Berl). 101, 410-414 van Diggelen OP, Thobois S, Tilikete C, Zabot MT, Keulemans JL, van Bunderen PA, Taschner PE, Losekoot M, Voznyi YV (2001). Adult neuronal ceroid lipofuscinosis with palmitoyl-protein thioesterase deficiency: first adult-onset patients of a childhood disease. Ann Neurol. 50, 269-272

Vanhanen SL, Puranen J, Autti T, Raininko R, Liewendahl K, Nikkinen P, Santavuori P, Suominen P, Vuori K, Hakkinen AM (2004). Neuroradiological findings (MRS, MRI, SPECT) in infantile neuronal ceroid-lipofuscinosis (infantile CLN1) at different stages of the disease. Neuropediatrics. 35, 27–35

Vanhanen SL, Sainio K, Lappi M, Santavuori P (1997). EEG and evoked potentials in infantile neuronal ceroid-lipofuscinosis. Dev Med Child Neurol. 39, 456–463

Vercruyssen A, Martin JJ, Ceuterick C, Jacobs K, Swerts L (1982). Adult ceroid- lipofuscinosis: diagnostic value of biopsies and of neurophysiological investigations. J Neurol, Neurosurg. Psychiatry. 45, 1056–1059

Verkruyse LA, Hofmann SL (1996). Lysosomal targeting of palmitoyl-protein thioesterase. J Biol Chem. 271, 15831-15836

Vesa J, Chin MH, Oelgeschlager K, Isosomppi J, DellAngelica EC, Jalanko A, Peltonen L (2002). Neuronal ceroid lipofuscinoses are connected at molecular level: interaction of CLN5 protein with CLN2 and CLN3. Mol Biol Cell. 13, 2410-2420

246 Vesa J, Hellsten E, Verkruyse LA, Camp LA, Rapola J, Santavuori P, Hofmann SL, Peltonen L (1995). Mutations in the palmitoyl protein thioesterase gene causing infantile neuronal ceroid lipofuscinosis. Nature. 376, 584-587

Vines D, Warburton MJ (1998). Purification and characterisation of a tripeptidyl aminopeptidase I from rat spleen. Biochim Biophys Acta. 1384, 233

Voigt T (1989). Development of glial cells in the cerebral wall of ferrets: direct tracing of their transformation from radial glia into astrocytes. J Comp Neurol. 289, 74-88

Voikar V, Koks S, Vasar E, Rauvala H (2001). Strain and gender differences in the behavior of mouse lines commonly used in transgenic studies. Physiol Behav. 72, 271– 281 von Figura K, Steckel F, Hasilik A (1983). Juvenile and adult metachromatic leukodystrophy: partial restoration of arylsulfatase A (cerebroside sulfatase) activity by inhibitors of thiol proteinases. Proc Natl Acad Sci U S A. 80, 6066-6070

Wade CM, Kulbokas EJ,III, Kirby AW, Zody MC, Mullikin JC, Lander ES, Lindblad- Toh K, Daly MJ (2002). The mosaic structure of variation in the laboratory mouse genome. Nature. 420, 574–578

Waheed A, Hasilik A, von Figura K (1982). UDP-N-acetylglucosamine:lysosomal enzyme precursor N-acetylglucosamine-1-phosphotransferase. Partial purification and characterization of the rat liver Golgi enzyme. J Biol Chem. 257, 12322-12331

Wahlsten D (1982). Deficiency of corpus callosum varies with strain and supplier of the mice. Brain Res. 239, 329-347

Walkley SU (1998). Cellular pathology of lysosomal storage disorders. Brain Pathol. 8, 175-193

Walkley SU, March PA (1993). Biology of neuronal dysfunction in storage disorders. J Inherit Metab Dis. 6, 284-287

247 Wang VY, Zoghbi HY (2001). Genetic regulation of cerebellar development. Nat Rev Neurosci. 2, 484-491

Wang Y, Gupta A, Markram H (1999). Anatomical and functional differentiation of glutamatergic synaptic innervation in the neocortex. J Physiol Paris. 93, 305-317

Warner TG, O'Brien JS (1983). Genetic defects in glycoprotein metabolism. Annu Rev Genet. 17, 395-441

Watanabe J, Asaka Y, Kanamura S (1996). Relationship between immunostaining intensity and antigen content in sections. Jnl of histochemistry and cytochemistry. 44, 1451-1458

Weibel ER (1979). Stereological methods. Vol. 1: practical methods for biological morphometry. London: Academic Press

Weibel ER (1980). Stereological methods. Vol. 2: theoretical foundations. London: Academic Press

Weimer JM, Custer AW, Benedict JW, Alexander NA, Kingsley E, Federoff HJ, Cooper JD, Pearce DA (2005a). Investigation of the Cln3-/- mouse demonstrates degeneration within the optic nerve and LGN rather than the retina as the source of visual deficits in Batten disease. Abstract. NCL 2005 P42

Weimer JM, Elshatory YM, Short DW, Alexander NA, Cooper JD, Pearce DA (2005b). Characterization of striatal dysfunction in the Cln3-/- mouse model of Batten disease. Abstract. NCL 2005 P41

Weimer JM, Federoff HJ, Pearce DA (2005c). A Cln3 null mutant mouse model of Batten disease reveals alterations in cerebellar postnatal development. Abstract. NCL 2005 O22

Weimer JM, Kriscenski-Perry E, Elshatory Y, Pearce DA (2002). The neuronal ceroid lipofuscinoses: mutations in different proteins result in similar disease. Neuromolecular Med. 1, 111-124

248 Weimer JM, Serour A, Alexander NA, Cooper JD, Federoff HJ, Pearce DA (2005d). Cln3 null mutant mice show loss of cerebellar neuron subpopulations resulting in motor coordination deficits. Abstract. NCL 2005 P43

Weissman T, Noctor SC, Clinton BK, Honig LS, Kriegstein AR (2003). Neurogenic radial glial cells in reptile, rodent and human: from mitosis to migration. Cereb Cortex. 13, 550-559

West MJ (1993). New stereological methods for counting neurons. Neurobiol Aging. 14, 275-285

West MJ (2002). Design-based stereological methods for counting neurons. Prog Brain Res. 135, 43-51

West MJ, Slomianka L, Gundersen HJ (1991). Unbiased stereological estimation of the total number of neurons in thesubdivisions of the rat hippocampus using the optical fractionator. Anat Rec. 231, 482-497

Wheeler RB, Sharp JD, Mitchell WA, Bate SL, Williams RE, Lake BD, Gardiner RM (1999). A new locus for variant late infantile neuronal ceroid lipofuscinosis-CLN7. Mol Genet Metab. 66, 337-338

Wheeler RB, Sharp JD, Schultz RA, Joslin JM, Williams RE, Mole SE (2002). The gene mutated in variant late-infantile neuronal ceroid lipofuscinosis (CLN6) and in nclf mutant mice encodes a novel predicted transmembrane protein. Am J Hum Genet. 70, 537-542

Williams MA, Fukuda M (1990). Accumulation of membrane glycoproteins in lysosomes requires a tyrosine residue at a particular position in the cytoplasmic tail. J Cell Biol. 111, 955-966

Winchester B, Vellodi A, Young E (2000). The molecular basis of lysosomal storage diseases and their treatment. Biochem Soc Trans. 28, 150-154

249 Windhagen A, Newcombe J, Dangond F, Strand C, Woodroofe MN, Cuzner ML, Hafler DA (1995). Expression of costimulatory molecules B7-1 (CD80), B7-2 (CD86), and interleukin 12 cytokine in multiple sclerosis lesions. J Exp Med. 182, 1985-1996

Wisniewski KE, Kida E, Golabek AA, Kaczmarski W, Connell F, Zhong N (2001a). Neuronal ceroid lipofuscinoses: classification and diagnosis. Adv Genet. 45, 1-34

Wisniewski KE, Zhong N, Kaczmarski W, Kaczmarski A, Kida E, Brown WT, Schwarz KO, Lazzarini AM, Rubin AJ, Stenroos ES, Johnson WG, Wisniewski TM (1998). Compound heterozygous genotype is associated with protracted juvenile neuronal ceroid lipofuscinosis. Ann Neurol. 43, 106–110

Wisniewski KE, Zhong N, Philippart M (2001b). Pheno/genotypic correlations of neuronal ceroid lipofuscinoses. Neurology. 57, 576-581

Witter MP, Amaral DG (1991). Entorhinal cortex of the monkey: V. Projections to the dentate gyrus, hippocampus, and subicular complex. J Comp Neurol. 307, 437-459

Wlodawer A, Li M, Dauter Z, Gustchina A, Uchida K, Oyama H, Dunn BM, Oda K (2001). Carboxyl proteinase from Pseudomonas defines a novel family of subtilisin-like enzymes. Nat Struct Biol. 8, 442-446

Wu VW, Schwartz JP (1998). Cell culture models for reactive gliosis: new perspectives. J Neurosci Res. 51, 675-681

Xue L, Fletcher GC, Tolkovsky AM (1999). Autophagy is activated by apoptotic signalling in sympathetic neurons: an alternative mechanism of death execution. Mol Cell Neurosci. 14, 180-198

Youdim MB, Lavie L (1994). Selective MAO-A and B inhibitors, radical scavengers and nitric oxide synthase inhibitors in Parkinson's disease. Life Sci. 55, 2077-2082

Yuan J, Lipinski M, Degterev A (2003). Diversity in the mechanisms of neuronal cell death. Neuron. 40, 401-413

250 Zeevalk GD, Schoepp DD and Nicklas WJ (1995). Excitotoxicity at both NMDA and non-NMDA glutamate receptors is antagonized by aurintricarboxylic acid: evidence for differing mechanisms of action. J Neurochem. 64, 1749–1758

Zeman W, Donahue S, Dyken P, Green J (1970). The neuronal ceroid lipofuscinoses (Batten-Vogt syndrome). In: Vinken PJ, Bruyn GW, eds. Handbook of clinical neurology Vol. 10. Amsterdam: North Holland. 588-679

Zeman W, Dyken P (1969). Neuronal ceroid-lipofuscinosis (Batten’s disease): Relationship to amaurotic family idiocy? Pediatrics. 44, 570–583

Zhong NA, Moroziewicz DN, Ju W, Wisniewski KE, Jurkiewicz A, Brown WT (2000). CLN-encoded proteins do not interact with each other. Neurogenetics. 3, 41-44

251