Viewed by Quick Freeze/Deep-Etch Electron Microscopy with the Prominent Inter-Filament “Cross- Bridges”

Home , Dynactin, Gigaxonin, Internexin, KIF5A, Spectrin

CHARACTERIZATION OF MOVING NEUROFILAMENTS IN CULTURED NEURONS

DISSERTATION

Presented in Partial Fulfillment of the Requirements for The Degree Doctor of Philosophy in the Graduate School of The Ohio State University

Yanping Yan, B.S.

******

The Ohio State University 2006

Dissertation Committee:

Professor Anthony Brown, Adviser Approved by Professor Berl R. Oakley

Professor Dale Vandre

Professor Richard W. Burry

Adviser Graduate Program in Neuroscience

ABSTRACT

Neurofilament proteins are the major cytoskeletal component of most neurons

and they are transported by slow axonal transport, whose mechanism is still

controversial. Live cell imaging has revealed that GFP-tagged neurofilament

proteins move rapidly along axons in predominantly filamentous form but the

overall rate is slow because they spend most of the time pausing. In this thesis I

test the hypotheses that these moving filamentous structures represent single

neurofilament polymers and that moving and pausing neurofilaments might differ

in their polypeptide composition or phosphorylation state. To test these hypotheses, I have developed a novel perfusion technique for capturing fluorescent filaments as they move through naturally occurring gaps in the axonal

neurofilament array of cultured sympathetic neurons. Using quantitative

immunofluorescence microscopy and correlative light and electron microscopy, I

demonstrate that the captured structures are single continuous neurofilament

polymers. To analyze their polypeptide composition, I processed the captured

filaments and filaments in detergent-splayed cytoskeletons for immunofluorescence microscopy using antibodies specific for various

neurofilament proteins. All neurofilaments contained four different neurofilament

proteins (NFL, NFM, α-internexin and peripherin) and on average each was

ii present along >95% of the neurofilament length. Since there was no difference between the polypeptide composition of the moving filaments and the overall neurofilament population, most of which are pausing at any point in time, these data indicate that moving and pausing neurofilaments do not differ in their polypeptide composition. Immunostaining with antibody RT97, which recognizes phosphoepitopes on NFM and NFH that have been correlated with slowing of neurofilament transport, revealed no evidence for hypophosphorylation of moving neurofilaments. All moving neurofilaments had the epitope distributed along almost their entire length and quantitative comparison of RT97 immunofluorescence intensity along moving and splayed neurofilaments further confirms that moving neurofilaments are not hypophosphorylated at this epitope relative to the overall population. I conclude (1) that neurofilament proteins move in the form of single assembled polymers, (2) that moving and pausing neurofilaments are complex heteropolymers comprised of at least four different subunit proteins, and (3) that phosphorylation at the RT97 epitope is not a key regulator of neurofilament movement in cultured sympathetic neurons.

iii

Dedicated to my family

ACKNOWLEDGMENTS

I would like to thank my adviser, Prof. Anthony Brown, for his unwavering support

and guidance throughout my graduate education. His guidance and insight were

instrumental in the success of the research presented in this dissertation.

I am grateful to Prof. Richard W. Burry, Kathy Wolken, and Brian Kemmenoe of

the Ohio State University Central Microscopy and Imaging Facility for their

insightful suggestions and the technical support that they provided me during the

electron microscopic studies.

I would like to thank Lei Wang, Atsuko Uchida and Kitty Jensen in the Brown laboratory for their stimulating discussions and for their help throughout the

research.

This research is supported by a grant from National Institute of Health.

VITA

February 07, 1977……………………Born-Shaoyang, P.R. China

2000……………………………...... B.S. Neurobiology and Biophysics University of Science & Technology of China

1999-2000…………………………….Undergraduate Research Assistant Shanghai Institute of Physiology, Chinese Academy of Sciences

2000 – 2001…………………………. Graduate Research Associate Ohio University

2001 – present……………………….Graduate Research Associate The Ohio State University

PUBLICATIONS

Research Publications

1. Yan, Y., Brown, A. (2005). Neurofilament polymer transport in axons. J. Neurosci. 25(30):7014-7021.

2. Ji, Y.H., Wang, W.X., Ye, J.G., He, L.L., Li, Y.J., Yan, Y.P., Zhou, Z. (2003). Martentoxin, a novel K+-channel-blocking peptide: purification, cDNA and genomic cloning, and electrophysiological and pharmacological characterization. J Neurochem. 84(2):325-335.

3. Ye, J.G., Wang, C.Y., Li, Y.J., Tan, Z.Y., Yan, Y.P., Li, C., Chen, J., Ji, Y.H. (2000). Purification, cDNA cloning and function assessment of BmK abT, a unique component from the Old World scorpion species. FEBS Lett. 479(3):136- 140.

vi FIELDS OF STUDY

Major Field: Neuroscience

vii

TABLE OF CONTENTS

Page

Abstract ….……………………………………………………………………………….ii

Dedication ...…………………………………………………………………………….iv

Acknowledgments……………………………………………………………………….v

Vita ………………………………………………………………………………………vi

List of Tables...... xiv

List of Figures……………………………………………………….………………….xv

CHAPTER 1

INTRODUCTION ...... 1 1.1 Neurofilaments and neurofilament function ...... 1 1.1.1 General introduction about neurofilaments ...... 1 1.1.2 Neurofilament function ...... 4 1.1.3 Molecular mechanisms of neurofilament organization ...... 6 1.1.4 Neurofilament phosphorylation and axonal radial growth...... 12 1.2 Neurofilament assembly ...... 14 1.2.1 Neurofilament assembly model...... 14 1.2.2 Assembly properties of neurofilament polypeptides...... 17 1.2.3 Molecular mechanism of neurofilament assembly...... 19 1.2.4 Roles of different regions of neurofilament proteins in the assembly...... 21 1.2.5 Role of phosphorylation in neurofilament assembly...... 21 1.3 Neurofilament protein expression ...... 24 1.3.1 Neurofilament protein expression in developing nervous system . 24 1.3.2 Neurofilament protein expression in adults ...... 28 1.3.3 Neurofilament protein expression in cultured neurons ...... 29 1.4 Neurofilament protein modification ...... 31 viii 1.4.1 Neurofilament phosphorylation ...... 31 1.4.2 Neurofilament associated non-proline directed kinases ...... 34 1.4.3 Neurofilament associated proline directed kinases ...... 35 1.4.4 Neurofilament associated phosphatases ...... 37 1.4.5 Regulation of neurofilament phosphorylation by macromolecular complexes...... 38 1.4.6 Neurofilament glycosylation ...... 39 1.5 Axonal transport...... 40 1.5.1 Fast and slow axonal transport ...... 40 1.5.2 The subunit transport model for cytoskeletal proteins ...... 41 1.5.3 The polymer transport model for cytoskeletal proteins...... 45 1.5.4 Recent live cell imaging studies on slow axonal transport ...... 45 1.5.5 Neurofilament transport motors...... 47 1.5.6 Regulation of neurofilament transport by phosphorylation ...... 48 1.6 Neurofilaments and neurodegenerative diseases...... 53 1.6.1 Transgenic mice with neurofilament protein overexpression...... 55 1.6.2 Neurodegenerative disease transgenic mouse model ...... 56 1.6.3 Neurofilament gene mutations associated with neurodegenerative diseases...... 58 1.6.4 Neurofilament transport and neurodegenerative diseases...... 60

CHAPTER 2

MATERIALS AND METHODS...... 62 2.1 Cell culture...... 62 2.1.1 SCG neuron culture ...... 62 2.1.2 SW13 cell culture ...... 62 2.2 Transfection...... 63 2.2.1 Plasmid DNA...... 63 2.2.2 Nuclear injection...... 64 2.2.3 Lipofection...... 65 2.3 Live cell imaging ...... 66 2.3.1 Perfusion chamber system...... 66 2.3.2 Image acquisition ...... 67 2.4 Capture, fixation and attempted reactivation of moving filaments.. 69 2.4.1 Capture of moving filament ...... 69 2.4.2 Fixation of moving filaments for immunofluorescence microscopy 69 2.4.3 Fixation of moving filaments for electron microscopy...... 71 2.4.4 Attempted reactivation of moving filaments...... 71 2.5 Neurofilament splaying ...... 72 ix 2.5.1 Use of sister coverslips ...... 72 2.5.2 Neurofilament splaying in the chamber ...... 72 2.5.3 Neurofilament splaying in open dishes...... 72 2.6 Immunostaining ...... 73 2.6.1 Immunostaining of captured and splayed filaments ...... 73 2.6.2 Immunostaining of SW13 cells...... 74 2.7 SDS-PAGE and western blotting ...... 79 2.7.1 Preparation of cytoskeletal proteins ...... 79 2.7.2 SDS-PAGE and immunoblotting ...... 79 2.7.3 Enzymatic dephosphorylation ...... 80 2.8 Analysis of fluorescence intensity along captured and splayed filaments...... 81 2.8.1 Image acquisition ...... 81 2.8.2 Criteria for the selection of splayed filaments for analysis...... 81 2.8.3 Quantification of fluorescence labeling along captured and splayed filaments ...... 82 2.8.4 Intrinsic flat field correction...... 82 2.8.5 Quantification of the fluorescence intensity along captured and splayed filaments ...... 84 2.8.6 Quantification of the fluorescence intensity along axons...... 89 2.9 Electron microscopy ...... 89 2.9.1 Strategy developed to locate the axons of interest...... 89 2.9.2 Electron microscopy...... 90 2.9.3 Reconstruction of the axon and measurement of neurofilament diameter...... 94

CHAPTER 3

LIGHT MICROSCOPIC ANALYSIS OF THE STRUCTURE OF MOVING NEUROFILAMENTS ...... 96 3.1 Introduction...... 96 3.2 Naturally occurring gaps ...... 97 3.3 Neurofilament splaying ...... 100 3.4 Detection of neurofilaments using GFP-NFM ...... 102 3.5 Permeabilization efficiency of 0.02% saponin in the perfusion chamber ...... 103 3.6 The capture of moving filaments...... 106 3.7 Attempted reactivation of movement after permeabilization ...... 109 3.8 Resistance of the captured filaments to detergent...... 112 x 3.9 Immunofluorescence microscopy of captured filaments ...... 112 3.10 Comparison of the captured and splayed filaments...... 113 3.11 Statistical analysis...... 116 3.12 Summary...... 119

CHAPTER 4

ELECTRON MICROSCOPIC ANALYSIS OF THE STRUCTURE OF MOVING NEUROFILAMENTS ...... 120 4.1 Introduction...... 120 4.2 Strategy and criteria for locating the captured filament...... 121 4.3 Ultra-thin sectioning ...... 122 4.4 Semi-thin sectioning ...... 124 4.5 The diameter of captured filaments ...... 125 4.6 Summary ...... 132

CHAPTER 5

POLYPEPTIDE COMPOSITION OF MOVING NEUROFILAMENTS...... 133 5.1 Introduction...... 133 5.2 Strategy ...... 133 5.3 Characterization of the specificity of the antibodies using Western blotting ...... 134 5.4 Characterization of the specificity of the antibodies using SW13 cl.2 Vim- cell system...... 137 5.5 Characterization of the staining quality of the antibodies using SW13 cl.2 Vim- cell system...... 141 5.6 Characterization of NFH antibodies...... 143 5.7 NFM incorporation along the filaments ...... 149 5.8 Peripherin incorporation along the filaments...... 150 5.9 Alpha-internexin incorporation along the filaments ...... 153 5.10 NFL incorporation along the filaments ...... 154 5.11 Summary ...... 156

xi CHAPTER 6

PHOSPHORYLATION STATE OF MOVING NEUROFILAMENTS...... 157 6.1 Introduction...... 157 6.2 Strategy ...... 158 6.3 Naturally occurring gaps in the neurofilament array lack RT97 and the SMI36 staining ...... 159 6.4 RT97 and SMI36 staining along splayed filaments from transfected and untransfected neurons………………………………………….160 6.5 RT97 and SMI36 staining of the captured filaments ...... 162 6.6 RMO55 and FNP7 staining of the captured filaments...... 166 6.7 Quantification of RT97 staining intensity...... 168 6.8 Quantitative analysis of the influence of the fixation conditions on RT97 staining...... 169 6.9 Quantitative comparison of RT97 staining along captured and splayed neurofilaments ...... 170 6.10 Summary ...... 174

CHAPTER 7

DISCUSSION 173 7.1 The capturing technique ...... 175 7.2 Neurofilament polymer transport in axons ...... 177 7.2.1 Validation of the polymer transport hypothesis ...... 177 7.2.2 Neurofilament polymers as carrier structures...... 179 7.2.3 Mechanism of neurofilament polymer transport ...... 180 7.2.4 Cut and run model for cytoskeletal polymer transport...... 181 7.3 Polypeptide composition of moving neurofilaments...... 184 7.3.1 Both moving and pausing neurofilaments in cultured SCG neurons are complex heteropolymers…………...... 184 7.3.2 Neurofilament polypeptides and the transport machinery ...... 186 7.4 Phosphorylation of neurofilament proteins and neurofilament transport...... 189 7.4.1 Regulation of neurofilament pausing behaviors ...... 189 7.4.2 Moving neurofilaments are not hypophosphorylated at the RT97 epitope compared with pausing neurofilaments...... 191 7.5 Future directions ...... 195

xii LIST OF REFERENCES...... 196

xiii

LIST OF TABLES

TABLE 1.1 Assembly properties of rodent and bovine neurofilament polypeptides...... 19

TABLE 2.1 Antibodies used in the experiments...... 75

TABLE 5.1 Characterization of NFH antibodies...... 146

xiv

LIST OF FIGURES

Figure 1.1 Comparison of the structures of intermediate filament subunit proteins expressed in neurons...... 3

Figure 1.2 Electron micrographs of neurofilaments...... 7

Figure 1.3 Schematic representations of three models for neurofilament interaction...... 8

Figure 1.4 Model diagram of neurofilament assembly...... 15

Figure 1.5 Generalized developmental profile of intermediate filaments that express in neurons...... 25

Figure 1.6 Schematic diagram of multiple sites on neurofilament proteins (NF-L, NF-M and NF-H) that can potentially be phosphorylated by different kinases...... 32

Figure 1.7 The polymer and subunit models for axonal transport of cytoskeletal proteins...... 42

Figure 1.8 Correlation between regional neurofilament accumulation and appearance of phosphoepitope RT97 on NFH during development.. 51

Figure 1.9 Motile behaviors of neurofilaments in cortical neurons...... 54

Figure 2.1 Perfusion chamber system...... 68

xv Figure 2.2 Schematic diagram of immunostaining protocol for neurofilaments used for quantitative analysis...... 78

Figure 2.3 Methods for analyzing the proportion of the neurofilament length that contains fluorescence...... 83

Figure 2.4 Examples of raw and flat-field-corrected images...... 85

Figure 2.5 The operation of a convolution matrix in the high pass heavy 7x7 filter...... 87

Figure 2.6 Methods for analyzing the fluorescence intensity along the filaments...... 88

Figure 2.7 Schematic diagram of the operation used to facilitate location of the axon in electron microscope...... 91

Figure 2.8 Strategy to locate the axon of interest...... 92

Figure 3.1 A naturally occurring gap...... 99

Figure 3.2 Detergent-induced splaying of axonal neurofilaments...... 102

Figure 3.3 GFP-NFM incorporation along single neurofilaments...... 104

Figure 3.4 Permeabilization with 0.02% saponin...... 106

Figure 3.5 Capture of a moving filament...... 108

Figure 3.6 Comparison of captured and splayed filaments...... 114

Figure 3.7 A captured filament immunostained for NFL...... 115

Figure 3.8 Fluorescence intensity analysis and comparison...... 118

Figure 4.1 The pre-fixation extraction technique...... 126

xvi Figure 4.2 Electron microscopy of a captured filament using ultra-thin sectioning...... 127

Figure 4.3 The post-fixation extraction technique...... 129

Figure 4.4 Electron microscopy of a captured filament using semi-thin sectioning...... 130

Figure 5.1 Characterization of the specificities of antibodies by immunoblotting...... 136

Figure 5.2 Characterization of the specificity of the antibodies using SW13 cl.2 Vim- cells...... 140

Figure 5.3 Characterization of the quality of the antibody staining using SW13 cl.2 Vim- cells...... 142

Figure 5.4 Characterization of NFH antibody RMO217...... 147

Figure 5.5 Characterization of NFH antibody AB1989...... 148

Figure 5.6 Characterization of the polypeptide distribution along captured and splayed neurofilaments...... 155

Figure 6.1 Comparison of RT97 and SMI36 staining of gaps in the axonal neurofilament array axons...... 160

Figure 6.2 Comparison of RT97 and SMI36 phosphoepitope distribution along splayed filaments from transfected and untransfected cells...... 163

Figure 6.3 Characterization of RT97 and SMI36 phosphoepitope distribution along captured filaments...... 165

Figure 6.4 Characterization of RMO55 and FNP7 phosphoepitope distribution along captured filaments...... 167

xvii Figure 6.5 Comparison of RT97/NFLAS immunofluorescence intensity ratio along splayed neurofilaments processed in the open dishes and those processed in open dishes...... 171

Figure 6.6 RT97 immunofluorescence intensity analysis and comparison...... 173

xviii CHAPTER 1

INTRODUCTION

1.1 Neurofilaments and neurofilament function

1.1.1 General introduction about neurofilaments

Neurofilaments are one of the most abundant cytoskeletal proteins in neurons. They convey mechanical strength on axons and dendrites and function as spacers contributing to the diameter of these processes, which determines the rate of nerve impulse conduction.

Neurofilaments are type IV intermediate filaments. These filaments are called intermediate because they are 10nm in diameter, which is between the 24nm thick microtubules and the 6-8nm thin microfilaments. Based on molecular structure homology, intermediate filaments are classified into six groups (I-VI). The whole family can be divided into two broad groups, based on whether they are expressed in the cytoplasm or in the nucleus. Type I-IV and type VI intermediate filaments are present in the cytoplasm while type V, nuclear lamins, are part of the nuclear lamina architecture (Aebi et al., 1986; Mckeon et al., 1986; Steinert and Roop, 1988). Based on the cell type where these cytoplasmic filaments are mainly expressed, they can be further divided into epithelial and non-epithelial groups. Acid and basic keratins are type I and II intermediate filaments (Fuchs et al., 1985), which are exclusively found in epithelia. Type III intermediate filaments include desmin, GFAP (glial fibrillary acidic protein), 1 peripherin and vimentin. Peripherin is a neuron specific intermediate filament (Portier et al., 1983; Leonard et al., 1988; Parysek et al., 1988; Greene 1989). Vimentin is mainly found in non-neurons in adults, but it is also expressed in neuroblasts and certain unusual neurons (Drager, 1983; Shaw and Weber, 1983; Shaw and Weber, 1984; Schwob et al., 1986). Neurofilament proteins have three major subunits: NFL (low molecular weight, 66 kD), NFM (intermediate molecular weight, 95-100 kD) and NFH (high molecular weight, 110-115 kD) (Lee and Cleveland, 1996). These triplet proteins share similarity in protein and genomic sequence with α-internexin, which is also neuron specific (Chiu et al., 1989; Kaplan et al., 1990), and they are all classified as type IV intermediate filament proteins. Nestin, which is originally found in neural epithelial stem cells and certain muscle tissue (Lendah et al., 1990), is defined as a type VI intermediate filament (Dahlstrand et al., 1992).

Like other intermediate filament proteins, neurofilament triplet proteins share an N-terminal globular head domain, α-helical coil-coiled rod domain and C-terminal tail domain (Figure 1.1). The head domains of neurofilament triplet proteins are rich in serines and threonines, and phosphorylation and O-glycosylation of these residues are believed to be important for regulation of neurofilament assembly (see details in section 1.2.5). The rod domain of NFL is similar to other intermediate filaments, with an interruption of the heptad repeat in coil1, while this domain in NFM and NFH is predicted to form a single continuous coiled coil (Figure 1.1). The diversity of neurofilament triplet proteins resides primarily in the carboxyl-terminal tail domain. For NFL, this region contains many glutamate residues comprising a segment referred to as the “E” segment. Distinct from NFL, NFM and NFH have much longer tail domains. In addition to E segments, they contain numerous lysine-serine-proline (KSP) repeats, with 42-51 in NFH, depending on the species, and much fewer in NFM (Shetty et al., 1993; Shaw, 1991). The serines in these KSP domains are heavily phosphorylated in axons (Julien and Mushynski, 1982; Julien and Mushynski, 1983; Carden et al., 1985;

2 Lee et al., 1988; Xu et al., 1992; Elhanany et al., 1994). Phosphorylation of the tail domains of NFM and NFH have been suggested to be involved in the elongation of neurofilaments and regulation of axonal calibers and neurofilament transport along the axon (Grant and Pant, 2000, see details in section 1.1.4 and 1.5.6).

Figure 1.1 Comparison of the structures of intermediate filament subunit proteins expressed in neurons. These intermediate filament proteins include a relatively conserved α helical rod domain and differ predominantly in N- and C- terminal domains. A unique characteristic of NFM and NFH is that their C-termini have multiple KSP repeats, many of which are phosphorylated in vivo. Also shown here, are posttranslational modifications, including phosphorylation and glycosylation on neurofilament subunits. The image is reproduced from Liu et al., (2004) with permission of Birkhäuser Publishing Inc.

3 1.1.2 Neurofilament function

Neurofilaments are required for axons to achieve their mature caliber in vertebrates. During radial axonal growth, there is a direct correlation between the axonal diameter and the number of neurofilaments (Friede and Samorajski, 1970; Hoffman, 1988; Cleveland et al., 1991). The study of a recessive mutant of a Japanese quail “quiverer” has demonstrated unequivocally the crucial role of neurofilaments in radial axonal growth. The “quiverer” quail is characterized by mild quivering, ataxia and a reduction in gross brain weight (Yamasaki et al., 1991). It turned out that there was a nonsense mutation in NFL and there is no detectable NFL either in axons or in the cell bodies (Ohara et al., 1993). NFM and NFH, which are unable to form neurofilaments in the absence of NFL (see section 1.2.2), form non-filamentous aggregates in the cell body. The average diameter of myelinated axons was significant smaller than in normal quails although the microtubule content was not affected. As a result, there is a reduction in nerve impulse conduction, which is responsible for the phenotype.

Studies on neurofilament knock out mice further support the important role of neurofilaments in controlling axonal caliber. In NFL single knock out mice or NFM and NFH double knock out mice, no neurofilaments are found in the axons and no axons attain large caliber in spite of an increase in microtubule number (Zhu et al., 1997; Elder et al., 1999b; Jacomy et al., 1999). In NFM knock out mice, NFL is down regulated and the number of neurofilaments is significantly reduced (Elder et al., 1998; Elder et al., 1999a). The axonal caliber for the ventral roots fail to increase in these mice as the animals mature (4 month decrease 20% and 2 years decrease >50%), while in the NFH knock out mice (Zhu et al., 1998; Rao et al., 1998; Kriz et al., 2000), in which there is no significant change in the neurofilament number in the mature axons, overall there is only a mild decrease in axonal caliber (2 years decrease 10%). It is quite possible that the increase in NFM levels and number of microtubules partially compensated for the absence of NFH in these mice. 4 Similar results were reported in a transgenic mouse expressing an NFH/β- galactosidase fusion protein in which the C-terminus of NFH was replaced by β- galactosidase. Neurofilament inclusions in the perikarya of neurons and depletion of axonal neurofilament proteins were associated with reduced axonal calibers (Eyer and Peterson, 1994). How the NFH/β-galactosidase fusion protein exactly caused the neurofilament inclusions in the cell body is still not clear. There are several possible ways that this fusion protein might have interfered with neurofilament function: first, this fusion protein could have impaired neurofilament assembly; second, it could have interfered with proper modification of neurofilaments, like phosphorylation and glycosylation in NFH tail domain (see details in section 1.4) and subsequently caused impaired neurofilament transport producing the misaccumulation; third, this fusion protein might have affected normal interfilament interaction and caused over-linked neurofilaments due to the tendency of galactosidase to form tetramers.

In addition to the studies with knock out mouse models, a series of studies have been carried out using transgenic mice overexpressing NFL, NFM and NFH individually and these studies demonstrate the importance of proper stoichiometry of neurofilament triplet proteins for the regulation of axonal caliber (Xu et al., 1993; Vickers et al., 1994; Wong et al., 1995; Marszalek et al., 1996; Xu et al., 1996). In all three models, there are neurofilament inclusions (see details in section 1.6.1) and the radial growth of axons is inhibited (Xu et al., 1996). Simultaneous overexpression of NFL with either NFM or NFH stimulated radial axonal growth of large axons (axon diameter> 3µm) in these mice (50% increase in NFL+NFM mice; 30% in NFL+NFH mice), whereas it caused no change in the small axons (axon diameter < 3µm). On the contrary, overexpressing of both NFM and NFH yielded the smallest axons among all the transgenic mice lines described, with an overall 55% reduction in axonal volume (Xu et al., 1996).

5 In principle, neurofilaments can affect axonal caliber through changes in the number of filaments and/or changes in the filament spacing in the axons, mediated by the interaction between filaments and/or between filaments and other axonal components. The impaired radial axonal growth in the various neurofilament knock out and transgenic mice described above could have been generated through either mechanism or a combination of both mechanisms. Modifications on neurofilament proteins (e.g. transgenic mice expressing NFH/β- galactosidase fusion protein) and changes in the expression level of different neurofilament polypeptides (e.g. neurofilament knock out and overexpression mice) could both affect the neurofilament assembly and neurofilament numbers in axons and thus both could affect axon caliber. The interaction between filaments and between filaments and other axonal components could also have been altered by changes in the polypeptide compositions and phosphorylation of neurofilaments in these mice (see details in next section 1.1.3).

1.1.3 Molecular mechanisms of neurofilament organization

As mentioned earlier, the interaction between neurofilament is one of the important factors that contribute to axonal radial growth. It is now generally believed that the carboxyl terminal domains of NFM and NFH mediate many of these inter-neurofilament interactions. During neurofilament assembly, the rod domains of the neurofilament subunits form the filament backbone and the carboxyl terminal domains of NFM and NFH, which contain more than 300 and 600 residues respectively, protrude from the core the filament to form the neurofilament “sidearms” (see details in section 1.2.1). Low angle rotary shadowing electron microscopy of neurofilaments reconstituted in vitro reveals that these sidearms extend 50-77nm from the filament core (Hisanaga and Hirokawa, 1988; Hisanaga and Hirokawa, 1990) (Figure 1.2A). But how these sidearms interact with each other and contribute to the axonal radial growth is still controversial.

Figure 1.2 Electron micrographs of neurofilaments. (A) shows a neurofilament assembled in vitro as visualized by glycerol spraying followed by low angle rotary shadowing. (B) shows neurofilaments viewed by quick freeze/deep-etch electron microscopy with the prominent inter-filament “crossbridges” . The arrows point two examples. Note cross bridges correspond to the protrusions sticking out of the core of the neurofilament in (A), which are formed by the C-terminal domain of NFH and NFM. Image A is reproduced from Hisanaga and Hirokawa (1988) with permission of Elsevier and image B is reproduced from Hirokawa (1982) with permission of The Rockefeller University Press.

Three distinct models have been proposed to explain the molecular mechanism. One model proposes that neurofilament sidearms interact with each other through binding or cross-bridging either directly or indirectly via accessory factors (Figure 1.3A). In support of this model, quick freeze/deep etching electron micrographs of nerve tissue have revealed extensive cross-links that bridge neighboring neurofilaments or microtubules (Hirokawa, 1982) (Figure 1.2B). Hirokawa and his colleagues further demonstrated that these cross bridges are indeed formed by the carboxyl terminal tail domains of NFM and NFH by using various deletion mutants lacking portions of the tail domains (Nakagawa et al., 1995; Chen et al., 2000). Properties of highly viscous gels formed by neurofilament in vitro have also been suggested to be a consequence of interneurofilament crossbridging through variable antiparallel overlaps of the

Figure 1.3 Schematic representations of three models for neurofilament interaction. The core of neurofilaments are shown in cross sections as round filled circles and the squiggly lines represent side arms formed by the carboxyl terminal tail domains of NFM and NFH sticking out from core of the filaments. (A) In the cross-bridge model, neurofilaments are crossed-linked by side arms that interact with each other, and the linkers are indicated by small filled rectangles. These side arms may interact directly or with the help of other accessory proteins such as BPAG1. (B) In the electrostatic model, neurofilaments repel each other through direct, colloidal electrostatic forces, which are mediated by a net negative charge on the extensive phosphorylated side arms. (C) In the polymer-brush model, the side arms are highly unstructured and they are constantly in rapid Brownian motion forming a so-called polymer brush. In a relative short period of time, on the order of nanoseconds, these side arms can adopt a very large number of conformations, essentially filling some characteristic space. Proteins entering this space tend to be excluded based on entropic considerations. Thus as two neurofilaments are brought together, the polymer brush gives rise to a repulsive interaction between the filaments (Mukhopadhyay et al., 2004). Neurofilaments in a, b and c are hypophosphorylated and neurofilaments in a’, b’ and c’ are hyperphosphorylated. Compared with neurofilaments in a, b and c, those in a’, b’ and c’ have acquired more negative charges on the sidearms. Thereby, the sidearms are more extended and more repulsive forces are generated between adjacent neurofilaments, which leads to more neurofilament spacing. The diagram is adapted from Mukhopadhyay et al (2004).

Figure 1.3

9 phosphorylable KSP domains of sidearms on adjacent neurofilaments (Leterrier and Eyer, 1987; Leterrier et al., 1996; Gou et al., 1998). Neuronal bullous pemphigoid antigen 1 (BPAG1n) was identified as a cross-linker between neurofilaments and actin filaments (Yang et al., 1996) but no cross-linking agent between neurofilaments has been identified.

In contrast to the attractive forces proposed in the first model, the other two models propose that the interaction between neurofilament is primarily repulsive, either due to the electrostatic or entropic forces of the sidearms. In the second model, neurofilaments acquire net negative charges on the sidearms through extensive phosphorylation and they repel each other through direct, colloidal electrostatic forces (Figure 1.3B). This model is consistent with the observed correlation between neurofilament phosphorylation and average interneurofilament spacing in vivo (de Waegh et al., 1992) but there hasn’t been any direct evidence in support of it.

Most recently, Hoh and his colleagues proposed a third model, known as the polymer brush model, in which the sidearms are highly unstructured polyelectrolyte chains that undergo rapid Brownian motion to occupy a large effective volume, producing an entropic repulsion when two filaments come near each other (Figure 1.3C) (Brown and Hoh, 1997; Kumar et al., 2002a; Kumar et al., 2002b; Mukhopadhyay et al., 2002; Kumar and Hoh, 2004). Evidence for this model comes primarily from atomic force microscopy (AFM) studies (Brown and Hoh, 1997; Hoh, 1998; Kumar and Hoh, 2004). In these studies, AFM imaging of isolated native neurofilaments revealed robust exclusion of small particles from the filament backbone (Brown and Hoh, 1997, Kumar and Hoh, 2004), which was attenuated when the filaments were enzymatically dephosphorylated (Kumar and Hoh, 2004). This exclusionary force was absent in the studies using NFL homopolymers, which lack side arms (Brown and Hoh, 1997). Thus the authors propose that the sidearms potentially formed by NFM and NFH are in constant thermal motion, which produces an entropic brushing force that can exclude 10 large molecules (including other neurofilament side arms). This entropic brush provides a molecular mechanism for maintaining interfilament spacing within an axon.

In line with these studies, Kumar et al. (2002a) report that repulsion due to either entropic or electrostatic forces can account for the axonal neurofilament distribution in wild type mice, while an attractive crossbridging interaction apparently cannot. Studies were further carried out to exam if repulsive forces between neurofilaments can explain the change in neurofilament distribution in the sciatic nerves of mice lacking myelin-associated glycoprotein (MAG). MAG is a membrane glycoprotein that transduces signals from the myelin sheath to the axonal cytoplasm and it is suggested to influence NF phosphorylation and spacing, and consequently axonal caliber (Yin et al., 1998, see details in next section 1.1.4). In MAG deficient mice, neurofilaments are densely packed and there is marked reduction in axonal diameter. It has been demonstrated that the difference in neurofilament distribution between wild-type and MAG deficient mice is best explained by a model in which neurofilaments interact through repulsive forces that weaken in MAG deficient mice (Kumar et al., 2002b). The fact neurofilaments move apart freely when released from the confines of the plasma membrane also suggests that Brownian forces are sufficient to separate neurofilaments, and that interactions between neurofilaments are either repulsive or very weak (Brown and Lasek, 1993; Brown, 1997).

In spite of all these differences, all three models agree on the important role of phosphorylation in the regulation of interfilament interaction and thus the axonal radial growth (Figure 1.3). When the carboxyl terminal regions of NFM and NFH become more extensively phosphorylated, the negative charge on the side arms is increased (Glicksman et al., 1987; Myers et al., 1987), which subsequently

11 leads to lateral extension of side arms (Hisanaga and Hirokawa, 1989) or an increase in repulsive force and thereby an increase in neurofilament spacing (Chen et al., 2000; Kumar and Hoh, 2004).

1.1.4 Neurofilament phosphorylation and axonal radial growth

Neurofilament-dependent axonal radial growth itself has been associated with phosphorylation of the carboxyl terminal tail domains of both NFM and NFH (de Waegh et al., 1992; Sanchez et al., 2000). Phosphorylation of these tail domains can regulate axonal caliber through two different ways. Firstly, phosphorylation at these regions can lead to an increase in neurofilament spacing (see details in above section 1.1.3) (Chen et al., 2000; Kumar and Hoh, 2004). Secondly, due to increased interfilament interaction mediated by the increased phosphorylation, the axonal transport of neurofilament proteins may slow down, leading to a local accumulation of neurofilaments, which further contributes to the axonal caliber (see details in section 1.5.5).

Several lines of evidence have suggested that the phosphorylation of neurofilament sidearms is regulated by myelinating cells (de Waegh et al., 1992; Yin et al., 1998). Neurofilaments are most extensively phosphorylated in the internode region of myelinated axons and least at the nodes of Ranvier, where axon caliber is significantly reduced (Mata et al., 1992; Hsieh et al., 1994). Further, neurofilament phosphorylation is considerably decreased in the Trembler mouse mutant that is marked by hypomyelinated axons with reduced calibers (de Waegh et al., 1992). It appears that myelin-associated glycoprotein (MAG) on the adaxonal glia cell membrane might be one of the signaling molecules that mediates the local interaction between myelinating glia and axons and signals phosphorylation of neurofilaments and the accompanying changes in the cytoskeleton (Yin et al., 1998).

12 Several studies have demonstrated that the probable axonal MAG receptor responsible for transducing the myelin signal into axons is the low affinity nerve growth factor receptor p75NTR (Wang et al., 2002: Wong et al., 2002; Yamashita et al, 2002), which subsequently activates the extracellular-signal regulated kinases (Erk1/2) and cyclin-dependent kinase-5 (Cdk5). In MAG knock-out mice, the decreased neurofilament phosphorylation was associated with decreased activity of Erk1/2 and Cdk5 (Yin et al., 1998). In the primary dorsal root ganglion (DRG) neurons and PC12 cells that were co-cultured with COS cells stably transfected with MAG, there was increased expression of phosphorylated NFM and NFH, which was also associated with increased activity of Erk1/2 and Cdk5 (Dashiell et al., 2002). Both Erk1/2 and Cdk5 have been shown in vitro to phosphorylate NFM and NFH respectively (see details in section 1.4.3).

Both NFM and NFH tail domains could be targets of this “outside-in” signaling cascade originating from myelinating cells. To test the role of NFH tail domain and its phosphorylation in the axonal radial growth, Cleveland and his colleagues created a tailless NFH “knock-in” mice, in which the NFH gene was replaced by one truncated in the tail domain (Rao et al., 2002). The loss of NFH tail domain along with all the possible phosphorylation sites slows the acquisition of normal axon diameter but there is no reduction in diameter for mature axons. However in these mice NFM phosphorylation was up-regulated, which could have compensated the loss of NFH phosphorylation sites. To further test the role of NFM tail domain in axon radial growth, both the Cleveland and Nixon labs created tailless NFM knock-in mice (Garcia, et al., 2003; Rao et al., 2003). In these mice, axon caliber reduction was accompanied by reduced spacing between neurofilaments and loss of long cross bridges with no change in neurofilament protein content (Rao et al., 2003). To conclude that the NFM tail and its phosphorylation are essential for radial growth of large myelinated axons, the Cleveland group compared tailless NFM knock-in, and tailless NFM and NFH double knock-in mice (Garcia et al., 2003). Since the axonal size distributions

13 from two different mutants were indistinguishable at six months, they concluded that the tail domain of NFM, but not NFH, was essential for establishing mature calibers of large axons.

1.2 Neurofilament assembly

1.2.1 Neurofilament assembly model

Most of what we know about neurofilament assembly is based on extrapolation from what is known for other intermediate filaments, such as keratin. Assembly of neurofilaments occurs through a complex series of steps (see Figure 1.4). Neurofilament subunits first make a parallel, un-staggered alpha helical coiled- coil dimer. Studies on the type I and type II keratin filaments provided strong evidence that this parallel un-staggered dimer formation is the first step of intermediate filament assembly (Hetzfeld and Weber, 1990). These dimers then laterally interact to form an anti-parallel tetramer, which has been suggested to be the basic subunit of intermediate filaments (Geisler et al., 1985; Geisler, 1992). The tetramer can be formed either in approximate registered mode or half staggered mode, but in the case of neurofilaments, the tetramers are mostly found formed in half staggered mode (Hisanaga and Hirokawa, 1990a; Heins et al., 1993; Steinert et al., 1993a; Steinert et al., 1993b; Stewart, 1993; Fuchs and Weber, 1994; Heins and Aebi, 1994). End to end association of the tetramers then form a protofilament. It has been postulated that the 10nm intermediate filament is eventually built up of four protofibrils, each formed of eight alpha helical coiled-coil dimers (four tetramers or two protofilaments). In vitro reassembly studies on keratin filaments have determined that there is an average of 32 subunits (in total sixteen alpha helical coiled-coil dimers) per filament cross section (Parry and Steinert, 1995). Some other studies on vimentin and keratin have given a range of from 24 to 40 subunits and beyond (Aebi et al., 1983; Steven et al., 1983; Engel et al., 1985). Low-angle rotary shadowing 14

Figure 1.4 Model diagram of neurofilament assembly. In the neurofilament assembly process, two NF subunits (NFL, and either NFM or NFH) first form parallel head-to-tail coiled-coil dimers, then two dimers form an anti-parallel half- staggered tetramer, two tetramers form a protofilament, and finally four protofilaments (two protofibrils) form the 10-nm NF. There are ~32 molecules in the cross section, and side arms formed by the hyperphosphorylated C-termini of NF-H and NF-M stick out of the stem of the filament. The diagram is reproduced from Liu et al. (2004) with permission of Birkhäuser Publishing Inc.

15 electron microscopy study of unraveling neurofilaments reconstructed in vitro from NFL showed that filaments were composed of predominantly four protofibrils but those with five were also observed (Hisanaga and Hirokawa, 1990b).

With exception of human NFL, NFL is incapable of forming homopolymers in vivo. Nevertheless, it can be induced to form homopolymers in vitro under appropriate solution conditions (see section 1.2.2). When 10nm neurofilaments are assembled with NFL alone in vitro, the filaments appear smooth walled as observed by negative staining electron microscopy, similar to non-neuronal intermediate filaments assembled in vitro (Geisler and Weber, 1981; Liem and Hutchison, 1982; Zackroff and Idler, 1982; Tokutake et al., 1984). On the other hand, if neurofilaments are assembled with all neurofilament triplet polypeptides, they appear rough walled, which suggests that NFM and NFH are responsible for the rough-wall appearance. Quick freeze/deep etch and low-angle rotary shadowing electron microscopy studies clearly demonstrate fine fibrous side protrusions extending from the side of the filaments, which potentially account for the rough walled appearance (Hirokawa, 1982; Hisanaga and Hirokawa, 1988; see figure 1.2 in section 1.1.3). All these observations have lead to the hypothesis that the backbones of neurofilaments are composed of NFL, and that NFM and NFH are just associated (hanging around) with the filament surface. However, immunoelectron microscopy using antibodies specific to the head, rod, or tail domains of individual neurofilament triplet proteins have demonstrated that all three proteins are incorporated integrally into the filaments (Balin & Lee 1991, Balin et a1 1991). Now it is generally believed that the rod domain of neurofilament triplet proteins form the core of the filament while the carboxyl- terminal tail domains of NFM and NFH form the side arms sticking out of the core (please also refer to section 1.1.3).

16 1.2.2 Assembly properties of neurofilament polypeptides

Assembly studies using purified neurofilament subunits from bovine, porcine and murine have shown that only NFL can form 10-nm homopolymer filaments in vitro (Geisler and Weber, 1981; Moon et al, 1981; Liem and Hutchison, 1982; Zackroff et al., 1982; Minami et al., 1984; Gardner et al., 1984; Lifsics and Williams, 1984; Hisanaga and Hirokawa, 1988; Balin and Lee, 1991; Heins et a., 1993). NFM can only assemble into 10-nm filaments by itself under limited conditions (highly unphysiological conditions), while NFH cannot under any conditions. Both NFM and NFH can be incorporated into filaments in the presence of NFL in vitro. But when rodent neurofilament triplet proteins were transiently expressed individually in a non-neuronal cell line (e.g. human adrenal carcinoma SW13 Vim- cell line) lacking cytoplasmic intermediate filaments, none of them was capable of homopolymeric assembly into filamentous arrays in vivo (Gill et al., 1990; Wong et al., 1990; Ching and Liem, 1993; Lee et al., 1993). NFL can co-assemble with either NFM or NFH into filamentous arrays in vivo while NFM and NFH cannot form filaments with each other. These data collectively suggest that the neurofilaments in vivo are obligate heteropolymers requiring NF- L and either NF-M or NF-H, at least in rodents.

Unlike rodent NFL, human NFL has been demonstrated to be capable of self- assembling into 10nm homopolymers in vivo (Carter, et al., 1998). Interestingly, a single amino acid change in the rod domain (e.g. Arg162 substituted for Gln162 in the rod domain) could make rat NFL capable of homopolymerization in vivo and the reciprocal substitution in human NFL could convert it into a protein (Gln161 substituted for Arg161) that was no longer able to self assemble (Carter, et al., 1998). It has been suggested the charge interaction provided by Arg161 may stabilize the coiled-coil associations of human NFL and thus facilitate its homopolymerization in vivo. In addition, rat NFL contains an extra serine (Ser252) in L12 linker region that bisects the first coil in the α helical rod domain of intermediate filaments (see Figure 1.1 in section 1.1.1) and appears to interfere 17 the dimerization and thus the homopolymerization in vivo (Carter, et al., 1998). Biological significance of human NFL's ability to homodimerize has not been explored, but these studies clearly indicate that subtle alterations in the rod domain or its proximal regions (e.g. the L12 linker) of an intermediate filament protein can drastically alter its assembly properties.

In addition to neurofilament triplet proteins, at least two other intermediate filament proteins, α-internexin (Pachter and Liem 1985; Fliegner et al., 1990; Kaplan et al., 1990) and peripherin (Portier et al., 1983; Escurat et al., 1988; Leonard et al., 1988; Parysek and Goldman 1988; Gorham et al., 1990) are also expressed in subpopulaions of differentiated neurons (see details in section 1.3.2). α-internexin has been shown to self-assemble into 10-nm filaments both in vitro and in vivo. When α-internexin was co-expressed with each of the neurofilament triplet proteins separately in a human adrenal carcinoma SW13 Vim- cell line lacking cytoplasmic intermediate filaments, they all formed typical intermediate filament networks, which suggests that α-internexin is capable of interacting with each of them to form heteropolymeric filaments (Ching and Liem, 1993).

Peripherin and neurofilament triplet proteins have been found to co-localize in sub-population of the filaments in pheochromocytoma PC12 cells and sciatic nerves by both light and immunoelectron microscopy (Parysek et al., 1991). In addition, there were two other subpopulations of filaments composed of only peripherin or only neurofilament triplet proteins. Later on, Parysek and her colleagues further demonstrated in SW13 Vim- cells that peripherin can indeed form homopolymeric filaments in vivo (Cui et al., 1998). The functional significance of the distinct populations of filaments composed of different polypetides in PC12 cells and sciatic nerves is still not clear.

18 The capability of α-internexin and peripherin to co-polymerize with neurofilament triplet proteins (Table 1.1) suggests that neurofilament heteropolymers can apparently accommodate quite a few different polypeptides and possibly a wide range of polypeptide ratios. This could be significant because the expression of neurofilament proteins and the polypeptide ratios do change markedly during neuronal development (Nixon and Shea 1992, see details in section 1.3.1). Neurofilaments composed of different polypeptides at different stages have been suggested to be important for neuron maturation including cell migration, neurite outgrowth and elongation, neural network establishment and finally axonal radial growth.

α-internexin Peripherin NFL NFM NFH in-vivo in-vitro in-vivo in-vitro in-vivo in-vitro in-vivo in-vitro in-vivo in-vitro α-internexin √ √ ? ? √ ? √ ? √ ? Peripherin ? ? √ √ ? ? ? ? ? ? NFL √ ? ? ? Χ √ √ √ √ √ NFM √ ? ? ? √ √ Χ Χ Χ Χ NFH √ ? ? ? √ √ Χ Χ Χ Χ

√ =polypeptides are able to co-polymerize with each other Χ =polypeptides are unable to co-polymerize with each other ? =no studies are available

Table 1.1 Assembly properties of rodent and bovine neurofilament polypeptides. Data composed from the literature (see text for citations).

1.2.3 Molecular mechanism of neurofilament assembly

How the neurofilament polypeptides interact with each other and eventually form the 10nm neurofilament is still not clear. Quite a few studies have tried to identify 19 the basic units for neurofilament assembly and address the details of the assembly process. Cross-linking analysis of native or partially disassembled neurofilament proteins revealed that each triplet polypeptide was capable of forming homodimers and that an NFM/NFH heterodimer was also possible, but NFL homodimers and NFL/NFM heterodimers were the most prominent species (Carden and Eagles, 1983; Carden and Eagles, 1986). The interactions between the rod domains of neurofilament triplet proteins have been examined using the yeast two-hybrid system. The result suggest that NFM and NFH are more likely to form heterodimers with NFL than homodimers or heterodimers with each other, which is consistent with the inability of NFM and NFH to self-assemble in vitro and in vivo (Carpenter and Wallace, 1996; Leung and Liem, 1996). Full length NFL had only a relative weak interaction with itself comparing to NFM and NFH.

Noveen and his colleagues demonstrated by nondenaturing polyacrylamide gel electrophoresis (“native gel”) the formation of two soluble heterotetramers, one containing NFL and NFM and the other containing NFL and NFH, but no heterotetramers containing NFM and NFH (Cohlberg et al., 1995). A more recent study from the Mushynski lab combined a novel “blue” native polyacrylamide gel electrophoresis with Western blot analysis, disulfide cross-linking and SDS polyacrylamide gel electrophoresis in the investigation of protein dimer formation involving the neurofilament triplet proteins, α-internexin and peripherin (Athlan and Mushynski, 1997). The study showed that NFL forms heterodimers with NFM or NFH and that both α-internexin and peripherin can form heterodimers with all neurofilament triplet proteins. Whether the requirement for heteropolymerization occurs at the dimer, tetramer, or higher-order oligomer level has not yet been determined, but all the observations strongly indicate that hetrodimeric subunits are the preferred building blocks for neurofilament assembly.

20 1.2.4 Roles of different regions of neurofilament proteins in the assembly

To examine the roles of different parts of the neurofilament polypeptides in neurofilament assembly, Cleveland and his colleagues generated a variety of truncated forms of NFL and NFM with an epitope tag fused to the C-terminus, expressed them in mouse fibroblast L cell lines, which has endogenous intermediate filament network formed by vimentin, and used a tag antibody to trace the fate of the different constructs (Gill et al, 1990; Wong and Cleveland, 1990). They found that: (1) the minimal NFL subunit that efficiently coassembled without disrupting endogenous intermediate arrays contained not only the rod domain, but also 75% of the head and nearly 50% of the tail (Gill et al., 1990). The loss of more than 30% of the head or 90% of the tail of NFL resulted in assembly incompetence; (2) the minimal NFM subunit that retained assembly competence included the rod domain, 30% of the head and 10% of the tail domain (Wong and Cleveland, 1990).

To further address the role of the end domains of neurofilament subunits in the assembly, recombinant headless, tailless and rod NFL fragments were generated and their assembly properties were explored (Heins et al., 1993). This study revealed that the amino terminal head domain promotes lateral association of protofilaments into protofibrils and ultimately 10nm neurofilaments, whereas the tails control this lateral assembly so that it terminates at the level of 10nm filaments. This is further supported by the report that NFL with a partial deletion of the tail domain assembles more quickly than full length NFL in vitro even under sub-optimal buffer conditions (Nakamura et al., 1993).

1.2.5 Role of phosphorylation in neurofilament assembly

Several studies have suggested that N-terminal head phosphorylation of NFL and NFM by PKA inhibits assembly into a heteropolymer in vitro and in vivo

21 (Nakamura et al., 1990; Hisanaga et al., 1990a; Hisanaga et al., 1990b; Hisanaga and Hirokawa, 1990; Streifel et al., 1996). Nixon and his colleagues mapped the major PKA phosphorylation site at Ser55 in NFL (Sihag and Nixon, 1991) and they have proposed that Ser55 phosphorylation regulates neurofilament assembly (Nixon, 1992).

The role of transient phosphorylation of NFL ser55 in neurofilament assembly was further studied in human adrenal carcinoma SW13 Vim+ cell line that has an endogenous vimentin intermediate filament network and an SW13 Vim- cell line that doesn’t have endogenous cytoplasmic intermediate filaments. Both cell lines were transfected with wild type human NFL or human NFL ser55 mutants, in which serine 55 was mutated to alanine to prevent phosphorylation, or to aspartate to mimic phosphorylation (Gibb et al., 1996). In SW13 Vim+ cells, both NFL ser55 mutants could incorporate into existing filament array in a manner quite similar to wild type NFL. In SW13 Vim- cells, wild type NFL and NFL alanine mutant transfection gave rise to a similar filament architecture, while the introduction of NFL aspartate mutant produced aggregates in the cytoplasm (Gibb et al., 1996). Though the mechanism of neurofilament assembly is not well understood, these data suggested that phosphorylation at ser55 of NFL might affect the very early protein-protein interaction during filament assembly.

To investigate the in vivo role of NFL Ser55 phosphorylation by PKA, a subsequent study produced a transgenic mouse with human NFL-ser55 aspartate mutant (mimicking permanent phosphorylation). These mice developed pathological accumulations of neurofilament aggregates in brain neuronal cells bodies within 4 weeks after birth (Gibb et al., 1998) and these inclusions were immunoreactive with human NFL antibody, but weakly with mouse NFM or NFH antibodies. Because of technical reasons, the level of mouse of NFL in these inclusions was not determined. Why this NFL mutant failed to incorporate into the endogenous neurofilament array in these mice like in SW13 Vim+ cells and

22 instead they formed aggregates even at low expression level (8% of total NFL in the mouse) is not clear. Nevertheless, the data do implicate the important role of this phosphorylation site in neurofilament assembly.

More recent studies from the Mushynski lab show that activation of PKA in dorsal root ganglion neurons treated with nanomolar concentrations of okadaic acid potentiated neurofilament solubilization and increased the phosphorylation of NFL at Ser2 (Sacher et al., 1992; Sacher et al., 1994; Giasson et al.,1996). Thus they suggest the phosphorylation of NFL at Ser2 might also be involved in the modulation of neurofilament assembly and dynamics. PKC and p34cdc2 phosphorylation sites in neurofilament proteins have also been reported (Gonda et al., 1990; Guan et al., 1992). Apparently, phosphorylation by p34cdc2 had no effect on neurofilament assembly, whereas PKC phosphorylation of non- filamentous NFL prevented the assembly in vitro and PKC phosphorylation of assembled filaments resulted in slow assembly (Gonda et al, 1990).

How the phosphorylation of the amino-terminal domain regulates neurofilament assembly is still unknown. It might act as a modulator of protein conformation and affect the lateral interaction of neurofilament proteins, thereby interfering with neurofilament assembly. It has been proposed that neurofilament phosphorylation is topographically regulated by compartment-specific macromolecular “protein machines” (see details in section 1.4.5) and that neurofilament assembly may also be compartment specific. This mechanism might help ensure that neurofilament assembly only happens in an environment where each subunit is in the proper conformation and at the right molecular stoichiometry.

23 1.3 Neurofilament protein expression

1.3.1 Neurofilament protein expression in developing nervous system

The polypeptide composition of neurofilaments varies temporally and spatially in the nervous system. Different subtypes of neuronal intermediate filament proteins appear and disappear sequentially during neuronal development and this is presumed to meet the changing demands for plasticity and stability of cell shape as neurons migrate, elaborate neurites, and establish a permanent fiber trajectory with desired calibers (Figure 1.5) (Nixon and Shea, 1992). For example, vimentin and nestin are mainly expressed during embryonic development in neuroectoderm cells and they may be important for neuronal migration and differentiation. α-internexin and peripherin are expressed more widely and heavily during early embryonic development, which suggests that they might be import for neurite outgrowth and extension, while expression of neurofilament triplet protein increases gradually during embryonic development and continues in the adults. The neurofilament triplet proteins are the most abundant proteins in most neurons in adults and they are important in axonal radial growth and stability.

Neural stem cells are derived from the outmost layer of embryo, the neuroectoderm. These stem cells contain both vimentin and nestin (Schnitzer et al., 1981; Tapscott et al., 1981; Hockfield and Mckay, 1985; Lendahl et al., 1990). As these neuron precursor cells migrate away to differentiate into neurons, expression of vimentin and nestin are gradually down-regulated. It has been suggested that vimentin plays a role in the initial stage of neuritogenesis (Boyneet al., 1996; Dubey et al., 2004) and nestin may help maintain the long processes of the radial glia (Nixon and Shea, 1992). In the mature nervous system, vimentin is most prominently expressed in glial cells (Menet et al., 2003)

Figure 1.5 Generalized developmental profile of intermediate filaments that express in neurons. The diagram is adapted from Nixon and Shea (1992).

25 and it is also found in a few unusual neurons, such as horizontal neurons in the mouse retina and cells in the rat olfactory epithelium (Drager, 1983; Schwob et al., 1986). Nestin is virtually absent from the mature nervous system except for some potential stem cells in the subventricular zone (Doetsch et al., 1997).

The expression of α-internexin has been shown to coincide with the onset of neuronal differentiation and it may be expressed in combination with vimentin or alone (Kaplan et al., 1990; Fliegner at al., 1994; Chien and Liem, 1995; Chien et al., 1996). At embryonic day 12 (E12) of the rat, α-internexin is already found at very high levels in cells not yet expressing NFL in detectable quantities (Kaplan et al., 1990). Similar results are found at E16. The distribution of α-internexin in the embryo is far more extensive than in adults and its distribution pattern given by immunohistochemical methods is similar to that of NFL in adults.

Some early studies have shown that neurofilament proteins can be detected at E12 in rat embryos using antisera against whole neurofilaments (Raju et al., 1981; Bignami et al., 1982). Since in those studies the antisera recognize all three neurofilament protein subunits, they could not address the question of which subunits were expressed. Subsequently quite a few studies using antibodies specific for each subunit demonstrated that NFL and NFM are co- expressed earlier than NFH in the rat during development (Shaw and Weber, 1982; Shaw and Weber, 1983; Pachter and Liem, 1984; Harry at al., 1985; Nona et al., 1985; Carden et al., 1987). NFL and NFM were first detected at embryonic day 12 (E12) and by E13 the coexpression of NFL and NFM was widespread by immunohistochemistry using NFL and NFM specific antibodies (Carden et al., 1987). Unequivocal staining of NFH first appeared at E15, when NFL and NFM had already attained their adult distribution pattern and most neurons had developed axons. For several weeks after birth, the expression level of NFH increased very slowly and remained well below those of NFL and NFM (Carden et al., 1987). This is consistent with studies on rat brain and optic nerve, which

26 have suggested that NFH is expressed in bulk only after birth (Shaw and Weber, 1983; Pachter and Liem, 1984). Since the delayed expression of NFH coincides with the stabilization of neuronal circuits, it has been speculated that NFH expression might be correlated with the formation of synapses (Shaw, 1998). The delayed expression and subsequent phosphorylation of NFH could also be involved in modulating axonal transport of neurofilament proteins and the radial growth of axons (Willard and Simon, 1993; Sanchez et al., 2000).

In contrast to the model of sequential expression of neurofilament protein subunits during development, one study examined the earliest appearance of neurofilaments in mouse neuroepithelium using antibodies specific for each subunit and concluded that all three neurofilament subunits appear early and simultaneously at E9 or E10 (Cochard and Paulin, 1984). In the same study, neurofilament proteins were found to coexist with vimentin in some cells for a short period of time. The different results from studies on rat and mouse might reflect specie variations or asynchronous programming of neuronal differentiation (Altman and Bayer, 1982; Altman and Bayer, 1984; Altman and Bayer, 1986; Carden et al., 1987).

Shortly after neurofilament proteins appear in the nervous system during development, peripherin starts to be expressed, first co-localizing with NFL (Escurat et al., 1990). Soon neurofilament protein expression spreads to every part of the nervous system, but peripherin remains localized to the periphery: motor neurons in the spinal cord, sensory neurons and neurons of the autonomic nervous system. Compared to adults, peripherin is more widely and heavily expressed early in development. The expression of peripherin increases after axotomy, while the expression of neurofilament triplet proteins decreases (Oblinger et al., 1989). Depletion of peripherin in PC12 cells mediated by

27 peripherin-siRNA inhibits the initiation, extension, and maintenance of neurites (Helfand et al., 2003b). These observations implicate a principal role of peripherin in facilitating axonal outgrowth.

1.3.2 Neurofilament protein expression in adults

In adults, different types of neurons express different combinations of neurofilament proteins, presumably due to various functional demands in these neurons. For example, in the adult brain, large neurons with myelinated axons like pyramidal neurons in the cerebrum and Purkinje cells in the cerebellum express predominantly neurofilament triplet proteins (NFL, NFM and NFH), which form neurofilaments with long extended sidearms and contribute to axonal caliber, while small neurons like cerebral interneurons and cerebellar granule cells have much less neurofilament triplet protein (Yen and Fields, 1981; Shaw et al., 1981; Trojanowski et al., 1985; Vickers and Costa et al., 1992).

In the sensory ganglia, the neurofilament triplet proteins expression follows a similar pattern: neurons responsible for proprioception have thick myelinated axons with a high expression level of neurofilament triplet proteins and the nerve impulse conduction is fast, while neurons responsible for pain and temperature have thin axons with a low expression level of neurofilament triplet proteins and the nerve impulse conduction is slow (Lawson et al., 1984; Lawson and Waddell, 1991). All the motor neurons in spinal cord contain neurofilament triplet proteins (Brody et al., 1989). The retina, auditory system and the enteric system contains a mix of cell types, among which some express neurofilament triplet proteins strongly and some don’t. Neurons that express abundant neurofilament triplet proteins tend to have large myelinated axons and those that express few neurofilament triplet proteins tend to have thin axons. NFL and NFM expression patterns are very similar to each other in adult animals except in a few quite unusual situations in which high levels of NFL and unusual low levels of NFM

28 have been described and vice versa (Shaw et al., 1981; Granger and Lazarides, 1983; Trojanowski et al., 1986; Kelly et al., 1992; Harris et al., 1993; Chien and Liem, 1995; Athlan et al., 1997).

α-internexin is predominantly expressed in the central nervous system. Not only is it found in abundance in large neurons with long projection axons, but also it is present in many small interneurons in the cerebrum and granule cells in the cerebellum, which express little or no neurofilament triplet proteins (Chiu et al., 1989; Kaplan et al., 1990; Chien et al., 1996). In the periphery, some neurons from sympathetic and enteric ganglia express α-internexin along with neurofilament triplet proteins or peripherin (Vickers et al., 1992; Eaker et al, 1994). These proteins presumably form mixed heteropolymers in these neurons.

Peripherin, as its name suggests, is generally found in the peripheral nervous system, such as sympathetic, parasympathetic and dorsal root ganglia (Ferri et al., 1990; Vickers et al., 1991). But it is also found in the central nervous system, such as in some motor neurons in the ventral horn of the spinal cord and the cranial nerve nuclei with sensory components (V, VI,VIII and IX) (Broday et al., 1989; Greene, 1989). In the cerebellum, a sub-population of climbing fibers and mossy fibers does not express any other neurofilament proteins except peripherin and thus presumably only peripherin homopolymers can be found in these processes. (Errante et al., 1998). Some interneurons in the cerebrum are also found to contain peripherin (Brody et al., 1989).

1.3.3 Neurofilament protein expression in cultured neurons

Several studies on cultured neurons have demonstrated that they resemble their in vivo counterparts in certain aspects of their developmental expression of neurofilament proteins (Lee, 1985; Foster et al., 1987; Benson et al., 1996; Athlan et al., 1997). For example, Lee and her colleagues found that NFL and 29 NFM could be detected from day 1 in cultured sympathetic neurons from superior cervical ganglia (SCG), E13 spinal cord neurons and also dorsal root ganglion (DRG) neurons (Lee, 1985; Foster et al., 1987). NFH started to appear at day 3 in DRG neurons, day 9 in spinal cord neurons and day 11 in sympathetic neurons. The diversity in the appearance of NFH in different cultures may have reflected the dependency of NFH expression on neuronal type (Foster et al., 1987).

Similar neurofilament expression patterns have been observed in studies on cultured hippocampal neurons (Benson et al., 1995). In addition to neurofilament triplet proteins, two addition neuronal intermediate filament proteins, α-internexin and peripherin, were also investigated in these studies. Immediately after plating, α-internexin immunoreactive filaments were evident in these neurons. Short NFL and NFM labeled filaments appeared as early as 1 day in culture, and by 3 days in culture all neurofilament triplet proteins were detected by immunofluorescence microscopy (Benson et al., 1995). Peripherin was not detected in these neurons.

A more recent study using cultured E15 DRG neurons (Athlan et al., 1997) produced results that were a little bit different from the earlier study with DRG neurons. Before and immediately after trypsinization, E15 DRG neurons mainly contained α-internexin. One or two days after plating, the peripherin expression level starts to increase. NFL and NFM level in these neurons was very low until two days after plating while the expression of NFH was delayed until three days later (day 5 in culture). The expression levels for all five polypeptides continued to increase until seventeen days after plating, a time point after which the change became very gradual. The low expression level of NFL and NFM in these young cultures contradicted the reports that NFL and NFM are already heavily expressed in freshly cultured DRG neurons from E13 mouse (Foster et al., 1987)

30 and E16.5 rat embryos (Brown, 1997; Brown, 1998). This contradiction might be due to the fact that the organisms and/or the age of the embryo, as well as analysis methods used, were different in these studies.

The expression pattern of neurofilament proteins in cultured neurons is quite similar to developing neurons and presumably reflects the distinct roles that neurofilaments may play during axon outgrowth, extension, radial axonal growth and synapse formation in both culture and in vivo. The overlapping expression of α-internexin or/and peripherin with neurofilament triplet proteins in cultured neurons suggest that neurofilaments are highly integrated structures and their subunit stoichiometry is highly variable. This could imply a considerable degree of plasticity to neurofilament polymers, including presumably dynamic subunit change.

1.4 Neurofilament protein modification

1.4.1 Neurofilament phosphorylation

Neurofilament proteins are among those highly phosphorylated proteins. Both the head and tail domain of the neurofilament triplet proteins are modified in vivo by phosphorylation, which is mediated by multiple and distinct classes of protein kinase (Figure 1.6). The head domain contains sites that may be phosphorylated by protein kinase A (PKA), protein kinase C (PKC) and calcium/calmodulin- dependent kinases, and phosphorylation of these sites has been implicated in the regulation of neurofilament assembly (details see section 1.2.5). The tail domain contains the glutamate (E) rich region (“E segment”) that can be phosphorylated by casein kinase I (CKI) and KSP consensus sequences (in the case of NFM and NFH) for proline-directed kinases such as cyclin-dependent kinases including Cdk5 and mitogen-activated protein (MAP) kinases.

Figure 1.6 Schematic diagram of multiple sites on neurofilament proteins (NF-L, NF-M and NF-H) that can potentially be phosphorylated by different kinases. The head domains of neurofilaments proteins, which are involved in the regulation of neurofilament assembly, are mainly phosphorylated by PKA, PKC and Ca2+/CaM. CK1 has been implicated in phosphorylation of the glutamate rich “E” segment in the tail domain. The multiple KSP repeats motifs of NFM and NFH are phosphorylated by proline-directed kinases, including Cdk5, MAPK and SAPK. The figure is adapted from Grant and Pant (2000).

Phosphorylation of the tail domains of NFM and NFH may be involved in the regulation of axonal radial growth (see details in section 1.1.4) and axonal transport of neurofilament proteins (see details in section 1.5.5).

Neurofilament phosphorylation in the tail domains of NFM and NFH varies both spatially and temporally in a well defined manner. During development, poorly phosphorylated neurofilaments always preceded more heavily phosphorylated forms as revealed by immunohistochemical studies using antibodies specific for phosphorylated and non-phosphorylated epitopes (Carden et al., 1987; Foster et al., 1987). At all stages of development, poorly phosphorylated neurofilaments are contained in the cell bodies, dendrites and proximal regions of axons while more extensively phosphorylated ones are found in axons (Sternberger and Sternberger, 1983; Carden et al., 1987; Lee et al., 1987). In addition, a proximal-

32 to-distal gradient of increasing neurofilament phosphorylation along axons both in culture and in vivo has been reported by several laboratories (Szaro et al., 1989; Archer at al., 1994; Bennett et al., 1994; Benson et al., 1996). Furthermore, more recently it has been reported that neurofilaments in cultured dorsal root ganglion (DRG) neurons can be composed of contiguous phosphorylated and non-phosphorylated epitope domains that are distributed non-randomly and that phosphorylated and non-phosphorylated segments of neurofilament polymer can co-exist side by side (Brown, 1997). Thus the gradient of increasing phosphorylation along these axons appears to reflect an increase in the proportions of the total neurofilament polymer that is phosphorylated rather than a uniform and continuous increase in phosphorylation across all polymers (Brown, 1997).

Although the temporal changes in neurofilament tail domain phosphorylation during neuronal development have been correlated with axonal radial growth, little is know about the functional significance of the spatial gradient of increasing phosphorylation along axons. Since phosphorylation of neurofilaments has been implicated in regulation of neurofilament transport (see details in section 1.5.5), it will be intriguing to determine whether there is also a proximal to distal gradient in the rate of neurofilament transport. The non-randomly distributed phosphoepitopes on the neurofilaments could reflect the fact that not every phosphorylation site is equally accessible due to different conformations of each protein in the assembled polymers, or it could be the result of sequential phosphorylation by synergistically interacting kinases that are associated with neurofilaments. These kinases could be activated by some spatial cues, such as myelin-associated glycoprotein (MAG), which has been demonstrated to induce a local increase in neurofilament phosphorylation in axons (de Waegh et al., 1992).

33 1.4.2 Neurofilament associated non-proline directed kinases

Protein kinases catalyze the addition of phosphate groups onto proteins. Most kinases act on both Serine (Ser) and Threonine (Thr), others act on Tyrosine (Tyr), and a number (dual specificity kinases) act on all three. Tyr phosphorylation of neurofilaments has not been reported. Neurofilament associated Ser/Thr kinases can be further divided into two groups: non-proline directed kinases and proline directed kinases.

PKA is one of the most important second messenger dependent non-proline directed kinases for neurofilament proteins. In vitro and in vivo, PKA can phosphorylate multiple sites in the head domains of NFL and NFM (Cleverley et al., 1998). Among the sites are NFL ser55 and NFM ser46 (Sihag and Nixon, 1990; Sihag and Nixon, 1991; Cleverley et al., 1998; Nakamura et al., 2000), which are involved in the regulation of neurofilament assembly (Grant and Pant, 2000, see details in section 1.2.5). PKC and CAMKII are two other kinases, which preferentially phosphorylate the head domains of neurofilament triplet proteins, and they may also regulate neurofilament assembly (Grant and Pant, 2000).

Casein Kinases (CK), which were initially found using the milk protein casein as a substrate, have a very large number known substrates. There are two families: CKI and CKII. Both of them have been found firmly bound to neurofilament preparations from mammalian and squid axons (Dosemeci et al., 1990; Floyd et al., 1991; Link et al., 1992; Hollander et al., 1996). In vitro, CKI phosphorylates Ser/Thr residues in the glutamate (E) rich region (“E segment”) of the carboxyl- terminal tail domain of neurofilament proteins (Sihag and Nixon, 1989; Dosemeci et al., 1990; Floyd et al., 1991; Link et al., 1992; Hollander et al., 1996; Bennett and Quintana, 1997). Phosphorylation at these sites may be involved in the regulation of the dynamic interchange between neurofilament oligomers and

34 polymers (Grant and Pant, 2000). Compared to CKI, CKII appears less active towards neurofilaments but in vitro it phosphorylates NFL at ser437 (Nakamura et al., 1999).

1.4.3 Neurofilament associated proline directed kinases

Three families of proline directed kinases have been identified to be the principal kinases that phosphorylate the multiple KSP sites in the tail domain of NFM and NFH: (1) the cyclin-dependent kinases including Cdk5, (2) the mitogen-activated protein (MAP) kinases including extracellular signal- regulated kinases 1/2 (Erk1/2), stress-activated protein kinase (SAPK) and P38 kinase, and (3) glycogen synthase kinase 3 (GSK3) (Roder et al., 1993; Giasson and Mushynski, 1996; Guidato et al., 1996; Sun et al., 1996; Giasson and Mushynski, 1997; Bajaj et al., 1999; Li et al., 1999a; Li et al., 1999b; Veeranna et al., 1998; Sharma et al., 1999; Brownlees et al., 2000). In contrast to the sites phosphorylated by non- proline directed kinases (see previous section for details), these KSP sites are usually extensively phosphorylated late in the development, during axonal radial growth and myelination.

Cdk5 activity has been implicated in diverse nervous system functions such as cellular adhesion, synaptic activity and maintenance of the neuronal cytoskeleton. This kinase preferentially phosphorylates KSPXK sites, in which X is a basic or neutral amino acid (Sun et al., 1996), and produces the RT97 phosphoepitope that is thought to be involved in the regulation of neurofilament transport (see details in section 1.5.5). In vitro phosphorylation of neurofilament proteins by Cdk5 has been demonstrated in several cell lines including COS and SW13 cell lines (Guidato et al., 1996; Sun et al., 1996). p35, an activator of Cdk5, has also been shown to bind at a specific site in the carboxyl terminal region of the rod domain of human NFM using the yeast two-hybrid system, which suggests that p35 may act as a substrate-targeting protein for Cdk5 (Qi et

35 al., 1998). In the nervous system, Cdk5 mediated neurofilament phosphorylation can be activated by integrin-mediated contact interactions during axonal outgrowth (Li et al, 2000) or by myelin-associated glycoprotein during myelination (see details in section 1.1.4) (Dashiell et al., 2002).

In vitro, it has been shown that Erk1/2 phosphorylates all KSP sites on neurofilament proteins, though it phosphorylates KSPXXK and KSPXXXK sites more efficiently (Veeranna et al., 1998). This suggests that Erk1/2 may play a dominant role in the phosphorylation of rodent neurofilament proteins, which are enriched in KSPXXXK motifs (around 80% of KSP repeats in the tail domain of rat NFH) (Grant and Pant, 2000). In vivo Erk1/2 activity has been implicated in the phosphorylation of neurofilament proteins and the neurite outgrowth in both hippocampal neurons and PC12 cells (Traverse et al., 1992; Pang et al., 1995; Veeranna et al., 1998).

SAPK is a major stress kinase in the nervous system and it has been suggested to be the key perikaryal kinase responsible for NFH tail domain phosphorylation under stress conditions (Grant and Pant, 2000). Stress factors (osmotic and UV) that activate SAPK gamma in PC12 cells also cause hyperphosphorylation of NFH in cultured dorsal root ganglion neurons (Giasson and Mushynski, 1996). In vitro, SAPK gamma also phosphorylates the tail domain of NFH and the phosphorylation occurs preferentially at KSPXE rather then KSPXK motifs (Giasson and Mushynski, 1996).

Another important kinase implicated in neurofilament phosphorylation is GSK3, which has two closely related forms called GSK3α and GSK3β. In vitro, GSK3 has been shown to phosphorylate native neurofilament triplet proteins purified from cows, with NFM being the best substrate (Guan et al., 1991). Interestingly, enzymatically dephosphorylated NFM was phosphorylated only half as much as native NFM by GSK3 (Guan et al., 1991), which suggested that prior

36 phosphorylation of neurofilaments by some other kinases might facilitate GSK3 phosphorylation. Similarly, phosphorylation of tau by GSK3 is potentiated by synergistic prior phosphorylation by non-proline directed kinases including PKA, PKC, CK1 and by Cdk5 (Singh et al., 1995; Sengupta et al., 1997). Recently, it has been shown that GSK3β phosphorylates rat NFH at Ser493 in the glutamate (E) segment both in vivo and in vitro (Sasaki et al., 2002). Secondary structure prediction suggests that phosphorylation of Ser493, which is adjacent to a proline residue, interrupts an α-helix in the glutamate rich “E” segment, which suggest that this phosphorylation site might play a role in sidearm structure (Sasaki et al., 2002).

1.4.4 Neurofilament associated phosphatases

The phosphates on neurofilament proteins are subject to turnover during axonal transport as demonstrated by radioisotopic pulse-labeling studies and the phosphorylation state of neurofilament proteins in vivo results from a dynamic balance of kinase and phosphatase activities (Nixon and Lewis, 1986). Phosphatase activity towards neurofilaments in vivo was first demonstrated in cultured dorsal root ganglion neurons that had been treated with okadaic acid, a phosphatase inhibitor (Sacher et al., 1992). The neurofilaments were fragmented in these cells, which suggested that phosphatase activity is necessary for the maintenance of the neurofilament array. Subsequent studies suggested the involvement of phosphatase PP2A in dephosphorylation of the head domain of neurofilament proteins and preservation of neurofilament polymers in these neurons (Sacher et al., 1994; Saito et al., 1995). Since phosphorylation of the head domain of NFL inhibits neurofilament assembly, PP2A might play a role in facilitating neurofilament assembly by dephosphorylating the NFL head domain in cell bodies prior to entering the axon hillock (Pant and Veeranna, 1995). It also has been shown that PP2A is capable of dephosphorylating the KSP sites in the tail domain of NFH that are phosphorylated by Cdk5 in vitro (Veeranna et al.,

37 1995). Recently an association between PP1 and NFL has also been reported (Terry-Lorenzo et al., 2000), but there is a lack of evidence that PP1 can dephosphorylate neurofilaments in vivo or in vitro.

1.4.5 Regulation of neurofilament phosphorylation by macromolecular complexes

The fact that neurofilaments are associated with so many kinases and a couple of phosphatases raises the question how these kinases and phosphatases are organized in the neuron and how they interact with each other thereby topographically regulating neurofilament phosphorylation. The kinases responsible for neurofilament phosphorylation appear to be distributed equally in the cell body and along the axon (Giasson and Mushynski, 1998). However, studies on the regulation of phosphorylation in neurons from squid giant axons using P13suc1 affinity chromatography suggest that neuronal phosphorylation of cytoskeletal proteins is compartmentalized into active axonal and inactive cell body-specific multimeric complexes of kinases, substrates and phosphatases (Takahashi et al., 1995; Grant et al., 1999).

All these data have led researchers to propose that neurofilament phosphorylation is topographically regulated by compartment-specific macromolecular kinase complexes (“protein machines”) (Alberts, 1998). Thus low phosphorylation activity in the cell body and proximal regions of the axon may be due to the presence of kinase inhibitors/or absence of kinase activating molecules, high levels of phosphatases and or/inactive conformational states of the substrate proteins (Grant and Pant, 2000). These “protein machines” may also be under constant modulation so they can regulate the dynamics of cytoskeletal protein interactions during axonal growth, synaptogenesis and stabilization.

38 1.4.6 Neurofilament glycosylation

Many cytoplasmic proteins including neurofilament triplet proteins are modified by addition of N-acetylglucosamine monosaccharides to the hydroxyls of serine or threonine residues, a process called O-linked glycosylation (O-GlcNAc). The O-GlcNAc moiety is added by O-GlcNAc transferase and cleaved by N-acetyl-β-

D-glucosoaminidase. O-GlcNAc appears to be highly dynamic in a manner quite similar to protein phosphorylation and is often reciprocal to phosphorylation at the same or adjacent sites (Slawson and Hart, 2003). O-GlcNAc modification has been implicated in playing roles in a variety of biological processes, including transcriptional regulation, cell activation, cell cycle regulation, the regulation of phosphorylation, and multimeric protein complex assembly (Holt et al., 1987; Jackson and Tjian, 1988; Hart et al., 1989; King and Hounsell, 1989; Kearse and Hart, 1991; Luthi et al., 1991; Chou et al., 1992; Haltiwanger et al., 1992; Reason et al., 1992; Kelly et al., 1993).

Four different O-GlcNAc sites have been identified in NFL and NFM respectively (Dong et al., 1993; Dong et al., 1996). The sites in NFL are all located in the head domain (Thr21, Ser27, Ser34 and Ser48), which has been implicated in the regulation of neurofilament assembly (see details in section 1.2.5). In NFM, three sites are found in the head domain (Thr19, Ser34 and Thr48) and the fourth one is located in the tail domain (Thr431). NFH is also extensively modified by O-GlcNAc at Thr53, Ser54 and Ser56 in the head domain and somewhat surprisingly, at multiple serine residues in the KSP repeats in the tail domain (Dong et al., 1996). The exact number of KSP repeats modified by O-GlcNAc has not been established yet.

Although the functional significance of O-GlcNAc modification on neurofilaments is still not well understood, O-glycosylation in the head domain has been suggested to play an important role in neurofilament assembly, perhaps by regulating phosphorylation and modulating/mediating the early protein-protein 39 interaction of the assembly cascade (see detail in section 1.2.5) (Dong et al., 1993; Dong et al., 1996). O-glycosylation in the tail domain might be involved in the regulation of interfilament spacing (Dong et al., 1996), which is generally believed to be modulated by phosphorylation of neurofilament tail domain (see details in section 1.2.3). In support of this, the reciprocal changes in NFM phosphorylation and O-GlcNAcylation in the tail domain have been recently reported in spinal cord tissue of a rat model of Amyotrophic Lateral Sclerosis (Ludemann et al., 2005), which is characterized by misaccumulated neurofilament proteins.

1.5 Axonal transport

1.5.1 Fast and slow axonal transport

Neurons are highly polarized with four morphologically identified regions: cell body, dendrites, the axon and axon terminal. Most proteins and organelles in the axon are synthesized in the cell body and then transported to the axon by a process called axonal transport. Axonal transport was first demonstrated by Weiss and Hiscoe almost six decades ago (Weiss and Hiscoe, 1948). They constricted regenerating peripheral nerves by allowing them to grow through short segments of artery. The constriction led to a gradual accumulation of axoplasm proximal (toward the cell body) to the site of the compression and a local distention in the axon. Subsequent ultrastructural studies have shown that the axonal enlargement was caused by an accumulation of neurofilaments and membrane bound organelles at the site of constriction (Schmidt and Plurad, 1985; LeBeau et al., 1988).

Most of what we know about axonal transport comes from radioisotopic pulse labeling studies with experimental animals (Brown, 2000; Brown, 2003b), which identified two components of the transport. The fast components of axonal

40 transport represent the movements of membranous organelles at rates of 50- 400mm per day, while the slow components represent the movements of cytoskeletal and cytosolic proteins at rates of 0.3-8mm per day. Slow axonal transport can be further divided into two groups: neurofilaments, microtubules and their associated proteins move in slow component ‘a’ at average rates of roughly 0.3–3 mm per day; microfilaments, as well as many other cytosolic proteins, move in slow component ‘b’ at average rates of roughly 2–8mm per day (Black and Lasek, 1979; Brown, 2000).

The general principles underlying fast axonal transport are clearly established today. It is a microtubule and microfilament dependent movement powered by molecular motor proteins (kinesin, dyneins, and myosins). In contrast, there are still a lot of questions surrounding slow axonal transport: In what form these proteins are transported? How do they interact with the transport machinery? How is the transport regulated?

1.5.2 The subunit transport model for cytoskeletal proteins

Cytoskeletal proteins are conveyed along axons as part of the slow component of axonal transport. In the last three decades, several conflicting ideas have been proposed for the transport mechanism and most of them have focused on the form in which these protein move. Two diametrically opposite theories have been proposed explaining how cytoskeletal proteins are transported down the axon: the subunit transport model states that cytoskeletal subunits are transported and that they assemble into polymers locally along the axon, while the polymer model states that cytoskeletal polymers are assembled in the cell body and then transported along the axon (Figure 1.7) (Baas and Brown, 1997; Hirokawa et al., 1997).

Figure 1.7 The polymer and subunit models for axonal transport of cytoskeletal proteins. (a) In the polymer transport model, polymers are assembled in the cell body and then actively transported along the axon. Subunits can diffuse in the axon but they are not actively transported. (b) In the subunit transport model, the polymers are stationery. Subunits are synthesized in the cell body and then actively transported into the axon, where they are assembled into polymers. The diagram is reproduced from Bray (1997) with permission of Elsevier.

The subunit transport hypothesis was first proposed by Bamburg and his colleagues when they demonstrated that the growth of axons was arrested when drugs that inhibited microtubule polymerization were applied to the growth cone, but no effect was observed when the drugs were applied to the cell body (Bamburg et al., 1986). They concluded that the growth of the axons was dependent on microtubule assembly in the growth cone and that tubulin was transported into the distal axons in unassembled form. However later analyses revealed that the drug concentration used in this study caused substantial microtubule disassembly (Yu and Baas, 1995). Thus the inhibition of axon growth may have been due to the sensitivity of the growth cone to microtubule disassembly rather than simply a lack of sufficient microtubule polymers to support axonal growth. Axonal growth was not affected if the drug was applied at a concentration that prevented microtubule assembly but induced little disassembly (Yu and Baas, 1995). Thus microtubule assembly at the tip of the

42 axon was not necessary for growth and the studies carried out earlier by Bamburg and his colleagues do not support the subunit transport hypothesis.

Numerous laboratories have attempted to detect the movement of cytoskeletal proteins directly in axons of cultured neurons by live cell imaging of a population of proteins that were marked using photobleaching or photoactivation, but they have all failed to visualize any movement (Lim et al., 1989; Lim et al., 1990; Okabe and Hirokawa, 1990; Tareda et al., 1994; Sabry et al., 1995; Tekeda et al., 1995). While movement was detected in cultured embryonic frog neurons (Reinsch et al, 1991; Okabe et al., 1993), subsequent studies demonstrated that this was probably a stretching artifact due to the rapid growth of those neurons and that it did not represent slow axonal transport (Okabe and Hirokawa, 1992; Chang et al., 1998). The failure of these approaches to detect movement of cytoskeletal polymers has been interpreted as support for the subunit model of transport (Hirokawa et al., 1997).

In addition to the photobleaching and photoactivation approaches, researchers devised several other approaches to try to detect the movement of cytoskeletal proteins in axons. Miller and Joshi injected chemically fixed fluorescently labeled microtubule fragments into the neuron (Miller and Joshi, 1996). They didn’t observe any movement of the labeled microtubules and thus they concluded that microtubules do not move in axons. However the lack of movement could also be due to technical problems. For example, the chemically fixed microtubules might not be able to interact properly with the transport machinery, which could have led to no transport. To investigate the form in which neurofilament proteins are transported along the axon, Terada and his colleagues introduced virally encoded neurofilament M into neurons from transgenic mice lacking neurofilament proteins (Terada et al., 1996). Because NFM cannot assemble into homopolymers (see section 1.2.2), detection of the transfected protein along the axon was interpreted as support for the subunit transport model. However, this

43 interpretation didn’t distinguish active transport and diffusion. As demonstrated by other studies in which fluorescent molecules could diffuse over hundreds of microns within hours (Popov and Poo, 1992), an alternative explanation for the data could be that the unassembled NFM proteins diffused over long distances down axons.

In another study using photoactivation of fluorescent tubulin followed by immunoeletron microscopy, no labeled tubulins were found outside of the activated regions if the cells were permeabilized before fixation. In contrast, labeled tubulins were found outside of the activated region with more labels located distally in unextracted cells. Thereby the author argued that the asymmetrical distribution was due to transport rather than diffusion and these tubulin molecules were transported in the form of heterodimers or oligomers (Funakoshi et al., 1996). However, it is surprising that no microtubules with incorporated labeled tubulin were detected near the photo activated region after one hour considering that microtubules are such dynamic structures and this makes the sensitivity of their technique questionable. Thus it is questionable whether the asymmetrical distribution of tubulins was really due to active transport. A more direct demonstration of subunit transport has come from studies on tubulin in squid giant axons, in which labeled tubulin was injected into axons and the fluorescence profiles were monitored by co focal laser scanning microscopy and fluorescence correlation spectroscopy (Galbraith et al., 1999; Terada et al., 2000). The diffusion coefficient for the transported complex formed by the injected tubulins was more consistent with tubulin oligomers than microtubule polymers. Thus it was concluded that these proteins were able to move in an unpolymerized form.

44 1.5.3 The polymer transport model for cytoskeletal proteins

The polymer transport model was first proposed by Lasek and colleagues two decades ago. In support of this model, Reese and his colleagues injected fluorescent taxol-stabilized microtubules or phalloidin labeled actin filaments into squid giant axons and reported saltatory movements at a net rate approximating that previously shown for slow transport of cytoskeletal proteins (Terasaki et al., 1995). Indirect evidence supporting microtubule polymer transport has also been obtained in cultured neurons injected with biotinylated tubulin either prior to or shortly after sending out processes (Yu et al., 1996; Slaughter et al., 1997). After some period of axon growth, both labeled and unlabeled microtubules were present in the newly grown region of the axon. Since the unlabeled microtubules in the newly grown axons must have assembled prior to the injection of biotinylated tubulin, the authors concluded that they were actively transported along the axon. Observations on fluorescently labeled neurofilament proteins injected into squid giant axons have also yielded indirect evidence for the movement of neurofilament polymers (Galbraith et al., 1999).

1.5.4 Recent live cell imaging studies on slow axonal transport

Recently live-cell imaging studies on cultured neurons have revealed that the movement of cytoskeletal proteins along axons is not slow after all (Brown, 2000, Shah et al., 2002; Brown, 2003b). Studies on GFP-tagged neurofilament proteins or fluorescently labeled tubulin in cultured mammalian neurons have shown that these proteins move bidirectionally in the form of filamentous structures and that the rate of movement is rapid but the frequency of movement is low (Wang et al., 2000; Roy et al., 2000; Wang and Brown, 2001; Ackerley et al., 2003; Uchida and Brown, 2004). These studies suggest that the movement of cytoskeletal and cytosolic proteins is characterized by rapid movements

45 interrupted by prolonged pauses. The filamentous appearance of the moving structures supports the hypothesis that cytoskeletal proteins move in the form of assembled polymers.

But why was no movement observed in other photobleaching and photoactivation studies? One possible explanation is that earlier experiments were all designed with the expectation of a slow and synchronous movement for cytoskeletal proteins. For example, the photobleached regions were small, the extent of bleaching was partial and the time lapse intervals were long (typically five minutes or more). Thus It is possible that individual filaments could have moved rapidly though the photobeached regions or moved out the photoactivated regions without being detected in these studies.

Interestingly, not all live cell imaging studies on GFP-tagged neurofilament proteins have revealed the movement of filamentous structures, and this has revived interest in the subunit hypothesis. For example, some studies on the movement of GFP-tagged neuronal intermediate filament proteins in neuronal cell lines have demonstrated that these proteins move predominantly in the form of punctate and short filamentous structure at short times after neuritogenesis (Yabe et al., 1999; Yabe et al 2001; Chan et al., 2003; Helfand et al., 2003a). At later times, they were predominantly filamentous. It was suggested that the punctate structures might represent precursors of neurofilament assembly (Yabe et al., 2001). However there has been no ultra-structural study of these punctate structures to differentiate them from short filaments or membrane bound organelles containing GFP labeled proteins. Moreover, the punctate structures could have been an artifact of the high expressional level of exogenous GFP fusion proteins relative to the endogenous proteins in these studies. GFP tagged neurofilament proteins have also been observed to transport in the form of punctate structures in cultured neurons from chick dorsal root ganglia (Theiss et al., 2005). However, the images in this study were very poor and it is quite

46 possible that the fluorescent dots represent speckled incorporation along neurofilament polymers. Thus neurofilament proteins observed to be transported in the form of puncta in this study might have been parts of neurofilament polymers. In another study, tubular structures were detected to move in nerve processes from neurons containing fluorescently labeled tubulins (Ma et al., 2004), but the authors reasoned that these structures could not be microtubules because they constantly changed their shape and their fluorescence intensity was not consistent with individual microtubules. So they concluded tubulin had to be transported in a non polymeric form.

1.5.5 Neurofilament transport motors

The observations that neurofilaments move rapidly along the axons with peak velocities as high as 3µm/s suggest that slow axonal transport is generated by fast motors. Several lines of evidence suggest that microtubule motors can interact with neurofilament proteins and move them along microtubules. It seems likely that the anterograde movement of neurofilaments is powered by a kinesin and the retrograde movement by dynein/dynactin.

For example, in extruded squid axoplasm, the microtubule associated molecular motor kinesin was found to co-localize with a motile non-filamentous form of neurofilament protein (Prahlad et al., 2000). The association between conventional kinesin and neurofilament proteins has also been reported (Yabe et al., 1999; Prahlad et al., Yabe et al., 2000), though the specificity of this interaction is not yet well established. Abnormal neurofilament accumulations have been reported in KIF5A (kinesin I) knock out mice, which suggests that neurofilament transport was impaired (Xia et al., 2005). Since fast axonal transport appeared to be intact, the authors concluded that KIF5A was the motor specific for anterograde neurofilament transport (Xia et al., 2005) but neurofilament transport was not studied directly. Studies with cultured neurons

47 from KIF5A knock out mice have shown that anterograde neurofilament movements are significantly reduced, but not completely diminished, which suggest some motor other than KIF5A is also capable of powering neurofilament transport in these cultured neurons (Uchida and Brown, unpublished data). The association of peripherin with microtubules and their associated motor proteins dynein and kinesin has also been demonstrated by both immunofluorescence microscopy and platinum replica immunogold electron microscopy in PC12 cells (Helfand et al., 2003a). In another study, a sub-population of neurofilaments isolated from spinal cord was found to be densely labeled with anti-dynein antibodies by immunoelectron microscopy (Shah et al, 2000). Furthermore, in the same study, bidirectional translocation of neurofilaments was observed along microtubules in vitro and this movement could be disrupted by vanadate, EHNA, or function-blocking dynein antibodies. Studies on non-neuronal cells have suggested that the retrograde movement of vimentin is also mediated by the dynein/dynactin complex (Helfand et al., 2002). AFM (atomic force microscopy) studies of the interaction between neurofilaments, microtubules and dynein/dynactin suggested that neurofilaments can associate with base of this motor complex (Wagner et al., 2004). This association was further confirmed by yeast two-hybrid and affinity chromatography assays and it could be disrupted by antibodies directed to either NFM or to dynein (Wagner et al., 2004). Recently, the role of cytoplasmic dynein in the axonal transport of neurofilaments has also been investigated using RNA interference and anti-dynein antibody in cultured neurons (He et al., 2005; Theiss et al., 2005). In both cases, retrograde neurofilament movements were significantly diminished. Myosin Va has also been shown to associate with neurofilaments in axons but how this fits into neurofilament transport is not clear (Rao et al., 2002).

1.5.6 Regulation of neurofilament transport by phosphorylation

According to a unified perspective, Brown has proposed that both fast and slow axonal transport are using same underlying mechanism but they move at 48 different rates due to differences in their duty ratio (Brown, 2003b). The high duty ratio of the membranous organelles ensures that they are delivered to their destination rapidly. In contrast, cytoskeletal proteins move in an intermittent manner, pausing more often and for longer periods of time. What regulates the pausing and moving behaviors of these proteins is not clear.

Considerable evidence has pointed to phosphorylation as the most likely mechanism for regulating neurofilament behavior. Neurofilament proteins are highly phosphorylated proteins with numerous phosphorylation sites, organized as repeating motifs (Grant and Pant, 2000; see details in section 1.4). Most phosphorylation sites are in KSP motifs of the tail domain of NFM and NFH. A lot of studies have suggested that phosphorylation of these sites correlates with a decrease in the rate of neurofilament transport (Lewis and Nixon, 1988; Watson et al., 1989; Archer et al., 1994; Toyoshima and Komiya, 1995; Jung and Shea, 1999; Jung et al., 2000a; Jung et al., 2000b; Yabe et al., 2001; Shea et al., 2004). Hypophosphorylated neurofilament subunits were selectively recovered within a standard microtubule associated protein preparation rich in kinesin (Saxton, 1994), while NFH bearing a developmentally-delayed carboxyl terminal phosphoepitope was selectively not co-precipitated by an anti-kinesin antibody (Jung et al., 2000b). Furthermore, hypophoshorylated NFM and NFH isoforms were transported approximately twice as fast as their extensively phosphorylated counterparts as demonstrated by autoradiographic analysis following metabolic radio-labeling Jung and Shea, 1999; Jung et al., 2000a). Up-regulated neurofilament phosphorylation in the sidearm domain has also been associated with slowing of axonal transport of neurofilament protein in cultured neurons in response to glutamate. It has also been shown that glutamate can activate members of the mitogen-activated protein kinase family, which subsequently phosphorylate the neurofilament side arm domain (Ackerley et al., 2000).

49 In the optic nerve, it has been clearly shown that signaling from the myelinating oligodendrocytes triggers phosphorylation of KSP repeats in the tail domains of NFH and, to a lesser extent NFM, and this results in local accumulation of neurofilament proteins (Nixon et al., 1994; Sanchez et al., 1996; Sanchez et al., 2000). Axons of retinal ganglion cells are unmyelinated within the retina and for a distance of ~100µm after they converge and form the optic nerve at the retinal excavation. At 100-150µm from the retinal excavation, there is a specialized meshwork of astrocytes called the lamina cribrosa (Figure 1.8A). Beyond this point, 95% of the axons in the optic nerve from adult mice are myelinated, large in caliber and contain abundant neurofilaments (Figure 1.8 B and C) (Nixon et al., 1994). Under the control of signals from oligodendrocytes, the cross-sectional areas of these axons increase an additional 200% during postnatal development (Sanchez et al., 1996). >80% of this radial axonal growth occurred between 21 and 30 days postnatally (Figure 1.8 D and E) (Sanchez et al., 2000), and is associated with regional accumulation of neurofilament proteins (Figure 1.8 F and G). To investigate the relationship of NFH carboxyl terminal phosphorylation to neurofilament organization, the levels of individual NFH phosphoepitopes during axonal radial growth were analyzed and RT97 phosphoepitope expression selectively coincided with the onset and rise of regional neurofilament accumulation (Figure 1.8 H and I) (Sanchez et al., 2000). Furthermore, RT97 phosphoepitope levels are selectively reduced in myelin-deficient Shiverer mutant mice, which exhibit decreased regional neurofilament accumulation.

There are three possible ways that RT97 phosphoepitope may have been associated with regional neurofilament accumulation. Firstly, the increased expression of RT97 phosphoepitope may happen to coincide with neurofilament accumulations and it does not have any causative affects. Secondly, RT97 phosphoepitope may be an actual trigger for neurofilament accumulation by increasing the interaction between adjacent neurofilaments and slowing down

Figure 1.8 Correlation between regional neurofilament accumulation and appearance of phosphoepitope RT97 on NFH during development. (A) Toluidine blue-stained sagittal section through the retina and proximal optic nerve of the adult mouse reveals the landmarks of the retino-optic junction. (B,C) Ultrastructure of two regions of the optic nerve from 21-day-old mouse located 50µm and 700µm from the retinal excavation respectively. Note the difference between axon diameter and neurofilament content. (D) Relative average axonal cross sectional area at the 50 and 700µm levels for representative subpopulations of the total axonal population were measured from electron micrographs of the optic nerve from mice at different ages. The percent increase in axonal cross-sectional area at the 700µm level over the area at the 50µm as a function of postnatal ages is shown in (E). (F) The absolute number of neurofilament at 50 and 700µm levels of the optic nerve were determined at different ages. Note the regional accumulation of neurofilaments, which is emphasized in (G) as the percentage increase in neurofilament number at the 700µm level over that at the 50µm. Relative content of different phosphoepitopes associated with NFH in triton-insoluable cytoskeletal fractions from intraretinal axons (H) and optic nerves (B) from mice at various postnatal ages. Immunoreactivity level of each phosphoepitopes as expressed as the ration to total NFH immunoreactivity determined with SMI33 antibody. Note that no RT97 immunoreactivity was detected in retinal axons and the marked increase in RT97 immunoreactivity in the optic nerve was correlated with the regional accumulation of neurofilaments and the increase in axonal caliber around 20 days postnatally (compare panel E, G and I). On the other hand, the expression of phosphoepitopes SMI31 and SMI34 increase gradually with time. The images are reproduced from Sanchez et al. (2000) with permission of The Rockefeller University Press.

Figure 1.8

52 axonal transport of neurofilaments. Thirdly, RT97 phosphoepitope may promote stability of neurofilaments by reducing their susceptibility to protease (Goldstein et al., 1987; Pant, 1988).

To further investigate the role of phosphorylation at the RT97 phosphoepitope in the regulation of neurofilament transport, Miller and his colleagues transfected cortical neurons with GFP tagged NFH mutant mimicking permanent phosphorylation (aspartate substitution mutant) and mutant mimicking dephosphoryation (Alanine substitution mutant) at this epitope (Ackerley et al., 2003). Live cell imaging studies on these cells demonstrated that those mutants mimicking permanent phosphorylation spend twice the time pausing comparing to those mimicking dephosphorylation. Figure 1.9 shows the comparison of the behavior for these two mutants. Consistent with this data, neurofilament transport was accelerated by the application of roscovitine, which is an inhibitor of the kinase Cdk5/p35 which produces the RT97 phosphoepitope (see details in section 1.4.3). All these observations provide strong circumstantial evidence for an important role of neurofilament phosphorylation in the transport regulation machinery.

1.6 Neurofilaments and neurodegenerative diseases

Abnormal accumulations of neurofilament proteins are the pathological hallmark for several human neurodegenerative diseases, including Amyotrophic Lateral Sclerosis (ALS, better known as Lou Gehrig's Disease), giant axonal neuropathy, Charcot-Marie-Tooth disease (CMT), Alzheimer’s disease (AD), Parkinson’s disease (PD) and diabetic neuropathy (Al-Chalabi and Miller, 2003). Many factors can potentially lead to the misaccumulation of neurofilament proteins. They include up-regulation of neurofilament protein synthesis, down- regulation of neurofilament protein degradation and impaired neurofilament protein transport. Interestingly, the abnormal neurofilament inclusions are often

Figure 1.9 Motile behaviors of neurofilaments in cortical neurons. (A) Representative movement characteristics of neurofilaments labeled with GFP- NFHwt, GFP-NFHala and GFP-NFHasp as indicated respectively. The y-axis represents the movement of the filament along the axon from its initial location, and the x-axis is the time. Asterisks indicate pauses during the translocation. Note the prolonged pauses during the translocation of the GFP-NFHasp filament. (B) Summary of the movement characteristics of GFP-NFHala and GFP-NFHasp filaments. The plots and table are reproduced from Ackerley et al. (2003) with permission of The Rockefeller University Press. .

54 associated with a decrease in the levels of NFL mRNA. NFL mRNA is selectively down-regulated by up to 70% in degenerating neurons of ALS and AD (Mclachlan et al., 1988; Bergeron et al., 1994; Wong et al., 2000; Menzies et al., 2002). In ALS, there is a reduction in the α-internexin or peripherin mRNA levels (Bergeron et al., 1994; Wong et al., 2000). Whether the decrease in NFL mRNA levels cause an alteration in the subunit stoichiometry for proper neurofilament assembly and transport, and subsequently cause the neurofilament inclusion, or the decrease in the mRNA level is just a cellular response to the inclusion is still not clear. However studies with transgenic mice and the identification of the neurofilament gene mutations that are linked to several neurodegenerative diseases further suggest disorganized neurofilaments are deleterious.

1.6.1 Transgenic mice with neurofilament protein overexpression

Studies with transgenic mice have found that overexpressing two to four times the normal levels of human NFH (Cote et al., 1993; Collard et al., 1995) or wild- type mouse NFL (Xu et al., 1993) leads to a striking ALS-like pathology, in which there were abnormal perikaryal accumulations of neurofilaments accompanied by axonal atrophy and motor dysfunction, but without extensive motor neuron death (Lariviere and Julien, 2003). More recent studies using transgenic mice overexpressing wild-type peripherin (Beaulieu et al., 1999; Beaulieu et al., 2000) found that they not only developed neurofilament inclusions but also exhibited massive motor neuron death during aging. Interestingly, when the peripherin was overexpressed in a NFL null mouse, the deficiency of NFL further accelerated these pathologies (Beaulieu et al., 1999). On the other hand, the peripherin- mediated motor neuron death in these NFL knock out mice could be rescued by co-overexpression of NFH (Beaulieu and Julien, 2003). In line with these studies the in vitro experiments overexpressing peripherin in the cultured motor neurons demonstrated that detrimental effect mediated by excess peripherin could be attenuated by coexpression of NFL (Robertson et al., 2001). Transgenic mouse

55 studies with other neurofilament polypeptides, such as NFM and α-internexin, have observed less abnormal filamentous accumulations and some neurodegeneration in the case of α-internexin overexpression (Vickers et al., 1994; Wong et al., 1995; Ching et al., 1999). All these findings with transgenic overexpression mouse models suggest the importance of the neurofilament subunit protein stoichiometry for the proper functioning of neurofilaments and that mis-expression of neurofilament polypeptide can lead to neurodegenerative disease-like pathology.

1.6.2 Neurodegenerative disease transgenic mouse model

Missense mutations in the gene encoding Cu/Zn superoxide dismutase 1 (SOD1) are responsible for ~20% of the familial ALS cases (Tu et al., 1997). Abnormal neurofilament aggregates have been reported in ALS cases with SOD1 mutation (Hirano et al., 1984; Rouleau et al., 1996; Ding et al., 2002). To further investigate ALS pathogenesis mediated by SOD1 mutation, transgenic mice with SOD1 mutation in different contexts of neurofilament proteins were generated. Surprisingly, neurofilaments were not required for the pathogenesis in SOD1 mutant since the expression of NFH-β-galactosidase, which trapped neurofilaments in the perikarya with few transported into the axon, offered no beneficial effect in SOD1 mutant mice (Eyer et al., 1998). This result was further confirmed by experiments, in which the SOD mutant mice crossed with NFL null mice. The absence of NFL leads to around a 15% extension of the life span of the SOD1 mutant (Williamson et al., 1998). Paradoxically, overexpression of human NFH also increases the lifespan of the SOD1 progeny by up to 65% (Couillard-Despres et al., 1998).

Two models have been proposed to explain how altering axonal neurofilaments can alleviate the toxicity of mutant SOD1 (Lobsiger et al., 2005). The first model has proposed that the perikaryal accumulations of neurofilaments produce a

56 beneficial effect and this protective role of neurofilaments is presumably due to multiphosphorylation sites in the tail domains of NFM and NFH that could have served as phosphorylation sinks for buffering the hyperactive Cdk5 in SOD1 mutants. Evidence supporting this hypothesis primarily comes from studies on mutant SOD1 mice generated in a context of one allele for each NF gene being disrupted (Nguyen et al., 2000). There was 40% decrease in axonal neurofilament content and caliber of motor neurons, but this didn’t extend the life span of SOD1 mice nor did it alleviate the loss of motor axons. Based on the analysis of the pathological changes in ALS mice, these authors proposed that perikaryal neurofilaments can alleviate the toxicity of mutant SOD1 by sequestering p25/Cdk5 complex and by acting as a phosphorylation sink for deregulated kinase activity, thereby reducing the potential toxicity of hyperphosphorylation of tau and of other Cdk5 substrates (Nguyan et al., 2001; Nguyan et al., 2002; Nguyan et al., 2003). However in the triple heterozygous knockout mouse, 40% reduction in neurofilaments might have caused overall reduced phosphorylation of neurofilament proteins and left other Cdk5 substrates at high risk of being hyperphosphorylated, whose potential toxicity could have covered the possible beneficial effects of reduced neurofilament burdens in these mice.

In an alternative hypothesis, hyperphosphorylated neurofilaments are deleterious and the protective effect of altering neurofilament content in the earlier study resulted from depletions of neurofilament proteins from the axon, which helped restore normal axonal transport in SOD1 mutant by reducing the interfilament interaction that were mediated by phosphorylated tail domain of NFM and NFH and subsequently released the neurofilament-dependent slowing of axon transport. If this hypothesis is correct, one would expect that removing NFM or NFH tail domain would slow mutant SOD1-mediated disease (Lobsiger et al., 2005). To test this, Cleveland and his colleagues generated a SOD1 disease model in which gene replacement was used to remove the tail domains of NFM,

57 NFH or both (Lobsiger et al., 2005). It was demonstrated that removal of the large phosphorylated tail domain of NFM and NFH delays disease onset and extends survival of mutant SOD1 mice by 2 months, which the authors thought strongly supported the second model. But in this study, there was no data on axonal transport supporting the hypothesis that the protective effect of depletion of neurofilaments from axon is a result of released axonal transport. As described above, both models can explain how altering neurofilament contents alleviate the toxicity of mutant SOD1 to some extent but there are still a lot of assumptions behind these models, which need to be addressed for us to better understand the association of neurofilament with the pathogenesis in these disease models.

Disease progression is also altered in transgenic mice overexpressing human tau protein (Ishihara et al., 1999), a mouse model of human taupathies, by modulating neurofilament expression. The formation of tau inclusion was delayed when the tau transgenic mice were crossed with NFL or NFH knock out mice (Ishihara et al., 2001). These results suggested a role for neurofilaments in the pathogenesis of neurofibrillary tau lesions in neurodegenerative disorders that contain both neurofilaments and tau proteins, such as AD. All of these studies using different neurodegenerative disease transgenic mouse models demonstrated that neurofilament proteins play a very important role in the pathogenesis of neurodegenerative disease, including ALS and taupathies.

1.6.3 Neurofilament gene mutations associated with neurodegenerative diseases

Recently, quite a few neurofilament gene mutations have been linked to neurodegenerative disease, including ALS, CMT and Parkinson disease. It has been suggested that the mutations associated with neurodegenerative diseases disrupt neurofilament assembly and axonal transport, which probably underlies the disease mechanism (Al-Chalabi and Miller, 2003).

58 Codon deletions or insertions in the KSP phosphorylation domains have been detected in a small number (approximately 1%) of sporadic cases of ALS (Figlewicz et al., 1994; Tomkins et al., 1998; Al-Chalabi et al., 1999) and such mutations were absent from over 1000 control DNA samples. However, recently Cleveland and his colleagues carried out a systematic analysis of the coding region and intron-exon boundaries of all three neurofilament genes from more than 200 non-SOD1 linked familial and sporadic ALS patients, along with >400 non-disease control individuals (Garcia et al., 2005). None of the variants within each of the three neurofilament subunits that are predicted to affect neurofilament function including neurofilament assembly properties could be unambiguously linked to dominantly inherited disease. Thus, they conclude that mutations in neurofilaments are possible risk factors that may contribute to pathogenesis in ALS in conjunction with one or more additional genetic or environmental factors, but are not significant primary causes of ALS (Garcia et al., 2005).

CMT, Charcot-Marie-Tooth disease, is an inherited neuropathy, which affects both sensory and motor neurons. CMT can be easily classified into three types: type 1 and 3 are due to demyelating and type 2 is an axonal disease. Several families have been identified in which heterozygosity for mutations of the NEFL gene (coding NFL) on chromosome 8 are associated with CMT2. The first of these changes is a proline at residue 8 to glutamine (Mersinayova et al., 2000). The second change is a leucine to a proline at residue 333, which is a highly conserved position (De Jonghe et al., 2001). A transgenic mouse model with the same mutation of NFL at a similar position (leucine to proline at residue 394) developed a very severe phenotype, in which there was massive degeneration of motor neurons and mortality at about 3-weeks postnatal (Lee et al., 1994). Similar mutations, proline-to-serine and proline-to-threonine substitution, at residue 22 were found in a Slovenian and a Japanese family respectively

59 (Georgiou et al., 2002; Yoshihara et al., 2002). Recently Jordanova et al. (2003) identified six pathogenic mis-since mutations and one 3-bp in-frame deletion in the NFL gene in 323 patients with different phenotypes.

A point mutation has been reported in the region of NEFM gene in an individual with Parkinson’s disease, who developed the disease at age of 16 (Lavedan et al, 2002). There are three other unaffected families carrying the same mutation, which suggests that this mutation in combination with other factors is responsible for the young onset of aggressive disease seen in that individual.

1.6.4 Neurofilament transport and neurodegenerative diseases

The precise mechanism by which neurofilament proteins accumulate in neurodegenerative disease is still unknown. Studies on axonal transport using ALS-linked SOD1 mutants demonstrated that dramatic defects in axonal transport (Collard et al., 1995; Willianson and Cleveland, 1999), including neurofilaments, tubulin and actin, arise months before degeneration, which suggests that damage to machinery of slow axonal transport is an early pathological feature. Thus the misaccumulation of neurofilaments is an early pathogenic event that plays an important mechanistic role in the neurodegenerative process rather than being not just an end-stage epiphenomenon.

Studies using transgenic mice expressing neurofilament protein and mutations that affect neurofilament assembly suggest that proper subunit stoichiometry and subsequent neurofilament assembly might be part of the regulatory machinery for axonal transport of neurofilament proteins. Another possibility is related to the phosphorylation state of neurofilament proteins. Phosphorylation of the head domain of neurofilament proteins is involved in the regulation of neurofilament assembly and phosphorylation of the tail domain has been suggested to be the

60 key regulator of neurofilament transport. Hyperphosphorylated neurofilament accumulations are a prominent feature of some neurodegenerative diseases. Increased neurofilament phosphorylation that occurrs in these disease state probably leads to impaired transport and subsequently misaccumulation of neurofilament proteins. Deregulation of cdk5/p35, which phosphorylates the neurofilament protein carboxyl domain, has been demonstrated in a mutant SOD1 transgenic mouse model of ALS, and neurofilament protein mutations associated with ALS all happen to be in the KSP regions of NFH.

CHAPTER 2

MATERIALS AND METHODS

2.1 Cell culture

2.1.1 SCG neuron culture

Neurons dissociated from superior cervical ganglia of neonatal (P0-P1) rats or mice were cultured on No. 1.5 square glass coverslips (22 x 22mm) as described by Brown (2003a). The coverslips were coated with poly-D-lysine (Sigma, Mw 70,000-150,000; 1mg/ml) and MatrigelTM (BD Biosciences; 10µg/ml) and the neurons were plated at low density (0.006 dissociated ganglia per cm2). Cultures were maintained at 37°C in Leibovitz’s L-15 medium (Invitrogen, Phenol Red- free) supplemented with 0.6% glucose, 2mM L-glutamine, 100ng/ml 2.5S nerve growth factor (BD Biosciences), 10% adult rat serum (Harlan Bioproducts), and 0.5% hydroxypropylmethylcellulose (MethocelTM, Dow Corning) as described by Brown (2003a). In some experiments No. 1.5 square glass photoetched coverslips (25 x 25mm, Bellco Biotechnology) were used to help locate the captured filaments in the gap after the immunostaining process.

2.1.2 SW13 cell culture

Human adrenal carcinoma SW13 cl.1 Vim+ and cl.2 Vim- cells (Sarria et al., 1990) were generously provided by Dr. Robert Evans from University of Colorado and cultured on No. 1 round coverslips (diameter 15 mm) in 12-well plates

62 (Becton Dickinson) at high density (105 cells per cm2). The cultures were maintained in DMEM/F12 medium (Dulbecco’s Modified Eagle Medium/ Ham’s F- 12 Nutrient Mixture, Invitrogen) supplemented with 5% fetal bovine serum (FBS)

and 10µg/ml gentamicin at 37°C in a humidified atmosphere of 5% CO2.

2.2 Transfection

2.2.1 Plasmid DNA

The GFP-rNFM expression vector, which directs the expression of the F64L/S65T variant of green fluorescent protein (Clontech) linked to the amino terminus of rat neurofilament protein M, has been described previously (Wang et al., 2000). All mouse neurofilament vectors were constructed and provided by Kitty Jensen (Jensen and Brown, unpublished data). The GFP-mNFL expression vector, which encodes GFP-tagged mouse NFL, and the GFP-mNFM expression vector, which encodes GFP-tagged mouse NFM, were constructed in a similar way to the rat construct. The pmNFL and pmNFM expression vectors encoding mouse NFL and mouse NFM respectively were constructed by removing GFP sequence using appropriate restriction enzymes. The pSRV-α expression vector encoding rat α-internexin (Ching and Liem, 1993) and pCI-NFH encoding rat NFH (Leung et al., 1999) were provided by Dr. Ronald Liem from Columbia University. The pBluescript-peripherin vector containing the human peripherin genomic DNA under the control of a human beta actin promoter was provided by Dr. Linda Parysek from University of Cincinnati. All plasmids were amplified in Escherichia coli (DH5α) and purified using Qiagen Maxi Prep Plasmid Purification kits (Qiagen). Purity of the plasmid was checked by electrophoresis on 0.8% agarose gels and was found to be more than 80% supercoiled. Concentration of the DNA was estimated spectroscopically by measuring the absorption at 260 nm and the 260nm/280nm ratio.

63 2.2.2 Nuclear injection

Two days after plating, cultured neurons were transfected with GFP-NFM expression vector by nuclear microinjection with the use of an Eppendorf InjectMan™ NI2 micromanipulator and FemtoJet™ microinjector (Brinkman Instruments) as described by Brown (2003a). Micropipettes were pulled from standard thick-wall borosilicate glass tubing (World Precision Instruments) with the use of a P-97 Flaming-Brown pipette puller (Sutter Instrument Company). In all experiments, the plasmid was co-injected with 1.25mg/ml (Mw 10,000) tetramethylrhodamine dextran (Sigma) to allow visual confirmation of the injection procedure.

The concentration of injected plasmid was selected to give an expression level sufficient to observe GFP fluorescence in single neurofilaments but not so much as to lead to overexpression. Signs of overexpression were aggregated and diffusible GFP fluorescence in both cell body and the axon. Using these criteria, plasmid GFP-rNFM was injected into rat neurons at concentration of 600µg/ml in 50mM potassium glutamate, pH 7.0 and plasmid GFP-mNFM was injected to mouse neurons at concentration of 100µg/ml. It is hard to estimate how many plasmids were actually delivered into the nucleus for each microinjection since the penetration for each cell varied, but the concentrations used for injection of both plasmids have reliably produced a high proportion of injected cells with medium expression level of GFP fusion protein, in which there was a nice neurofilament network in the cell body observed by intrinsic GFP fluorescence. Comparably lower concentrations of GFP-mNFM were used for mouse neurons, since many of the cells overexpressed the exogenous protein if higher concentration was used (Atsuko Uchida and Nael Alami, Brown lab, personal communication).

64 2.2.3 Lipofection

The SW13 cells were transfected with plasmids using Lipofectamine 2000 (Invitrogen) according to the manufacturer’s instruction with some modifications. First plasmids and lipofectamine 2000 solution were diluted separately in 100µl DMEM/F12 medium. The dilution of lipofectamine 2000 solution used was dependent on the total amount of DNA used for each co-transfection. If there were less than three plasmids for the co-transfection, 4µl Lipofectamine in 2000 solution was diluted in 100µl DMEM/F12 medium. If there were more than three plasmids co-transfected, the amount of Lipofectamine solution was increased accordingly (2µl for each additional plasmid). 100µl diluted plasmids (10ng/µl each) and 100µl diluted Lipofectamine 2000 solution were then combined and incubated for 20 minutes at room temperature. Cells were then transfected with 200µl combined DMEM/F12 medium containing diluted Lipofectamine 2000 solution and plasmid DNA per well of 1x 105 cells for 4h at 37°C. After removing the transfection mixture, cells were maintained in fresh medium with FBS and gentamicin as described earlier for 24 hours before fixation. If the dilution for Lipofectamine 2000 was more than 50 fold (4µl in final 200µl combined medium), cells appeared to be healthy following an overnight incubation in transfection solution and it was not necessary to change the medium after 4 hours. The amount of Lipofectamine 2000 solution and DNA used for each transfection was much lower than the manufacturer recommended, because higher concentrations resulted in substantial cell death without a significant increase in transfection efficiency (Atsuko Uchida and kitty Jensen, Brown lab, personal communication).

65 2.3 Live cell imaging

2.3.1 Perfusion chamber system

For live-cell imaging, coverslips were removed from the culture dish and mounted in a RC-21B closed bath perfusion chamber (Warner Instruments) using silcone grease (Dow Corning Compound 111) as the sealant. A drop of warm imaging medium (see below) was placed on the coverslip to prevent the cells from drying out during chamber assembly. The chamber had an internal volume of 260µl and a laminar flow design to maximize the rate of solution exchange during perfusion. After assembly, the chamber was gently filled with imaging medium using a syringe that was connected to the inlet port with polyethylene tubing. Care was taken to avoid introducing bubbles into the chamber. The inlet tubing was then clamped with a mini 9/16’’ binder clip and the whole chamber was moved to the microscope stage where the inlet tubing was connected to a reservoir containing permeabilization solution. The permeabilization solution consisted of 0.02% saponin (Sigma) in a solution composed of NPHEM (0.19M NaCl, 60mM sodium PIPES, 25mM sodium HEPES, 10mM sodium EGTA, 2mM MgCl2, pH 6.9), 10µg/ml Bestatin (Sigma), 10µg/ml Leupeptin (Sigma) and 10µg/ml E64 (Sigma). In some experiments, I substituted 5mM calcium for the EGTA in the NPHEM in order to induce the disassembly of microtubules (Black et al., 1984; O’Brien et al., 1997). The perfusion solution was warmed to 35°C before it entered the chamber using an in-line SH-27B solution heater (Warner Instruments) (Figure 2.1).

In earlier experiments, the imaging medium consisted of oxygen-depleted culture medium (Wang et al., 2000) lacking Methocel™ (Methocel™ was omitted because it increased the viscosity of the medium, and thus the resistance to flow). To deplete the medium of oxygen, I added EC Oxyrase™ (Oxyrase Inc., 1:100 dilution) plus 20 mM sodium succinate and 20 mM sodium lactate and pre-

66 incubated the mixture in a sealed syringe with no air space for 1-2 hours at 37ºC. Immediately prior to use, the mixture was filtered using a 0.2µm syringe filter to eliminate aggregates of the EC Oxyrase. The purpose of depleting oxygen from the medium was to minimize photobleaching and photodamage. But I felt that there was still significant photobleaching using oxygen-depleted imaging medium. To confirm this, the cells were observed using medium with or without the depletion of oxygen. It turned out that there was no difference in photobleaching between the two different conditions (Atsuko Uchida, Brown lab, personal communication). Thus in the later experiments with mouse neurons, I skipped the oxygen depletion procedure.

2.3.2 Image acquisition

Cells in the perfusion chamber were observed by epifluorescence and phase contrast microscopy with a Nikon Quantum TE300 inverted microscope 3 days after transfection. Image acquisition was performed with the use of a MicroMax 512BFT cooled CCD camera (Roper Scientific), a Nikon 100x/1.4NA Plan Apo oil immersion objective and a FITC/EGFP filter set (Chroma Technology, HQ 41001). The temperature on the microscope stage was maintained at approximately 35°C by using a Nicholson ASI-400 Air Stream Incubator (Nevtek). For time-lapse imaging, the epifluorescent illumination was attenuated by 70- 90% using neutral density filters, and images were acquired at 4s intervals with 1s exposure using MetaMorphTM software (Universal Imaging) as described by Brown (2003a). The focus was found to drift during imaging, presumably due to either mechanical or thermal instability. To compensate for this, I made fine adjustments to the focus in between time-lapse image acquisition using a remote motorized focus attachment (Prior Scientific).

Figure 2.1 Perfusion chamber system. For live-cell imaging, cells were cultured on coverslips and mounted in an RC-21B closed bath perfusion chamber using silicone grease as the sealant. The chamber was placed on the stage of an inverted microscope. The inlet tubing (inner diameter 1/32’’) was connected to a reservoir containing permeabilization solution, which was around 20 inches above microscope stage, and the solution was warmed to 35°C before it entered the chamber using an in-line SH-27B solution heater. The mini binder clip on the inlet tubing was used to start and stop the flow. The outlet tubing (inner diameter 1/16”) was kept as short as possible (about 7 inches) to avoid siphoning.

68 2.4 Capture, fixation and attempted reactivation of moving filaments

2.4.1 Capture of moving filament

Gaps were located along the axon from cells expressing GFP-NFM and observed by time-lapse fluorescence imaging. When suitable GFP-tagged neurofilament proteins entered the gap, the flow of permeabilization solution was initiated by removing the binder clip on the inlet tubing. The flow rate was about 3.8ml/min, which was determined by the height of the solution reservoir above the stage. The flow was maintained for about one minute and stopped by replacing the clip. Upon initiation of the solution flow, I found that there was frequently a large focus shift due to thermal expansion or contraction of the chamber or chamber mounting hardware. To overcome this problem, the in–line SH-27B solution heater (Warner Instruments) was included in the system to warm the solution to 35°C before it entered the chamber. SH-27B is a rapid flow in-line solution heater and it accommodates perfusion flows up to 10 ml/min.

2.4.2 Fixation of moving filaments for immunofluorescence microscopy

The cells were fixed by perfusing 3ml 4% (w/v) formaldehyde in NPHEM into the chamber by the use of a syringe manually or under gravity at an approximate rate of 2ml/min. During the fixation, the axon of interest was marked by scoring a circle on the coverslip using a Leitz diamond-scoring object marker. Ten to fifteen minutes later, the chamber was removed from the microscope stage, submerged in phosphate buffered saline (PBS, 10 mM phosphate buffer saline, 138 mM NaCl, 2.7 mM KCl, pH 7.4, Sigma) and then disassembled (if the chamber was disassembled in air, the fixative drew completely off the coverslip and the surface tension of the liquid caused the cells to detach). As described earlier, the coverslips were mounted in the chamber using silicone grease as sealant, which

69 sealed the chamber very well but also impaired disassembly. To avoid breaking the coverslips during disassembly, two thin double edged razor blades were inserted under the coverslip (between the coverslip and the body of the chamber) at opposite corners. Then the blades were slowly and carefully inserted toward the open area of the chamber. Eventually the coverslip would loosen, at what time finely tipped forceps were used to lift up the coverslip carefully from the chamber and place it in fresh fixative for another twenty minutes at room temperature. After fixation, the dish was rinsed with PBS and then treated with PBSNT (0.1% Triton X-100 + 0.3M NaCl in PBS) for 15 min (“post-fixation extraction”) to ensure complete demembranation of the cells prior to immunostaining (Brown, 1997).

While the above fixation protocol was adequate for most of the primary antibodies used, two of the antibodies (7C5, a monoclonal antibody for peripherin, and RT97, a monoclonal antibody specific for phosphorylated neurofilament proteins) stained poorly. If filaments were only fixed in the chamber for 15 minutes without post-fixation using 2ml 2% (w/v) formaldehyde instead of 3ml 4% (w/v) formaldehyde, the staining along the filaments for these two antibodies was much stronger and similar to those fixed in open dishes using 2ml 4% (w/v) formaldehyde for twenty minutes. The possible explanation is that the laminar flow design of the perfusion chamber allows a more thorough solution exchange, which produces a much more efficient fixation in the chamber comparing to open dishes. The overfixation appeared to destroy 7C5 and RT97 epitopes, which presumably are sensitive to formaldehyde fixation. Thus for these two antibodies, captured filaments were fixed in the chamber for 15 minutes using 2ml 2% (w/v) formaldehyde as described earlier.

70 2.4.3 Fixation of moving filaments for electron microscopy

For correlative light and electron microscopy, 3ml 2% glutaraldehyde (EM grade: Polysciences) in 150mM sodium phosphate, pH 7.4 (320 mOs, the osmolarity was checked using a vapor pressure osmometer) was perfused into the chamber under gravity at the approximate rate of 2ml/min. Five or ten minutes after stopping the flow, the axon of interest was marked the same way as described earlier. Fifteen to twenty minutes later, the chamber was disassembled as described above. Then the coverslip with the cells was placed in fresh fixative for another twenty minutes at room temperature. After the fixation, the dish was rinsed twice for five minutes each in sodium phosphate buffer and left in fresh buffer overnight in the refrigerator.

2.4.4 Attempted reactivation of moving filaments

To explore the mechanism underlying the capture of moving filament, I tried to reactivate the moving filaments after they stopped. For these experiments, I used a different concentration of saponin (0.008%) to permeabilize the axon and stop the moving filaments. The goal of using much lower concentration of saponin was to stop the moving filament while causing minimum disturbance in the transport machinery. The 0.008% saponin could sufficiently permeabilize the axon and capture moving structures (see results).

The permeabilization solution also included 10µM taxol 6.5 µM phalloidin and 100µM AMP-PNP. The cells were imaged and permeabilized as described above. To reactivate the movement, 8-10ml 2mM ATP in NPHEM buffer was flowed into the chamber under gravity at the rate of around 3ml/ml. In some experiments, I substituted 10mM ATP for the 100µM AMP-PNP in the permeabilization solution.

71 2.5 Neurofilament splaying

2.5.1 Use of sister coverslips

In experiments comparing captured filaments and splayed filaments or splayed filaments under different conditions, I performed the capture and splaying on sister coverslips (i.e. coverslips from the same culture batch) and then prepared them for immunostaining side –by-side in an identical manner.

2.5.2 Neurofilament splaying in the chamber

Coverslips were assembled in the imaging chamber and the chamber was placed on the microscope stage as described earlier. To cause the neurofilaments to splay from each other, 2 ml permeabilization solution containing 0.5% Triton X- 100 (Sigma) in NPHEM was flowed into chamber under gravity at the rate of 3.8ml/min. Five minutes later, the cells were fixed in the chamber by perfusing 2ml 2% (w/v) formaldehyde in NPHEM under gravity at the rate of approximately 2ml/min. Fifteen minutes later, the chamber was removed from the stage, submerged in PBS and disassembled as described earlier.

2.5.3 Neurofilament splaying in open dishes

To cause the axonal neurofilaments to splay apart from each other, coverslips in the open dishes were rinsed once with PBS, once with NPHEM and then treated with 0.5% Triton X-100 (Sigma) in NPHEM at room temperature After five minutes, the cells were fixed at room temperature in a solution of 4% (w/v) formaldehyde in NPHEM for twenty minutes. After the fixation, the dishes were rinsed with PBS and treated with PBSNT as described above.

72 2.6 Immunostaining

2.6.1 Immunostaining of captured and splayed filaments

The immunostaining was carried out as described elsewhere (Brown, 1997). Briefly, the cells were rinsed with PBS and then “blocked” with 4% normal goat serum (Jackson Immunoresearch) in PBS (“Blocking solution”). A glass cloning ring was attached to the coverslip using silicone grease to create a well for the antibody with cells in the center. First the cells were stained with primary antibodies diluted in blocking solution (please refer to Table 2.1 for the information of all the primary antibodies used in the experiments). Unbound primary antibodies were rinsed off with PBS and the cells were blocked again prior to incubation with secondary antibody. All the primary mouse monoclonal antibodies were visualized using Alexa 647 conjugated goat anti-mouse secondary antibody (Invitrogen) diluted 1:200 in blocking solution and all the primary rabbit polyclonal antibodies were visualized using Rhodamine Red-X TM - conjugated goat anti-rabbit antibody (Jackson Immunoresearch) diluted 1:200 in blocking solution. Unbound secondary antibody was rinsed off with PBS and the coverslip was mounted on glass slides in prolong gold antifade medium (Invitrogen). All the antibody incubations were performed for 45 min at 37°C in a humidified box.

For quantitative analysis of immunofluorescence intensity of splayed and captured neurofilaments, the immunostaining protocol was identical to that described above except that the secondary antibody staining was followed by blocking in 4% normal donkey serum blocking solution and then a tertiary Alexa 488 conjugated donkey anti-goat antibody (Invitrogen) diluted 1:200 in blocking solution(see Figure 2.2). In this way, all routine inspection and focusing were performed on the Alexa 488 channel, and the rhodamine and Alexa 647 channels was reserved exclusively for image acquisition to avoid the possibility

73 of photobleaching during routine inspection and focusing. In some experiments, the intrinsic GFP along filaments from cells expressing GFP-NFM was used for inspection and focusing instead of the tertiary antibody fluorescence (figure 2.2). For quantitative analysis, all images were acquired using the 100x/1.4 NA Apo phase contrast objective and identical exposure times, thereby ensuring identical and minimal photobleaching (Brown, 1997). To confirm that excitation of Alexa

488 (λmax absorption 495 nm/ λmax emission 519 nm) won’t cause photobleaching of Rhodamine Red X (λmax absorption 580 nm/ λmax emission 590 nm), I compared the fluorescence intensity along individual filament before and after Alexa 488 had been continuously excited for 5 or 10 minutes. I found no bleaching of Rhodamine Red X during excitation of Alexa 488.

2.6.2 Immunostaining of SW13 cells

The cells were rinsed briefly twice in 37°C PBS and then fixed in 4% (w/v) formaldehyde at 37°C for 15 minutes. After the fixation, the cells were rinsed with PBS and then treated with 2 ml 0.25% Triton X-100 in PBS at room temperature for 15 min (“post-fixation extraction”) to ensure complete demembranation of the cells prior to immunostaining. Then the cells were processed for immunostaining as described for the SCG neurons except that there was no ring attached to the coverslip. For the antibody incubation, the coverslips bearing the SW13 cells were moved from the plates to a tray with a layer of parafilm covered at the bottom. Because of the hydrophobic property of the parafilm, the antibody solution remained on the coverslip.

74 Antibody Supplier Type Dilution Reactivity Comments name NFLAS Virginia lee Rabbit 1:300 Phosphorylated & This antiserum was raised against a synthetic (Univ. Penn.) antiserum Non-phosphorylated peptide identical to the 20 amino acid sequence at NFL the extreme carboxy terminal of human NFL (Trojanowski et al., 1989). RMO270 Zymed Mouse 1:100 Phosphorylated & RMO270 was raised against purified NFM from rat monoclonal non-phosphorylated spinal cord (Lee et al., 1987) NFM RMO55 Virginia lee Mouse 1:100 Phosphorylated NFM RMO55 was raised against purified NFM from rat monoclonal spinal cord (Lee et al., 1987) and it binds to a phosphorylated epitope in the tail domain of rat NFM (Virginia Lee, personal communication) FNP7 Virginia lee Mouse 1:100 Non-phosphorylated FNP7 was raised to a non-phophorylated monoclonal NFM and NFH synthetic peptide containing the sequence 75 KSPVPKSPVEE (Virginia Lee, personal communication), which is located between residues 603 and 613 in the tail domain of rat and mouse NFM. The sequence contains two phosphorylation sites and both of them can be phosphorylated in vivo (Xu et al., 1992). 7C5 Gerry Shaw Mouse 1:100 Phosphorylated & 7C5 was raised against recombinant rat (Univ. monoclonal non-phosphorylated peripherin purified from E. coli. Florida) peripherin

Continued

Table 2.1 Antibodies used in the experiments Table 2.1 continued

Antibody Supplier Type Dilution Reactivity Comments name αBB Ronald Liem Rabbit 1:200 Phosphorylated & αBB was raised against a C-terminal peptide (Univ. polyclonal non-phosphorylated amino acid residue 305-340 of rat α-internexin Columbia) α-internexin purified from E. coli (Ching et al., 1999). RT97 Biodesign Mouse 1:50 Phosphorylated NFH RT 97 was raised against semi-purified rat monoclonal and NFM neurofilament proteins (Wood and Anderton, 1981) and recognizes phosphorylated epitopes (KS/TP repeats 43-52 on the tail domain of mouse NFH, 43 is KTP site) on the NFH (Bajaj and Miller, 1997), and to a lesser extent, on NFM. In the absence of NFH, the affinity of 76 RT97 for NFM is increased (Sanchez et al., 2000). SMI36 Stemberger Mouse 1:100 Phosphorylated NFH SMI36 recognizes phosphorylated NFH and, to Monoclonal monoclonal and NFM less extent, phosphorylated NFM (Sternberger and Sternberger, 1983; Martus, 1988). N52 Sigma Mouse 1:100 Phosphorylated & N52 was raised to the carboxy terminal tail monoclonal non-phosphorylated segment of enzymatically dephosphorylated pig NFH NFH and it binds to an epitope in the tail domain of rat NFH between residues 559-1072 (Harris et al., 1991).

Continued

Table 2.1 Continued

Antibody Supplier Type Dilution Reactivity Comments name RMO217 Virginia lee Mouse 1:100 Phosphorylated NFH RMO217 was raised against purified NFM from monoclonal rat spinal cord (Lee et al., 1987) 3G3 Gerry Shaw Mouse 1:100 Phosphorylated & 3G3 was raised to a recombinant fusion protein monoclonal non-phosphorylated containing the c-terminus of NFH (Harris et al., NFH 1993). NFHAS Virginia lee Rabbit 1:100 Phosphorylated & polyclonal non-phosphorylated NFH AB1989 Chemicon Rabbit 1:100 Phosphorylated & AB1989 was raised to arecombinant fusion (R12) polyclonal non-phosphorylated protein containing the extreme carboxy 77 NFH terminus (residues 846-1072) of rat NFH purified from E.coli (Harris et al., 1991).

R49 Gerry Shaw Rabbit 1:100 Phosphorylated & R49 was raised to a recombinant fusion protein polyclonal non-phosphorylated containing the extreme carboxy terminus NFH (residues 846-1072) of rat NFH purified from E.coli (Harris et al., 1991).

R14 Gerry Shaw Rabbit 1:100 Phosphorylated & R14 was raised to a recombinant fusion polyclonal non-phosphorylated proteins containing the C-terminal residues NFH 559-794 of rat NFH purified from E.coli (Harris et al., 1991).

CPCA- Gerry Shaw Chicken 1:1000 Phosphorylated NFH NF-H polyclonal

Figure 2.2 Schematic diagram of immunostaining protocol for neurofilaments used for quantitative analysis. As shown in (A), the neurofilament (represented by thickened grey curved line) is visualized with specific primary neurofilament antibody (in grey) following by Rhodamin Red-XTM or Alex 647 conjugated secondary antibody (in red). The antibodies in green represent the tertiary antibody, which is normally Alex 488 conjugated. All the routine inspection and focusing were carried out in Alexa 488 channel and the secondary antibody channels were reserved exclusively for image acquisition to ensure minimum and equal photobleaching of the secondary antibody fluorescence. In some cases, the neurofilaments were labeled with GFP, as shown in (B). The filaments were processed for regular double-labeling immunofluorescence microscopy and the intrinsic GFP fluorescence was used for focusing.

78 2.7 SDS-PAGE and western blotting

2.7.1 Preparation of cytoskeletal proteins

A crude preparation of mouse cytoskeletal proteins was obtained from adult mouse essentially as described by Uchida (Uchida et al., 1999; Uchida et al.,

2004). Adult mice (10-20 weeks) were decapitated under deep CO2 anesthesia. The spinal cord and sciatic nerve with ventral roots (L4 and L5) were dissected out and cut into short segments (5-10mm). The segments were crushed into fine powder in liquid nitrogen and homogenized in 1ml ice-cold extraction solution composed of Triton-buffer (1% Triton X-100, 50mM Tris-HCl, pH 7.5, 5mM EGTA, 2mM MgCl2, 100mM NaCl and 1mM Dithiothreitol), 10µg/ml Bestatin (Sigma), 10µg/ml Leupeptin (Sigma) and 10µg/ml E64 (Sigma). The homogenate was centrifuged at 100,000 g for one hour at 4°C and the precipitate was used as the cytoskeletal fraction. The proteins in the precipitate were dissolved in 500µl (for proteins from the spinal cord) or 200µl (for the proteins from the sciatic nerve) sodium dodecyl sulfate (SDS) sample buffer composed of 62.5mM Tris- HCl pH 6.8, 2% SDS, 5% 2-Mercapto-Ethanol and 7% glycerol.

2.7.2 SDS-PAGE and immunoblotting

The specificity of the monoclonal and polyclonal antibodies was confirmed by immunoblotting using mouse cytoskeletal proteins, which were prepared as described above. The proteins were separated by SDS-PAGE on 7.5% polyacrylamide gels using a mini-vertical gel electrophoresis unit (Amersham Biosciences) and transferred onto Immobilon-P PVDF (Immobilon-P transfer membrane, 0.45µm pore size, Millipore) using a mini trans-blot electrophoretic

79 transfer cell (Bio-Rad Laboratories). Transferred proteins were visualized by staining a strip of the PVDF blot with a solution of 0.025% Coomassie Brilliant Blue R250 (Sigma) in 40% methanol and 7% acetic acid.

The immunostaining was carried out as described by Brown (Brown, 1998) except that the strips of PVDF blots were first wetted using 100% methanol since they are hydrophobic. Next they were rinsed briefly and then three times with Tris buffered saline (TBS; 50 mM Tris-HCl, 138 mM NaCl, 2.7 mM KCl, pH 7.4) each for five minutes. After that the strips of the PVDF blot were blocked with a 5% (w/v) solution of Carnation brand non-fat dry milk in TBS (“blocking solution’) for one hour and then they were probed with the primary antibodies. The antibody incubations were carried out on a shaker at room temperature for one hour. Unbound primary antibodies was rinsed off with TBS containing 0.1% Tween-20 (Sigma) four times each for 5 minutes and then with TBS alone twice each for 2 minutes. After being rinsed, the blot strips were blocked again for one hour in the blocking solution and then incubated in the secondary antibodies (goat anti- mouse IgG or goat anti-rabbit IgG, conjugated to horseradish peroxidase) at a dilution of 1:50,000 in TBS containing 1% normal goat serum. Unbound secondary antibodies were rinsed off with TBS containing 0.1% Tween-20 and then TBS alone as described above. The bound antibodies were visualized using ECL plus western blotting detection reagents system (Amersham Biosciences).

2.7.3 Enzymatic dephosphorylation

To test the dependence of the antibody binding on phosphorylation state by immunoblotting, the neurofilament proteins were dephosphorylated enzymatically by treating the PVDF blots with 3units/ml alkaline phosphatase (purified from Escherichia coli, Sigma) in dephosphorylation buffer (50mM Tris-HCl, 138mM

NaCl, 2.7mM KCl, 1mM ZnSO4, 1mM MgCl2, pH 8.0) for at least 18 hours at room temperature on a shaker (Sternberger and Sternberger, 1983; Carden et

80 al., 1985; Brown, 1998). The alkaline phosphatase was removed by rinsing the blots with 400mM sodium phosphate, 100mM NaCl, 50mM NaEDTA, pH 7.2. Then the blots were processed for immunoblotting as described earlier. Since the PVDF membrane is hydrophobic, they were wetted and rinsed before the enzymatic dephosphorylation as described above. The rinsing after methanol treatment is very important, since methanol denatures protein and any left-over might affect the alkaline phosphatase treatment.

2.8 Analysis of fluorescence intensity along captured and splayed filaments

2.8.1 Image acquisition

All images of immunofluorescence were acquired with the use of a Quantix cooled CCD camera (Roper Scientific) equipped with a Kodak KAF1400 chip. The camera was operated at maximum gain with a readout speed of 5Mhz and no pixel binning.

2.8.2 Criteria for the selection of splayed filaments for analysis

To select splayed neurofilaments for analysis, I used the following criteria: 1) the filaments were located in the central area (660 X 520 pixel) of the image. 2) the fluorescence intensity along the filaments appeared uniform and there were no sharp bends (which cause segmentation artifacts in the segmented mask procedure, see below), 3) the filaments could be traced unambiguously for at least 3µm. Paired filaments were analyzed if the two filaments were co-aligned for a distance of at least 2µm.

81 2.8.3 Quantification of fluorescence labeling along captured and splayed filaments

To quantify the incorporation of GFP-NFM along individual filaments, splayed axonal cytoskeletons were immunostained using an antibody to NFL (see above) and GFP-NFM was visualized directly using the intrinsic GFP fluorescence. To quantify the distribution of different polypeptides and phosphorylation epitopes along captured and splayed filaments, the coverslips were processed for double- labeling immunofluorescence miscroscopy using NFL or NFM antibody to visualize the entire filament length and some other primary antibody as described above in immunostaining section. NFL and NFM are present along entire length of all neurofilaments in these axons (See Results). Using MetaMorph™ software (Universal Imaging), a line was drawn along the medial axis of each filament in the image of NFL or NFM fluorescence and then one or more lines were drawn on the image of the second fluorescence corresponding to the portion or portions of each filament that contained the fluorescence (Figure 2.3). For each filament, the total length of the line(s) on the other fluorescence image was divided by the length of the line on the NFL or NFM image to determine the percent of the filament length containing corresponding fluorescence. As can be seen from the description above, the quantification of fluorescence labeling along individual filament was not affected by the non-uniformity of brightness in the images and thus the analysis was not restricted to the filaments located in the center quarter of the images.

2.8.4 Intrinsic flat field correction

To quantify the fluorescence intensity along the filaments, each image was first corrected for non-uniformity in image brightness across the field of view, which is inherent in light microscopic images. To do this, an image of the background fluorescence was obtained for each raw image by a digital filtering procedure that is called intrinsic flat field correction (Brown, 2003a). This procedure involved 82

Figure 2.3 Methods for analyzing the proportion of the neurofilament length that contains fluorescence. This figure shows images of a splayed filament double stained with NFLAS antibody specific for NFL (A) and the RT97 antibody specific for phosphorylated NFH and NFM (B). To measure the length of the filament, a line was drawn along the medial axis of the filament in the NFL image, which is represented by the red line in image (A). The portions of the filament that contains RT97 immunofluroscence are represented in green lines in image (B). The lines in (A) and (B) are superimposed in (C): portions of the neurofilament that stained with RT97 are shown in yellow and the portions that doesn’t stain are shown in red. Scale bar= 2µm.

subtraction of the bias offset value and then sequential application of a local minimum filter, then a local maximum and finally a local average (smoothing) filter to each image using Oncor-Image software. The filter size used in this process is 121x121 pixels. This is biggest filter that can be used to preserve the background without any artifact (Anthony Brown, personal communication). This procedure has the effect of “filtering out” bright objects that are smaller than 121 pixels in either dimension. If the axons occupy a big portion of the image, the glow from the axonal fluorescence will also contribute to the background image because of the limited filter size. Thus this filtering procedure only works when each object (in this case, axon) is smaller than the filter size and the objects occupy a small fraction of the area. The local maximum filtering then eliminates the pixels in the background with lowest value, which could represent the noise. Finally the average filtering further reduces the noise and evens out pixel values 83 in the background image with respect to the values of neighboring pixels. The background images were created using Oncor Imaging software (Oncor).

The created background image was then used to correct the raw image using the equation for the flat-field correction:

Corr. Image = M * [(Raw image – bias) / Background image)]

Where M is the mean pixel value in the background image. The flat-field correction was carried out using MetaMorphTM software (Universal Imaging). Examples of a raw flat-field corrected image are shown in Figure 2. 4.

To take advantage of the full dynamic range of the camera chip without saturating it, normally I used 200ms to acquire the raw image for image analysis. But for obtaining the background fluorescence image, I acquired a separate image using higher exposure time (500ms). In this way I could raise the signal to noise ratio and obtained a more accurate representation of the background. Since saturated pixels were filtered out during the local minimum filtering, background image creation was not affected by the saturation that might be caused by this high exposure time.

2.8.5 Quantification of the fluorescence intensity along captured and splayed filaments

The fluorescence intensity was analyzed along filaments in the flat-field corrected images using the segmented mask method described before (Brown, 1997; Brown et al., 1993; Brown et al., 1992). The procedures for creating the mask and for subsequent measurement of the images were carried out using MetaMorphTM software (Universal Imaging). First, a high pass heavy 7x7 matrix convolution filter (Figure 2.5) was applied to the flat-field corrected image to

Figure 2.4 Examples of raw and flat-field-corrected images. (A) Raw image with 200ms exposure and (B) background image generated from a raw image of exact same field view with 500ms exposure. (C) Corrected image. Note that the filaments or axons pointed by the white arrows appeared to be fainter in (A) and looked quite similar to other filaments and axons in (C). Scale bar=150µm.

accentuate the edges of the neurofilaments. Then, a thresholding operation was performed on the filtered image, and extraneous objects were removed using eroding operation and selectively painting. The resulting binary mask included only the neurofilaments of interest. The edited mask was dilated by four pixels in the directions perpendicular to the longitudinal axis of the neurofilament and then divided into contiguous 0.847µm-long segments by performing “create segment regions” operation. A complementary background mask that surrounded the neurofilament mask was also created so that each segment in the neurofilament mask was flanked by two segments in the background mask (Figure 2.6). To quantify the fluorescence intensity along the neurofilament, the segmented mask of the neurofilament was overlaid on the flat-field corrected image of the NFL fluorescence. To correct for the background, the mean pixel intensity was calculated for each pair of segments in the background mask and this value was subtracted from each segment in the neurofilament mask. The total fluorescence intensity for each segment in the neurofilament mask was then calculated by summing the corrected intensities of the individual pixels in that segment. The

85 evaluation of the efficacy of the masking procedure was described else where (Brown et al., 1992). The mask of the filament was overlaid on the image of NFL fluorescence from which it derived and the proportion of the filament fluorescence contained within the mask at different sites was estimated. To do that, a line was drawn perpendicular to the long axis of the mask at different points along the axon and pixel intensity along the line was determined using the “linescan” tool in MetaMorphTM software (Universal Imaging), which normally provided a bell shaped curve of pixel intensity versus distance along the line. The created mask was estimated to cover more 95% area of the curve, which indicates that more than 95% of the total NFL fluorescence was contained in the mask.

To compare RT97 immunofluorescence intensity along the individual filaments, the filament and background mask were created and fluorescence intensity was quantified in exactly the same way as described above with the following modifications. For the analysis, RT97 fluorescence intensity was eventually divided by NFL fluorescence intensity. Uneven illumination is related to the microscope optical system and should affect different chromophores equally. Thus RT97/NFL immunofluorescence intensity ratio should not have been affected by the non-uniformity of image brightness. Because of the above reasons, all the quantification was performed using the raw images without flat field correction. To determine the size of the mask that should include more than 95% of both RT97 and NFL fluorescence, I initially used the same mask size (with four pixels dilation) that I used for NFL fluorescence analysis and estimated the efficacy of the mask procedure as described above. Finally I determined that the mask needed to be two pixels wider on both sides comparing to the previous one described above to include more than 95% of both RT97 and NFL fluorescence. Because the fluorescence intensity ratio was used for the analysis, the mask was not divided into segments.

Figure 2.5 The operation of a convolution matrix in the high pass heavy 7x7 filter. The filter (kernel) used for sharpening the image have numeric values covering a 7x7 pixel area (a&a’). The kernel is multiplied against the values of 49 (7x7) pixels covered by the mask in the original images (b&b’), a process called convolution. The resulting 49 values (c&c’) are summed to obtain a value that is assigned to the central pixel location in the new filtered images (d and d’). The kernel then moves by 1 pixel to calculate the next pixel value, the process repeats until all the pixels in the original image have been recalculated. Notice the values in the kernel used for sharpening. The central values is emphasized (matrix value 45) relative to the 48 surrounding values. When the central pixel value is relative high, it will become even higher after the convolution (A); when the central pixel value is relative low, it will become even lower after the convolution (B). The sum of the values in the kernel is 1.

Figure 2.6 Methods for analyzing the fluorescence intensity along the filaments. (A) An image of unpaired neurofilament with the segmented masks (multiple colors). The masks have been estimated to include more than 95% of the filament fluorescence. (B) An image of the same filament as in (A) with segmented background masks with corresponding colors. These segmented masks were used in quantitative analysis of NFL immunofluorescence intensity along individual filament. (C) An image of unpaired neurofilament with unsegmented mask and (D) the same filament with corresponding background mask. The unsegmented masks were used in the quantitative analysis of the ratio of RT97 to NFL immunofluorescence intensity along individual filament. The neurofilament mask in (C) is two pixels wilder on bother side than those in (A) and it was adjusted to include more than 95% fluorescence in both NFL and RT97 images. Scale bar=1.5µm.

88 2.8.6 Quantification of the fluorescence intensity along axons

To quantify the change in the fluorescence intensity along the axon in time lapse images, masks were created around the axon as described above by dilating the binary axon mask four pixels in the directions perpendicular to the longitudinal axis of the axon without segmentation. The mask was estimated to include 95% fluorescence in the axon. The background mask was created exactly as described above. The total fluorescence intensity for the axon in each image was then calculated by summing the intensities of the individual pixels in the mask, which had been corrected for the background as described above. The fluorescence intensity of the same axon for each image acquired after the start of perfusion was then normalized to the one in the image immediately before the start of perfusion.

2.9 Electron microscopy

2.9.1 Strategy developed to locate the axons of interest

In order to ensure that I could identify the location of the axon of interest unambiguously in the electron microscope, all surrounding axons were cut away using a glass fiber mounted on the Eppendorf InjectMan™ NI2 micromanipulator (Brinkman Instruments) the day after the moving filaments were captured and fixed using glutaraldehyde. The glass fiber for cutting was pulled to a diameter of 20-30µm from 1-mm-diam borosilicate glass rods (World Precision Instruments) using a Sutter P-87 Flaming-Brown pipette puller. In most cases, the surrounding axons were interacting with the axons of interest by intersecting or fasciculating outside of the region of the gap. To avoid stretching or detaching the axon of interest while scraping away these surrounding axons, I first severed the axon of interest at points far away from the region containing the captured filament, thus leaving that part of the axon in isolation. To sever the axon, the glass rod was 89 lowered down until the tip was touching the coverslip and I could feel there was some resistance between the coverslip and glass rod. Then I moved the glass rod against the coverslip at the point where I wanted the axon to be served in a direction perpendicular to the longitudinal axis of the axon until it was served. The resistance is very critical in this manipulation. If it was too low, the axon could not be cut efficiently; if it was too high, the glass rod couldn’t be manipulated properly and chances were high that the axon could be easily damaged. Once the axon of interest was served and isolated from surrounding ones, the other axons in the vicinity including the flanking segments of the axon, which was served from the region containing the captured filaments, were easily scraped away by moving the glass rod against the coverslip in directions away from the region of interest. After that, the axon was further shortened to only include the gap containing the captured filament and the flanking 30-40µm-long regions. Six to eight cell bodies were then placed around the axon to mark the location (Figure 2.7 & 2.8). To do this, the glass fiber was used to sever the axons around nearby cell bodies and then push them across the glass coverslip to the area where the axon was located. In order to ensure that the cell body remained attached through the electron microscopy processing, the cell body was pushed against the coverslip from above using the glass rod.

2.9.2 Electron microscopy

For those experiments involving semi-thin sectioning, the fixed cells were treated with 5% Trition X-100 (Sigma) in sodium phosphate buffer for 5 minutes to extract membranes and thereby facilitate subsequent detection of the captured neurofilaments in electron microscope. After marking the location of the axon of interest, coverslips bearing captured filaments were rinsed twice in sodium phosphate buffer, postfixed for 15 min in 1% osmium tetroxide (Electron Microscopy Sciences), rinsed twice for two minutes each in 3.6% NaCl, rinsed twice for two minutes each in water, stained for thirty minutes with uranyl acetate

Figure 2.1 Schematic diagram of the operation used to facilitate location of the axon in electron microscope. Lines of different thickness represent axons of different diameters and filled circles represent cell bodies. (A) Schematic drawing of a typical light microscope image using 40x objective, in which the axon of interest in the center is surrounded by other axons and cell bodies. Note the context is very clear in the image since the field of view is large (several hundred micrometers wide). In the electron microscope, the field of view is very small (only around 5µm wide at a magnification of 22K) and the context will be lost. The other surrounding axons might be easily mistaken for the axon of interest. To help reliably relocate the axon in the electron microscope, all the surrounding axons were cut away and the axon was shortened to include only the region of interest containing the gap and captured filament and the flanking 30-40µm regions, as shown in (B). Then several cell bodies were moved into the area to mark the location (C).

Figure 2.8 Strategy to locate the axon of interest. (A) Phase contrast image of an axon (very faint marked by the arrows) surrounded by nine cell bodies, before being processed for electron microscopy. The axon was shortened only including the gap containing the captured filaments and the flanking 30-40µm- long regions. The cells bodies were moved from places nearby to mark the location of the axon. The quality of the image is not great because both the axon and cell bodies had been permeabilized and fixed. (B) Bright field image of the resin block, showing the location of the axon of interest marked by the cell bodies. The image is fuzzy because the light was passing through thick plastic. Note that the axon is not visible and thus it would be impossible to locate without the cell bodies surrounding it. Scale bar=30 µm.

92 (Polysciences) in the dark. The coverslips were washed thoroughly before the treatment of uranyl acetate because any residue of phosphate buffer might cause precipitation (uranyl phosphate). The cells were then dehydrated first in 50% ethanol for ten minutes, next through 70, 80, 95% ethanol respectively for 5 minutes and finally three times for five minutes in 100% ethanol. For infiltration, the coverslips were treated first with propylene oxide (Electron Microscopy Sciences) twice for five minutes each, with a 1:1 mixture of propylene oxide and Poly/Bed 812 epoxy resin (Polysciences) for thirty minutes, then with a 1:2 mixture for another thirty minutes, finally with resin alone for thirty minutes twice (the protocol was adapted from Kathy Wolken in Campus Microscopy & Imaging Facility at the Ohio State University). Adequate infiltration was achieved by placing the glass container with specimen on a slow rotary mixer during infiltration. When it was ready for polymerization, the coverslip was held on its edge for some time to let most of the resin drip into a waste pool and was wiped at the bottom using tissues to remove excess resin. Then the coverslip was rested on an embedding mold with the cell side facing up. A BEEM capsule was filled with fresh resin and flipped over onto the coverslip around the area where the axon of interest was located. To estimate the location of the axon, one corner of the coverslip bearing the captured filaments was cut off using a high precision fine diamond scribe (Electron Microscopy Sciences). The relative position of the axon to this particular corner was then estimated using the light microscope and in this way the BEEM capsule could be placed in the proper location. The resin was polymerized in a vacuum oven at 65°C for two days. After embedding, the coverslip was dissolved by being exposed to 48% hydrofluoric acid for ten minutes. The area, where the axon of interest was located, could be easily found under phase contrast with the help of the surrounding cells.

Trimming of the block was performed using a single-edged razor blade under a dissecting microscope. Then ultra-thin sections (silver/gold; approximately 80- 100 nm) and semi-thin sections (blue/green; approximately 200-300 nm) were

93 cut with a diamond knife (DuPont) using a Reichert-Jung Ultracut microtome, retrieved on hexagonal 200 mesh copper grids, stained with uranyl acetate (15 minutes for ultra-thin sections and half an hour for semi-thin sections) and lead citrate (30 seconds for ultra-thin sections and three minutes for semi-thin section). In order to avoid loss of sections during staining, the grid was coated with an adhesive. Adhesive-covered grids were prepared by dissolving one-inch double sided Scotch-brand transparent adhesive tape (1” wide) in 5ml acetone. The grids were picked up with finely tipped forceps and dipped into this solution briefly, then dried in air. After completely drying, the grids were checked using a dissection microscope and any grids with aggregates of adhesive were discarded. The adhesive-covered grids were stored on clean parafilm in Petri dishes.

The sections were observed using a Philips CM-12 electron microscope at 80KV. For each section, overlapping images of the entire gap and flanking regions were acquired on photographic film at a magnification of 22k and 45K and digitized using a flatbed scanner at a resolution of 1200 dpi.

2.9.3 Reconstruction of the axon and measurement of neurofilament diameter

Montages of the axon were constructed from the digitized images using Adobe Photoshop and the optimal overlap of adjacent images was ensured by creating “transparent” effects with each image. Then the outline of the axon and the neurofilaments were traced in the Montages using Adobe Illustrator. The line corresponding to the neurofilaments were thickened to make them more visible. Neurofilament diameter was measured using the “caliper” tool in MetaMorphTM (Universal imaging). For the filaments in the ultrathin sections, I performed measurements every 100 pixels along the filaments; for those in the semi-thin sections, I performed measurements every 300 pixels. Confusing points along 94 the filament, where the filament is too close to the membrane, or other ultrastructures, were excluded from the measurement. In total there were 20-35 measurements for each captured filament.

CHAPTER 3

LIGHT MICROSCOPIC ANALYSIS OF THE STRUCTURE OF MOVING NEUROFILAMENTS

3.1 Introduction

Radioisotopic pulse labeling studies revealed that neurofilament proteins are transported along the axon in slow components at the rate of 0.004-0.04µm/s (Black and Lasek, 1979), but the transport machinery is still controversial. The main controversy has been focused on the form in which neurofilament proteins are transported along the axon: polymers or subunits? Recently live-cell imaging studies with cultured neurons have demonstrated that GFP tagged neurofilament proteins actually move predominantly in a filamentous form along the axons in a rapid intermittent bidirectional manner (Roy et al., 2000; Wang and Brown, 2001; Ackerley et al., 2003; Uchida and Brown, 2004). But the actual structure of these moving proteins still remains a mystery because of the diffraction limit of light microscopy (Terada and Hirokawa, 2000; Terada, 2003). For example, they could be neurofilament polymers. Alternatively, they could be membrane bound tubules containing GFP tagged neurofilament proteins (Terada, 2003).

To resolve this issue, I have sought to devise a definitive test of the hypothesis that these moving structures observed by fluorescence microscopy are indeed single neurofilament polymers. This proved to be challenging because the movement is both rapid and infrequent. To circumvent this problem, I took

96 advantage of naturally occurring gaps in the axonal neurofilament array of cultured rat SCG neurons (Wang et al., 2000) and captured the moving filaments within these gaps. Since these gaps lack neurofilaments, they are the only places in the axon where moving filaments can be studied in isolation. The length of gaps in the neurofilament array ranges from several micrometers to forty micrometers or even longer, but gaps longer than 20µm are rare. Neurofilaments move rapidly through these gaps, with peak velocity as high as 3 µm/s and average velocity (excluding pauses) of 0.4 to 0.7 µm/s (Brown, 2000; Roy et al., 2000; Wang et al., 2000). It takes several tens of seconds for a filament to move through a 20-µm-long gap. To stop the moving filament in the middle of the gap, the movement needs to be detected fast and reliably. To observe neurofilaments by fluorescence microscopy, I transfected cultured neurons with GFP-tagged NFM. I confirmed that gaps visible in axons of neurons expressing GFP-NFM were devoid of neurofilament proteins and GFP fluorescence on individual neurofilament could be relied on to detect the movement of neurofilaments. My strategy was to observe a filament entering a gap and then capture it, before it exits, by rapidly permeabilizing the axon. When axons are permeabilized, some soluble factors necessary for the transport machinery will diffuse out of the axon and the moving filament will eventually stop. Using quantitative immunofluorescence and electron microscopy, I demonstrated in this chapter and subsequent chapter that these captured filaments are indeed single neurofilament polymers.

3.2 Naturally occurring gaps

All the experiments described in this and next chapters were performed on cultured neurons from the superior cervical ganglia (SCG) of neonatal rats. It has been shown previously that these cultured neurons frequently exhibit short regions of axon that lack neurofilaments entirely (Wang et al., 2000). Neurofilament is one of the most abundant proteins in neurons and normally they 97 form a continuous neurofilament array that extends along most of the length of the axon. Thus these gaps represent discontinuities in the neurofilament array. Gaps are also visible in axons expressing GFP-tagged neurofilament proteins because most GFP-tagged proteins are incorporated into the endogenous neurofilament array and there are very few diffusible subunits (Wang et al., 2000). These gaps appear to arise due to the relatively low abundance of neurofilaments in these neurons, and they have also been reported in neurons from rat brain cortex (Ackerley, 2003). It is unlikely that the gaps in these cultured neurons have any functional significance, because neurons can survive and function in the absence of neurofilaments. For example, in NFL knock out or NFM and NFH double knock out mice, there are no neurofilaments found in the axons but axons can extend and function properly except that there are no large axons in these animals (Zhu et al., 1997; Elder et al., 1999b; Jacomy et al., 1999). As described earlier, the main function of neurofilament proteins is to contribute to the axonal caliber, especially for large axons, which is usually myelinated (see section 1.1.2 and 1.1.4 of Chapter 1). Since there is no myelination happening in these culture dishes, short segments of axons lacking neurofilament in these cultured neurons do not appear to have significant function. In addition, studies of neurofilament transport using these naturally occurring neurofilament gaps and using photobleaching technique revealed similar motile behaviors of neurofilaments in these cultured SCG neurons (Roy et al., 2000; Wang et al., 2000; Wang and Brown, 2001).

To characterize the structure of moving neurofilaments, I took advantage of these gaps in the GFP fluorescence in axons of neurons expressing GFP-NFM and developed a rapid permeabilization technique to capture moving filaments in the middle of the gaps while acquiring time-lapse fluorescence images. To confirm that these gaps in GFP fluorescence are really devoid of neurofilaments, I transfected neurons with GFP-NFM fusion protein and then fixed them with 4% (w/v) formaldehyde and stained the cells with a polyclonal antibody NFLAS,

98 specific for NFL. To confirm that the axon is continuous across the gaps in the neurofilament array, I also stained for actin using rhodamine phalloidin. For each gap, I acquired images of the NFL immunofluorescence, the rhodamine phalloidin fluorescence, and the intrinsic GFP fluorescence (Figure 3.1). In total I examined 38 gaps from 6 cells in four different dishes and the gaps in the GFP fluorescence all corresponded to those in the neurofilament array. During the whole thesis research project, I have observed more than 200 gaps in GFP fluorescence and processed them either for immunofluorescence microscopy or electron microscopy. In all cases, these gaps were devoid of neurofilaments. I will show later that this is because all neurofilaments contained GFP-NFM along at least part of their length (see section 3.4)

Figure 3.1 A naturally occurring gap. Cultured SCG neurons expressing GFP- NFM exhibit naturally occurring gaps in the axonal neurofilament array. (A) GFP- NFM in an axon visualized directly, without the use of the antibody. (B) Total neurofilaments visualized by immunofluorescence microscopy using a polyclonal antibody specific for NFL. (C) Actin filaments visualized by fluorescence microscopy using rhodamine-conjugated phalloidin. Note that the gap in the GFP fluorescence corresponds to a gap in the NFL staining, which confirms that the gap in the GFP fluorescence lacks neurofilaments. The continuity of the axon across the gap is indicated by the presence of actin filaments. Scale bar=5µm.

99 3.3 Neurofilament splaying

Individual neurofilaments in axons are packed very close to each other and they cannot be resolved by light fluorescence microscopy. However, when the axons are treated with high concentration of detergent, the neurofilaments loop out from the axonal bundles and separate from each other as observed by immunofluorescence microscopy using an antibody specific for NFL, which is present in all neurofilaments (Ching & Liem, 1993; Lee et al., 1993). I used this splaying technique extensively during my thesis research project to visualize single neurofilament using immunofluorescence microscopy and thus in this section I will discuss how this technique works.

The extent of the splaying is related to the concentration of detergent and the age of the culture. The higher the detergent concentration, the more extensive is the splaying. As the culture gets older, it also gets harder to cause neurofilament splaying using the same concentration of detergent, perhaps due to increased fasciculation of axons. The exact mechanism underlying the splaying is unknown. One of the possible explanations is that: normally in the axon neurofilaments are loosely interacting with each other or other structures. When the axonal plasma membrane is gradually dissolved by the detergent, these flexible polymers move apart passively by thermal motion. I also noticed that when the same concentration of detergent was applied with force into the dish, the neurofilament splaying was more extensive, which suggests that splaying can be enhanced if the solution is flowing at the time of permeabilization.

Since the dissociation of neurofilament into protofibrils only occurs in the presence urea (Krishnan et al., 1979), units smaller than single neurofilament polymers won’t exist under conditions used for splaying. Filaments from the splayed preparation must be either single or bundled neurofilaments, but they cannot be distinguished based on diameter because most of the splayed filaments have an apparent thickness of about 250-300nm, which corresponds to 100 the diffraction-limited resolution of light microscope (Figure 3.2). However, inspection of the images indicates that filaments do differ in brightness of NFL immunofluorescence. Most filaments appear to have uniform and comparable NFL staining. Other filaments appear to be brighter and often the brighter filament can be seen to arise from paring or crossing over of one or more of the former filaments.

If the unpaired filaments represent single neurofilaments, their NFL fluorescence intensity should represent the smallest fluorescence unit in the splayed cytoskeleton. The fluorescence intensity of other bundled neurofilaments could be presented as integral multiples of the amplitude of this unit. Consistent with this expectation, quantitative analysis of the unpaired and paired filaments revealed a bimodal distribution of the fluorescence intensity (Brown, 1997). The first peak occurred at the average fluorescence intensity of unpaired filaments and the value of the second peak, which represented the average fluorescence intensity of paired filaments formed by co-aligning two unpaired filaments, is twice that of first. Situations where co-alignment of three or more filaments can be identified unambiguously are very rare, so the analysis of these bundled filaments is not available (Brown, 1997). Nevertheless, all these observations and analysis demonstrate that unpaired filaments in the splayed preparations represent single neurofilament, and single and paired neurofilaments can be distinguished based on their staining intensity (Brown, 1997).

3.4 Detection of neurofilaments using GFP-NFM

To track neurofilament movements, cultured SCG neurons were first transfected with GFP-NFM fusion proteins by nuclear injection. It has been shown that three days after the transfection GFP-NFM was present in neurofilaments throughout the axonal arbor and ≥96% of this fusion protein remained in the axon after permeabilization with a non-ionic detergent (Wang et al., 2000). To further

101

Figure 3.2 Detergent-induced splaying of axonal neurofilaments. To splay the neurofilaments apart, the neuron was treated with 0.5% saponin for five minutes, and then fixed and processed for immunofluorescence microscopy using a specific antibody for NFL to visualize the neurofilaments. Note the numerous isolated filaments that have separated completely from the axonal neurofilament bundle so that one or both ends are visible. Inset shows splayed filaments at higher magnification. Quantitative analysis has shown that the fluorescence doubles where there is a cross-over or pairing of two splayed filaments pointed by the arrowheads (Brown, 1997). Scale bar = 25µm for the low magnification image; 5µm for the high magnification inset. The image is reproduced from Brown (1997) with permission of John Wiley & Sons, Inc.

confirm that GFP fluorescence could be used to detect moving neurofilaments, I quantified the incorporation of GFP-NFM along individual filaments by immunofluorescence microscopy.

Three days after transfection, I induced the axonal neurofilaments to splay apart from each other by permeabilizing the neuron with 0.5% Triton X-100 in the presence of 0.19M NaCl. To visualize the whole length of individual neurofilaments, the splayed cytoskeletons were stained with a polyclonal antibody specific for NFL. Images of the NFL immunofluorescence and GFP intrinsic fluorescence were acquired randomly along the axons of transfected neurons. Single neurofilaments were identified based on the intensity of their

102 staining with NFL antibody and the proportion of their length that had incorporated GFP-NFM was measured as described in section 2.8.3 of Chapter2. GFP-NFM incorporation was continuous along some neurofilaments and discontinuous along others (Figure 3.3). Fully labeled and partially labeled neurofilaments were found side by side in the same regions of axons. Totally I analyzed 196 unpaired filaments from 14 cells in 10 dishes of two different experiments. All neurofilaments contained GFP-NFM along >15% of their length and 88% contained GFP-NFM along >50% of their length. Thus all neurofilaments from these neurons incorporated GFP-NFM and most filaments had more than half of their length labeled with GFP fluorescence. The data suggest the GFP fluorescence can be relied on to detect moving filaments.

3.5 Permeabilization efficiency of 0.02% saponin in the perfusion chamber

Having established that gaps in GFP fluorescence were devoid of neurofilament proteins and GFP fluorescence on individual filaments could be relied to detect moving neurofilaments, I moved forward and started developing the perfusion chamber system to capture moving neurofilament proteins. As mentioned earlier, it only takes a filament tens of seconds to move though the gap. To stop the filament in the middle of the gap, the axon must be permeabilized as fast as possible without neurofilaments being splayed. For the present study, I used saponin, which is a plant glycoside. The typical molecule of a saponin has a large lipophilic region and a short side chain of sugar residue which interacts with non- lipid components. For this reason, saponins are soap-like and they are thought to cause permeabilization by interaction with sterols in the cell membrane. It has been shown previously that 0.02% saponin is sufficient to permeabilize axons of

103

Figure 3.3 GFP-NFM incorporation along single neurofilaments. (A) Axonal neurofilaments were induced to splay apart from each other by permeabilization of neurons with 0.5% Triton X-100 in the presence of 0.19M NaCl 3 days after they were transfected by nuclear injection with the GFP-NFM expression vector. Neurofilaments were detected by immunofluorescence microscopy using a polyclonal antibody specific for NFL and GFP-NFM was detected directly using the intrinsic GFP fluorescence. GFP-NFM incorporation was continuous along some filaments (left) and discontinuous along others (right). The white arrowheads mark the splayed filaments. Scale bar=1µm. (B) Histogram of the proportion of the length of each filament that contained GFP-NFM for 196 filaments from splayed preparations.

104 cultured neurons (Koehnle and Brown, 1999) and that it causes minimal neurofilament splaying (Brown, 1997), presumably because it permeabilizes the axonal membrane without stripping it away completely (Seeman, 1967).

To confirm that 0.02% saponin is sufficient to permeabilize axons in the perfusion chamber, I injected 10mg/ml tetramethylrhodamine dextran (Mw 10,000, Sigma) into the cytoplasm of the cells five days after plating. Three hours later the coverslip containing the microinjected cells was assembled into the chamber and moved onto the microscope stage as described in section 2.3.1 of Chapter 2. The cells in the perfusion chamber were permeabilized by flowing 0.02% saponin permeabilization solution under gravity at a rate of 3.8ml/min for about one minute and the loss of fluorescence during the perfusion was monitored by time- lapse fluorescence imaging. The fluorescent intensity along the axon at different time points after the perfusion was quantitatively analyzed using the modified segmented mask method (Brown et al., 1992; Brown et al., 1993) as described in section 2.8.6 of Chapter 2. Initially there was a transient and small increase (from 3% to 39%) in the fluorescence intensity along the axon immediately after the start of perfusion. The explanation for the increase is not clear and it could be due to the influence of environment change on the fluorescence of tetramethylrhodamine. Before the perfusion, the axons were observed in imaging medium which contains serum and after the start of perfusion the medium was gradually replaced by NPHEM buffer solution (see section 2.3.1 of Chapter 2). After the transient increase in fluorescence, there were two phases of fluorescence loss along axons. In the first phase, there was a small and gradual decrease in fluorescence intensity for 10-20 seconds. In second phase, the fluorescence was lost completely within 4-12 seconds and often the axon assumed a beaded appearance immediately prior to the complete loss of fluorescence. The difference in the duration of these two phases indicates that there was some variation in the rate of permeabilization. One minute after the start of the perfusion the average proportion of the fluorescence that remained in

105 the axon was 1.9% (min=0.8%, max=2.9%, n=14) (Figure 3.4). This experiment indicates that I can permeabilize axons in the perfusion chamber within one minute of starting the flow of perfusate.

Figure 3.4 Permeabilization with 0.02% saponin. (A) An axon containing tetramethylrhodamine dextran immediately before the flow of permeabilization solution. (B) The same axon one minute after the start of the flow. The images were photographed and processed identically to allow direct comparison. The background is brighter in A due to weak autofluorescence of the culture medium. Scale bar=5µm.

3.6 The capture of moving filaments

To capture moving neurofilaments, coverslips containing neurons expressing GFP-NFM were mounted in the perfusion chamber and gaps in the GFP fluorescence were located and observed by time-lapse imaging. When the leading edge of a fluorescent filament moved into the gap, the binder clip on the inlet tubing was released, initiating rapid flow of permeabilization solution into the chamber (see section 2.3.1 of Chapter 2). The image that was acquired immediately prior to the start of the flow was considered to be time zero (start 106 point). In total, I analyzed the kinetics of 42 captured structures, all of which were filamentous in shape (Figure 3.5). Some filaments stopped almost immediately after the flow while others continued to move for many seconds before stopping. The average stop time for thirty-seven captured filaments was 11 seconds (minimum=4 seconds; maximum=36 seconds). It was not possible to estimate the stop time for the other five filaments because of focus shifts during the flow of perfusate (see section 2.3.2 of Chapter 2). The difference in the time for the filaments to stop may reflect variability in the time taken for the axons to become permeabilized. However, the average time for the filaments to stop (11 seconds) was much shorter than the time taken to completely permeabilize the plasma membrane as judged by loss of fluorescent dextrans (up to thirty seconds; see previous section). This suggests the transport machinery was disturbed very early during the permeabilization. It could be that some small soluble factors such as ATP diffused out of the axon immediately after saponin partitioned into the membrane lipid bilayer prior to the membrane being sufficiently permeable to allow the Mw. 10,000 fluorescent dextran to diffuse out of axons.

Of the 42 filaments that I analyzed, 25 stopped with both their leading and trailing ends visible within the gap. Of the other 17 filaments, 14 stopped before the trailing end of the filament had entered the gap and 3 stopped after the leading end had moved out of the gap. In 9 cases, a portion of the captured filament was observed to jiggle laterally, which I attribute to Brownian motion. This probably represents the early sign of splaying and it indicates that some lateral mobility of neurofilaments occurs, even in 0.02% saponin. In 8 of these cases the jiggling was at the end of the filament and in one case it was in the middle. In all but one case, the end that was jiggling was the trailing end, not the leading end. In some cases, the free end was long enough to loop back on itself, giving the end of the filament the appearance of a barb or eye bolt after fixation (Brown, 1997).

107

Figure 3.5 Capture of a moving filament. The images shown here were selected from a time-lapse movie acquired at 4 second intervals. The white arrowheads mark the leading and trailing ends of the moving filament. The filament moved rapidly into the gap and then paused. The flow of permeabilization solution was started during the pause. The filament resumed movement after the start of the flow but stopped before exiting the gap. Proximal is right and distal is left. Scale bar=6µm.

108 It has shown previously that about 70-80% of the moving filaments in cultured SCG neurons move anterogradely (Wang et al., 2000), but in the present study I chose to capture similar numbers of anterograde and retrograde captured filaments. Nineteen of the captured filaments were moving anterogradely at the time of capture and twenty-three were moving retrogradely. Most of them continued moving in the same direction after starting the flow of perfusate until they stopped. One of them switched direction once after the perfusion, and one switched direction twice. The lengths of the anterograde filaments ranged from 2.1 to 23.8µm (mean=9.7µm, n=15) and the lengths of the retrograde filaments ranged from 2.6 to 36.0 µm (mean=16.2µm, n=18). All appeared to be diffraction- limited in width. I was unable to estimate the length of the other nine captured moving structures because both ends were never visible within the gap at the same time.

3.7 Attempted reactivation of movement after permeabilization

The mechanism by which permeabilization caused the moving filaments to stop is unknown. There are several possibilities. First, it could be due to the disassembly or disorganization of microtubules and/or microfilaments, which are potential substrates for neurofilament transport. It is known that microtubules and microfilaments are rather dynamic structures and any alteration in monomer- polymer equilibrium during the permeabilization might have disrupted polymer organization. Second, it could be due to disruption of the interaction between neurofilaments and their molecular motors. Third, it could be due to the fact that some soluble factors required for the translocation diffused out of the axon when the plasma membrane was permeabilized. In vitro studies of organelle transport using permeabilization models have suggested that loss of soluble factors (in most cases, it was ATP) was the essential factor that causes the membranous organelle movement to stop (Vale et al., 1985a; Vale et al., 1985b; Koonce and Schliwa, 1986; Koonce et al., 1986). Upon addition of ATP, these organelle 109 movements could be reactivated again. To test the hypothesis that the loss of ATP was the mechanism underlying the cessation of neurofilament movement in the perfusion chamber after permeabilization, I captured moving filaments in gaps and then tried to reactivate the movement by adding ATP. These experiments were not successful but they are described here because they may be useful information for future experiments.

To increase my chances of reactivating the movement, I tried to use the lowest possible concentration of saponin that could still permeabilize the axon fast enough to capture a moving filament in the middle of the gap. In this way the disturbance to the transport machinery was likely reduced to the minimum. To determine the lowest concentration of saponin that can permeabilize the axon efficiently, 10mg/ml sulforhodamine 101 (Mw 606.71, Sigma) was injected into the cytoplasm of the cells five days after the plating. Sulforhodamine was used because its molecular weight is very close to that of ATP (Mw 507.2). Three hours later the coverslips containing the microinjected cells were observed in the perfusion chamber as described in section 2.3.2 of Chapter 2. Normally I was able to find some fluorescence labeled membranous organelle moving along the axons from these injected cells. These organelles presumably incorporated the fluorescent dye via autophagocytosis or endocytosis. I took advantage of these moving membranous structures and tried to find the lowest concentration of saponin that could stop the movements within twenty seconds. Since saponin with concentration less than 0.004% cannot permeabilize the plasma membrane (Anthony Brown, personal communication), I tried saponin concentrations starting from 0.004% with a 0.002% step increase. When I used 0.004% and 0.006% saponin, it took around eighty (n=2) and forty seconds (n=2) respectively for the membranous organelle movements to stop, while the movement stopped in around twenty seconds after 0.008% saponin was flowed into the chamber

110 (n=1). These observations suggest that it might also be possible to stop neurofilament movement within twenty seconds if the axons were permeabilized using 0.008% saponin.

To further stabilize the transport machinery, I also included 10µM taxol (to stabilize microtubules) and 6.5 µM phalloidin (to stabilize microfilaments) in the permeabilization solution. Taxol-stabilized microtubule and phalloidin-stabilized microfilament have been demonstrated to be able to support organelle and neurofilament motility in vitro (Kron and Spudich 1986; Schroer et al., 1989; Shah et al., 2000). First I included 10nm ATP in the permeabilization solution to see if the moving filament continued the movement in the presence of ATP during the permeabilization. The movements stopped as usual (average time=6 seconds, n=3). The explanation for the cessation of movement in the presence of stabilized microtubules, microfilament and ATP is not clear. It could be that the interaction between neurofilaments and the transport carriers was disturbed during permeabilization. Thus I substituted 100µM AMP-PNP for 10nm ATP in the permeabilization solution. AMP-PNP is a non-hydrolysable analogue of ATP. It can “lock” the cargo onto the cytoskeleton (Lasek and Brady, 1985; Dabora and sheetz, 1988) and it is interchangeable with ATP (Koonce and Schliwa, 1986). Using this permeabilization solution I captured four filaments and all stopped within 20 seconds (average=7 seconds, n=4) after the start of perfusate. I observed these captured filaments for 15-20 minutes after the perfusion of 8- 10ml 2mM ATP in NPHEM into the chamber and all of them remained in their position. Thus I was unable to reactivate the movement in permeabilized axons in these few studies. Due to the fact that there are many undefined factors in the model and experimental approach, I didn’t explore the reactivation experiments further. However these data do suggest that the cessation of neurofilament movement in these permeabilized axons may not simply be due to the loss of ATP.

111 3.8 Resistance of the captured filaments to detergent

Membranous organelles transported along axons can also be tubular in shape (Kaether et al., 2000). So it is theoretically possible that the filamentous structures containing GFP-NFM are actually membranous tubules. To confirm that captured fluorescent structures were not membrane bound tubules containing neurofilament proteins, fifteen of the captured structures were observed in the chamber system on the microscope stage for more than ten minutes after the flow of permeabilization solution was stopped. All of them remained intact without any change in their shape or their position relative to the gap during the period of observation. In six experiments 2-3ml 0.5% Triton X-100 (6-9 chamber volumes) was perfused into the chamber after the filaments had stopped and the filaments were observed for a further 5 minutes. In all cases, the filaments remained and their appearance did not change. Given that 0.5% Triton X-100 is known to be sufficient to permeabilize membranous organelles (Lutter et al., 2000), this suggests that the moving filaments do not represent membranous tubules.

3.9 Immunofluorescence microscopy of captured filaments

If the moving filaments are single neurofilament polymers, they should contain similar amounts of neurofilament protein subunits. To test this hypothesis, I sought to use immunofluorescence microscopy to compare captured filaments with single and bundled neurofilaments in splayed preparations. For immuofluorescence microscopy, I initially captured the filament by perfusing 4% (w/v) formaldehyde directly into the imaging chamber, which stopped the movement within 10 seconds (data not shown), and then I processed the coverslip for immunostaining using a polyclonal antibody specific for NFL as described above (see section 2.6.1 of Chapter 2). All neurons on the coverslip stained poorly, including the captured filament, which suggested poor fixation.

112 The exact reason for the poor fixation is not clear. One of the possible explanations is that the coverslip was maintained in imaging medium when the fixative was flowed into the chamber. So the fixative might have reacted extensively with the protein in the medium instead of the cells. To obtain good immunofluorescence images of the captured filaments, I captured the moving filaments by first permeabilizing the cells using 0.02% saponin as described above and then fixed the cells by flowing 4% (w/v) formaldehyde into the chamber (see section 2.4.2 of Chapter 2). The captured filaments were then processed for immunofluorescence microscopy using NFL antibody. The staining along the captured filaments was very uniform and comparable to each other in the same experiment. Thus the permeabilization solution not only permeabilized the axons causing the movement to stop but also ensured that subsequent fixation of the captured filaments was adequate.

3.10 Comparison of the captured and splayed filaments

The captured filaments appeared to be similar to single splayed filaments in their brightness (Figure 3.6). To test this quantitatively, I analyzed the fluorescence intensity along the length of captured filaments and along the length of single and paired splayed filaments from sister culture dishes. For each batch of culture dishes, I typically captured 1-3 filaments (one captured filament per coverslip) and then processed two additional coverslips for splaying. To permit quantitative comparison of the fluorescence intensities of filaments from different culture dishes within each experiment, I stained the coverslips side-by-side to ensure identical treatment. To compare the intensity of single and bundled filaments, I took advantage of the fact that single filaments in the splayed preparations occasionally pair up along part of their length (Brown, 1997). As it has been noted previously, this co-alignment indicates that the filaments are sufficiently close to each other that they cannot be resolved in the light microscope and does not necessarily imply the existence of any interaction between the filaments

113 (Brown, 1997). I typically analyzed 10-15 unpaired filaments and 2-3 paired filaments on each of the two splayed coverslips (total of 20-30 unpaired filaments and 4-6 paired filaments). The number of paired filaments analyzed was far fewer than the number of unpaired filaments because the co-alignment of two filaments was relatively rare. In total, I analyzed 17 captured filaments and 213 splayed filaments (169 unpaired and 44 paired).

Figure 3.6 Comparison of captured and splayed filaments. (A) Immunofluorescence of a captured filament stained with a polyclonal antibody specific for NFL. (B) Immunofluorescence of several splayed filaments stained with a polyclonal antibody specific for NFL. The white arrowheads mark places where the filaments are paired. Note the increase in brightness. The captured and splayed filaments were from sister coverslips and were processed and imaged identically to ensure that their intensities can be compared (see section 2.6.1 of Chapter 2). Note that the captured filament is similar in brightness to the unpaired splayed filaments and less bright than paired splayed filaments. Scale bar=2µm.

114 All the gaps containing the captured filaments were devoid of neurofilaments except the captured ones. Eight of the captured filaments were moving anterogradely and the other nine were moving retrogradely. Thirteen of the captured filaments stopped with both ends in the gap. The length of the anterograde filaments ranged from 4.2 to 23.8µm in length (mean=11.1µm, n=6) and the length of the retrograde filaments ranged from 6.4 to 33.8µm (mean=21.0µm, n=7). In three cases, the captured filaments appeared to be shorter in the image of GFP fluorescence comparing to the image of the NFL fluorescence (Figure 3.7). This indicates that these three captured filaments had not incorporated GFP-NFM along their entire length. This is expected since the data in Figure 3.3 showed that GFP-NFM incorporation was not uniform along all neurofilaments.

Figure 3.7 A captured filament immunostained for NFL. (A) A captured filament in a gap, visualized by the intrinsic GFP fluorescence. (B) The same structure visualized by immunofluorescence microscopy using a polyclonal antibody specific for NFL. The white arrowheads mark the location of the ends of the filament based on the NFL immunofluorescence. Comparison of the length of the captured filament in A and B indicates that it had incorporated GFP-NFM along 65% of its length. Scale bar=5µm.

115 3.11 Statistical analysis

For practical reasons, it was only possible to get several captured filaments from each experiment. I processed the filaments from different experiments for immunofluorescence microscopy with a NFL specific antibody using identical conditions. NFL is present in every filament, and all single filaments have similar NFL content (Ching & Liem, 1993; Lee et al., 1993; Brown, 1997). If all unpaired filaments from different experiments could have been processed together and identically, they should have had comparable NFL immunofluorescence intensity, which should represent the smallest fluorescence unit on all coverslips. However, the NFL immunofluorescence intensity along unpaired filaments from splayed preparations varied significantly between different experiments. There are several factors that might lead to the variations in the immunostaining results from different experiments: antibody concentration, antibody incubation temperature, and time. In each experiment the antibody concentration, incubation temperature and time were all kept the same, it suggests that other unknown factors contribute to the variations in the NFL immunofluorescence intensity observed between different experiments. Regardlee of the reasons, the filaments processed in different experiments cannot be compared directly.

In order to make a statistical comparison between the captured filaments and splayed filaments, I combined the data from different experiments by normalizing all the fluorescence intensities for the captured, unpaired and paired splayed filaments from each experiment to the mean of the fluorescence intensity of the unpaired splayed filaments for that experiment, which in this case acted as the internal reference standard. Figure 3.8 shows three histograms, which summarize the quantitative data for all analyzed splayed and captured filaments. The combined data for the paired and unpaired splayed filaments yield a bimodal distribution that cannot be fit with a single Gaussian distribution (p<0.05, Kolmogorov-Smirnov GOF test, combined histogram not shown), but the separate data sets for paired and unpaired splayed filaments do each match a 116 Gaussian distribution (p>0.01 for both sets of data, Kolmogorov-Smirnov GOF test). The mean of the normalized data for paired neurofilaments (1.98±0.30, s.d.) was twice the mean of unpaired neurofilaments (1.00±0.15, s.d.) as expected (Brown, 1997). The mean of the normalized date for the captured filaments was 1.10±0.35 (s.d.), which is not significantly different from that of unpaired filaments (p>0.05, unpaired t-test), but different from paired filaments (p<0.05, unpaired t-test). These observations indicate that the NFL protein content of the moving filaments is not significantly different from that of single neurofilament polymers in splayed axonal cytoskeleton.

Compared to the histogram of the unpaired splayed filaments, the histogram for the captured filaments is broader, with three captured filaments partially overlapping with the histogram of paired filaments. Since neurofilament subunit stoichoimetry is not fixed for each neurofilament (see section 1.3 of Chapter 1), the variation of NFL content in individual filaments might account for the variation in NFL immunofluorescence intensity for both splayed and captured filaments. In addition there are two other factors that might account for the variations in fluorescence intensity among captured filaments. First, there might be partial folding or bending within the moving filament. A previous study in the Brown lab has reported that neurofilaments are capable of folding and unfolding during the transport (Wang and Brown, 2001). Second, folding or bending in the filament might occur during the cessation of movement. For example, when the moving filament stopped during permeabilization, not every part of the filament stopped at the same time, which might cause folding or bending within the filament since the neurofilament is such a flexible structure. Because of the diffraction limit of the light microscope, it was not possible to distinguish straight filaments from those with slight folding or bending.

117

Figure 3.8 Fluorescence intensity analysis and comparison. (A) Histogram of the normalized fluorescence intensities for 169 unpaired filaments from splayed axonal cytoskeletons. (B) Histogram of normalized fluorescence intensities for 44 paired filaments from splayed axonal cytoskeletons. The curves represent Gaussian fits of the binned data. (C) Histogram of normalized fluorescence intensities for 17 captured filaments. The mean normalized fluorescence intensity of the captured filaments is not significantly different from the unpaired filaments (p>0.05, unpaired t-test) but it is significantly different from the paired filaments (p<0.05, unpaired t-test).

118 3.12 Summary

In this study, I developed a rapid perfusion technique to capture moving GFP tagged neurofilament proteins within the gaps of neurofilament array. Filaments that moved into gaps were stopped by perfusing 0.02% saponin into the chamber, which permeabilizes the axonal plasma membrane. The moving filament stopped within seconds, but the mechanism responsible for stopping the movement is not clear. The captured filaments remained even after perfusion with 0.5% Triton X-100, which indicates that they are not membranous. To test the hypothesis that captured filaments were single neurofilament polymers, I visualized them by immuofluorescence microscopy using an antibody specific for NFL, which is present in all neurofilaments. Using quantitative digital imaging processing and analysis techniques, I compared the fluorescence intensity of the captured filament with unpaired and paired filaments from splayed cytoskeletal preparations of sister coverslips that were otherwise processed identically. Statistical comparisons demonstrated that the mean normalized fluorescence intensity of captured filaments was not significantly different from that of unpaired filaments, but it was significantly different from that of paired filaments. The data indicate that NFL protein content of moving filaments is not significantly different from that of single neurofilament polymers from the splayed axonal cytoskeleton. All these observation support that moving GFP tagged neurofilament proteins are single neurofilament polymers.

119

CHAPTER 4

ELECTRON MICROSCOPIC ANALYSIS OF THE STRUCTURE OF MOVING NEUROFILAMENTS

4.1 Introduction

Previously I demonstrated by quantitative fluorescence microscopy that moving GFP-tagged neurofilament proteins in cultured SCG neurons have similar NFL protein contents as single neurofilaments from detergent-splayed cytoskeleton preparations, which supports the hypothesis that neurofilament proteins are transported along axons as single assembled polymers. However fluorescence microscopy doesn’t provide any fine structure information about these moving neurofilament proteins. So far, electron microscopy is the only technique that provides extremely high resolution images and detailed information on ultrastructure. In this case, correlative light and electron microscopy is the gold standard to demonstrate that the moving neurofilament proteins observed by fluorescence microscopy are actually 10-nm neurofilament polymers. To accomplish that, I captured the moving neurofilament proteins within gaps in the neurofilament array using the rapid perfusion technique as described in the previous chapter and processed them for electron microscopy using ultra-thin and semi-thin sectioning. Then I relocated the captured filaments in the electron microscope and demonstrated that they were indeed 10-nm single neurofilament polymers.

120 4.2 Strategy and criteria for locating the captured filament

To demonstrate unequivocally that the captured filaments are indeed neurofilament polymers, I performed correlative light and electron microscopy. These studies were technically difficult because the axons that contained gaps were typically only 100-200nm in diameter, and they were contained entirely within the first two or three ultra thin sections to come off the resin block during sectioning. There were several challenges that needed to be overcome to ensure successful relocation of the captured filaments in the electron microscope.

First, the block alignment needed extra accuracy so that axon loss during the first few strokes in sectioning could be minimized. If the block face is not aligned perfectly with the edge of the diamond knife, the first few sections will be partial and it is quite possible that the axon of interest can be at the edge of some partial sections. When these partial sections are picked up on grids, the very edges of sections tend to fold on themselves, which will make it impossible to observe the axon in these sections. In reality, there are always some errors in the block alignment and it is very hard to get the complete sections for the first few cutting strokes. To make the block alignment more accurate and consistent, I usually trimmed the block face to a small trapezoid, with the largest side (usually the lower side) not exceeding 500µm, and centered the area of interest in the block face. Secondly, axons with the gaps were usually contained in more than one ultra-thin section. Thus preparation of serial sections containing the axons of interest was a prerequisite for the reconstruction of the gap and captured filament. To do that, I did serial sectioning. During sectioning, the small trapezoidal block face also proved to be very important for the formation of a straight unbroken ribbon, which is essential in keeping serial sections in sequence. The last but not least of the challenges in these studies was to relocate the captured filament in the electron microscope. As described earlier (see section 2.9.1 of Chapter 2), to assist me in locating the axon of interest, I

121 severed the axons around the gap after glutaraldehyde fixation and cleared them away from the area. I also positioned several nerve cell bodies in the vicinity of the gap to mark the position, which could be easily identified using low power magnification in the electron microscope.

I was able to relocate the gap using electron microscopy based on its location within the surrounding cell bodies, cellular debris or varicosities along the axon and also the absence of neurofilaments. As described in section 3.4 of Chapter 3, a significant number of neurofilaments from neurons expressing GFP-NFM were only partially labeled with GFP. Thereby I expected that in some cases the gaps as observed by electron microscopy might appear to be shorter than observed by fluorescence microscopy if the edge of the gap only contained neurofilaments that were partially labeled with GFP. After locating the gap, I used the following criteria to locate captured filaments. The captured filaments should be at the same corresponding location in the gaps as observed by fluorescence microscopy. However, given the observation using light microscopy, I expected that some captured filaments might appear to be longer in the electron microscope, but none of them should be shorter.

4.3 Ultra-thin sectioning

In the previous chapter, I captured the moving neurofilaments by permeabilizing the axon using 0.02% saponin. For electron microscopy, to obtain satisfactory preservation of ultrastructures inside the cells it would be more desirable to fix the neuron without permeabilization. Thereby initially I captured moving filaments by perfusing the chamber with 2% glutaraldehyde fixative in 150mM sodium phosphate, pH 7.4, which normally stopped the movement within 8 seconds (data not shown). I then processed the axons for transmission electron microscopy using silver/gold (80-100nm) ultra thin sections. Using this approach I successfully relocated three axons that contained the captured filaments in the

122 electron microscope, but everything was packed very densely in these thin axons compared to thicker axons (compare Figure 4.1 A and B). I could unambiguously locate the gap with the help of varicosities and cellular debris along the axon, but I couldn’t unambiguously trace the captured filament, which was presumably obscured by the densely packed microtubules and other membranous structures.

To facilitate detection of neurofilaments I captured the moving filaments by permeabilizing the axon with 0.02% saponin as I did for the immunofluorescence microscopy studies described in Chapter 3, except that I substituted 5mM calcium chloride for the EGTA in the permeabilization solution to depolymerize axonal microtubules (Black et al., 1984; O’Brien et al., 1997). Note that the perfusion solution contained a mixture of protease inhibitors to prevent calcium- dependent proteolysis. The axon was then fixed using 2% glutaraldehyde and processed for transmission electron microscopy using ultrathin sectioning. Using this approach, microtubules disassembled and most organelles diffused out of the axon exposing the neurofilaments. Most of the plasma membrane still remained (Figure 4.1C), which is consistent with the data from previous studies showing that saponin peameabilizes the plasma membrane by creating membrane holes without stripping the membrane away (Seeman, 1967).

In total, I successfully located three captured filaments using this approach. These filaments measured 5.5µm, 6.5µm and 13.2µm in length in the images of the GFP fluorescence. Two of the filaments were moving anterogradely and one was moving retrogradely. In each case electron microscopy confirmed that the gap was devoid of neurofilaments except for a single filament in the same location as the captured filament. However, in no case was the entire axon or filament contained within a single section. Two of the axons were contained within two sections and one was contained within three. In all three axons, the captured filament traced a serpentine course, passing in and out of each section multiple times. Figure 4.2 shows an example of one of these filaments, which

123 was moving retrogradely at the time of capture. I attempted to serially reconstruct the filaments in adjacent sections, but encountered difficulty in aligning the images due to differential shrinkage or expansion of the sections during processing and observation. In one case, the serial reconstruction of the captured filaments was unsuccessful mainly due to the fact that in one of the serial sections the gird bar obscured most of the gap.

4.4 Semi-thin sectioning

To avoid the need for serial reconstruction of the captured filament in ultra-thin sections, I switched to blue-green (200-300 nm) semi-thin sections so that the entire axon would be contained in one section. Three gaps were successfully relocated in the electron microscope using this approach, but the captured filaments were generally obscured along portions of their length by the membranous elements in these thicker sections, especially where there was a varicosity (Figure 4.3 A). To overcome this problem, I treated the axons with 0.5% Triton X-100 after glutaraldehyde fixation. Triton X-100 treatment removed most of the plasma membrane and internal membrane, leaving an axonal “ghost” composed primarily of neurofilaments (Figure 4.3 B and C). Since varicosities were no longer visible in these axons, I had to rely on cellular debris surrounding the axon to identify the gap and locate captured neurofilaments.

Using this approach, I successfully captured and located two anterogradely moving filaments in the electron microscope. Each axon was contained entirely within one section and the captured filaments could be traced continuously along their entire length. In both cases the gap was devoid of neurofilaments except for a single filament in same place as the captured filament. Both filaments measured longer by electron microscopy than was apparent by fluorescence microscopy. One of the filaments (shown in Figure 4.4) measured 5.6µm in the image of the intrinsic GFP fluorescence and 8.9µm in the electron micrographs,

124 which indicates that it had incorporated GFP-NFM along 63% of its length. The other captured filament (not shown) measured 6.6µm in the image of the intrinsic GFP fluorescence and 8.9µm in the electron micrographs, which indicates that it had incorporated GFP-NFM along 74% of its length. In both cases, the non- fluorescent parts of these two filaments happened to be located at their trailing ends.

4.5 The diameter of captured filaments

To confirm that the captured filaments were neurofilament polymers, I measured their diameters at 20-35 evenly spaced points along their length and also performed similar measurements on filaments in regions flanking the gaps. The average diameter of the three captured filaments was 10.2±0.7nm in the ultra- thin sections and 10.5±0.8nm in the semi thin sections. In comparison, the average diameter of filaments flanking the gaps was 10.4±0.8nm in the ultra-thin sections and 10.3±0.8nm in the semi-thin sections. I estimate that the calibration of the magnification in the electron microscope was accurate to within ±10%. Thus the captured filaments were single, continuous 10nm-diameter neurofilament polymers.

125

Figure 4.1 The pre-fixation extraction technique. (A) A relative thick axon and (B) a relatively thin axon both fixed directly using 2% glutaraldehyde without pre- extraction. Note that the neurofilaments in (A) can be traced easily and the arrowheads pointed to two examples. Unfortunately these thick axons do not exhibit gaps. In (B) the membrane and microtubules were densely packed in the axon and a short neurofilament as pointed by the arrow heads can be traced but no other neurofilaments could be found, which were presumably obscured by the microtubules and membranous structures. (C) A thin axon permeabilized using 0.02% saponin including 5mM calcium and then fixed with 2% glutaraldehyde. Note that there were no microtubules left in the axon and neurofilaments pointed by arrow heads could be traced unambiguously along their lengths. Scale bar=300nm.

126

Figure 4.2 Electron microscopy of a captured filament using ultra-thin sectioning. (A) GFP fluorescence of a permeabilized axon before fixation, showing a captured filament in a gap. This filament was moving retrogradely at the time of capture. Proximal is left and distal is right. (B) Phase contrast image of the same axon. The blebs along the axon are a consequence of the saponin treatment. (C) Schematic line drawing of the axon and neurofilaments. The filaments and axon were traced from a montage generated from the digitized electron micrographs. The entire axon was contained within 3 ultra-thin (silver- gold) sections. The section shown here was the first to come off the block (i.e. the section immediately adjacent to the glass coverslip). The neurofilaments are represented by the red lines and the axon outline is represented by the black lines. The red lines have been thickened to make the location of the filaments visible, so their thickness is not to scale. The shaded regions represent cellular debris, also visible in the phase contrast image. The varicosities and cellular debris facilitated alignment of the fluorescence image with the electron micrographs. The arrows in A, B and C mark the ends of the filament. (D,E,F) Electron micrographs of the three boxed regions shown in C. Note the two segments of the captured filament marked by white arrowheads shown in D, the absence of neurofilaments in the gap shown in E, and presence of several filaments in the region flanking the gap shown in F. The membrane is largely intact, which indicates that saponin perforates the membrane without stripping it away. Note that microtubules are absent from these images because we destabilized axonal microtubules by including 5mM calcium in the permeabilization solution in order to permit unambiguous detection of the captured neurofilament. (A,B) Scale=8µm. (C) Scale bar=2µm. (D,E,F) Scale bar=150nm.

127

Figure 4.2

128

Figure 4.3 The post-fixation extraction technique. (A) Electron micrograph of an axon containing the captured filament. The axon was permeabilized with 0.02% saponin, fixed with 2% glutaraldehyde and then processed for electron microscopy without any post-fixation extraction. Note that parts of the filament pointed by the arrow heads can be identified unambiguously and the part of the filament in the big varicosity pointed by the arrows cannot be identified unambiguously due to cloudy membranous structures. (B,C) Electron micrographs of axons processed identically as the one in (A) except that these axons were post-extracted with 0.5% triton X-100 after glutaraldehyde fixation. Note that there is not much membrane structures left along the axon and the neurofilaments can be traced unambiguously. Scale bar=200nm.

129

Figure 4.4 Electron microscopy of a captured filament using semi-thin sectioning. (A) GFP fluorescence of a permeabilized axon before fixation, showing a captured filament in a gap. This filament was moving anterogradely at the time of capture. Proximal is left and distal is right. (B) Phase contrast image of the same axon. The blebs along the axons are a consequence of the saponin treatment. (C) Schematic line drawing of the axon and neurofilaments. The filament was traced from a montage generated from the digitized electron micrographs. The entire axon was contained within 1 semi-thin (blue-green) section which was the first to come off the block (i.e. the section immediately adjacent to the glass coverslip). The neurofilament is represented by the red line. The axon membrane was extracted (see below), so the outline of the axon was drawn artificially based on the phase contrast image and is represented by the black dashed lines. The red line has been thickened to make the location of the filaments visible, so its thickness is not to scale. The shaded region represents a large piece of cellular debris, also visible in the phase contrast image, which facilitated alignment of the fluorescence image with the electron micrographs. The arrows in A, B and C mark the ends of the filament. One end of the filament extended further in the electron micrographs than in the image of the GFP fluorescence, indicating that this neurofilament had not incorporated GFP-NFM along its entire length. (D) Electron micrograph of the boxed region shown in C. The axon membrane is not present because we treated with 0.5% Triton X-100 after glutaraldehyde fixation to permit unambiguous detection of the neurofilament in these semi-thin sections. Note the continuity of the captured filament, compared to the discontinuity of the captured filament in the ultra-thin section in Figure 3.10. (A,B) Scale bar=8µm. (C) Scale bar=2µm. (D) Scale bar=100nm.

130

Figure 4.4

131 4.6 Summary

In this chapter, the structure of moving neurofilament proteins was further analyzed by correlative light and electron microscopy. First, the moving structures were captured using the rapid perfusion technique and processed for electron microscopy using ultra-thin sectioning. These axons were usually 100- 200nm thick and contained within two or three 80-100nm ultra-thin sections. Only single neurofilament polymers could be found in the gaps at locations that corresponded to the captured filaments. The captured filaments appeared to trace a serpentine course in the axon, passing in and out each section several times. To further demonstrate that captured filaments are continuous 10-nm single neurofilament polymers, I processed the captured filaments for electron microscopy using semi-thin sections. Using this approach, the axons were contained within one 200-250nm semi-thin section and I found that the captured filaments were indeed continuous neurofilament polymers that measured 10-nm in diameter.

132

CHAPTER 5

POLYPEPTIDE COMPOSITION OF MOVING NEUROFILAMENTS

5.1 Introduction

The data in the previous two chapters demonstrate that neurofilament proteins are transported as assembled polymers in cultured SCG neurons. But what is the mechanism of movement and how is it regulated? To address these questions, it is very important to know the polypeptide composition of the moving neurofilaments. However, polypeptide composition of neurofilaments in cultured SCG neurons has not been previously characterized. In this chapter, I analyzed the polypeptide composition of neurofilaments in these neurons and determined whether there was a difference between moving and pausing neurofilaments in this respect.

5.2 Strategy

As described earlier (see details in section 1.3 of Chapter 1), NFL, NFM, NFH, peripherin and α-internexin are the intermediate filament proteins that are expressed in differentiated neurons. Thus I chose to analyze the distribution of these five polypeptides along the neurofilaments in cultured SCG neurons from neonatal mice by immunofluorescence microscopy using corresponding specific antibodies. To do this, I first characterized the specificity of the antibody for each polypeptide using western blotting. For each specific antibody as determined by

133 the western blotting, its specificity and quality for immunostaining was further confirmed using SW13 cell systems. Surprisingly, even though there are quite a few NFH antibodies that are commercially available or widely used, I did not find an NFH antibody that met the criteria for specificity and quality of staining. Thereby I was not able to determine NFH distribution along neurofilaments in these neurons.

To characterize the polypeptide composition of the overall neurofilament population, most of which are pausing at any point in time, I took advantage of the splaying technique developed by Brown (Brown, 1997) and analyzed the splayed filaments by immunofluorescence microscopy using various antibodies that had been characterized as described above. To determine whether there was a difference in the polypeptide composition between moving and pausing neurofilaments, I characterized the distribution of each polypeptide along the moving filaments in exactly the same way as described for splayed filaments. To get the moving filaments, I took advantage of the perfusion chamber system that I developed earlier (see Chapter 3). To visualize neurofilaments, the neurons were transfected with GFP-tagged mouse NFM by nuclear injection two days after plating. Neurons expressing GFP-NFM were observed in the perfusion chamber and the moving filaments were captured as they were moving through gaps in the neurofilament array by permeabilizing the membrane using 0.02% saponin (see section 2.4.1 of Chapter 2).

5.3 Characterization of the specificity of the antibodies using Western blotting

To confirm the specificity of the antibodies that I used for characterizing the polypeptide composition of neurofilaments I performed immunoblotting using cytoskeletal protein fractions prepared from mouse spinal cord and sciatic nerve. Previous studies have shown that native neurofilaments purified from these

134 tissues are phosphorylated at numerous epitopes (Uchida et al., 1999). To test the dependence of the antibody binding on the phosphorylation state, I dephosphorylated the proteins after blotting by incubating the PVDF membrane strips with alkaline phosphatase (see Section 2.7.3 of Chapter 2). Normally dephosphorylation induces a motility shift for proteins, but I performed the alkaline phosphatase treatment after the proteins had been separated by SDS- PAGE and thus there was no motility shift after dephosphoryation.

Figure 5.1 shows the results of the immunoblotting. Mouse monoclonal antibody RMO55 bound only to a band of apparent molecular weight 150kD on the untreated membrane strip (Figure 5.1E), which corresponds to NFM, but didn’t recognize anything on the enzymatically dephosphorylated strip (Figure 5.1F). This is consistent with previous reports that RMO55 is specific for phosphorylated NFM (Lee et al., 1987; Brown, 1998) and it confirms that the dephosphorylation treatment of the membrane is efficient under the conditions used here. The rabbit NFL polyclonal antibody, NFLAS, only bound to a 68kD band, which corresponds to NFL (Figure 5.1A) and this was not affected by enzymatic dephosphorylation (Figure 5.1B). The data are consistent with previous reports that NFLAS antiserum is specific for NFL and binds independently of phosphorylation state (Trojanowski et al., 1989; Brown, 1998).

Mouse monoclonal antibody RMO270 bound specifically to both native (Figure 5.1C) and enzymatically dephosphorylated NFM without significant difference (Figure 5.1D). This indicates that RMO270 binds to NFM independently of its phosphorylation state, which is consistent with a previous report for this antibody (Lee et al., 1987). Mouse monoclonal antibody 7C5 bound only to a 58kD band, which corresponds to peripherin, on both untreated and enzymatically dephosphorylated membrane strips (Figure 5.1 G & H), with slightly stronger staining for the dephosphorylated protein. These data suggest that 7C5 antibody has higher affinity for dephosphorylated peripherin. Rabbit polyclonal antibody

135

Figure 5.1 Characterization of the specificities of antibodies by immunoblotting. The neurofilament preparation in PVDF blot strips (A-H) was from sciatic nerve and that in the blot strips (I, J) was from spinal cord. The blot strips were stained with NFLAS (A, B), RMO270 (C, D), RMO55 (E, F), 7C5 (G, H) and alphaBB (I, J). The blot strips marked P (A, C, E, G, I) represent the untreated neurofilament preparation and the blot strips marked dP (B, D, F, H, J) represent the same protein preparation after enzymatic dephosphorylation with alkaline phosphatase on the membrane prior to immunostaining. The arrow heads mark the approximate locations of pre-stained marker proteins with molecular masses of 150, 70 and 50kD and the arrows mark the approximate locations of NFL, NFM, peripherin and alpha-internexin.

136 αBB reacted only with a 66kD band, corresponding to α-internexin (Figure 5.1 I & J). Comparison of the antibody staining between untreated and emzymatically dephosphorylated strips indicates that this antibody is weakly dependent on phosphorylation, or that it might be mixture of phosphorylation-dependent and phosphorylation-independent antibodies against α-internexin.

5.4 Characterization of the specificity of the antibodies using SW13 cl.2 Vim- cell system

Immunoblotting examines antibody specificity for denatured proteins, which might not reflect the antibody specificity for proteins in their native state. To confirm the specificity of each antibody for the corresponding polypeptide in cells using immunofluorescence microscopy, I took advantage of the SW13 cl.2 Vim- cell line, which is derived from human adrenal carcinoma and lacks its own cytoplasmic intermediate filament network due to a silencing of vimentin expression. By transient expressing different polypeptides in these SW13 cells, I constructed filamentous networks with different polypeptide compositions in these cells and use them to assess the specificity of each antibody.

For example, to confirm the specificity of NFLAS for NFL, I transfected SW13 cl.2 Vim- cells with two different combinations of expression vectors and compared NFLAS staining of cells transfected under different conditions. Under one condition, I transfected the cells with a mixture of five different plasmids including pmNFL, pmNFM, pCI-NFH, pSRV-α and pBluescript-peripherin, which encode mouse NFL, mouse NFM, rat NFH, rat α-internexin and human peripherin respectively. Under the other condition, I transfected the cells with the identical plasmids except the one encoding NFL so that each of the polypeptides except NFL would be expressed in some of the cells. All transfections were carried out using Lipofectamine 2000 (Invitrogen) (see Section 2.2.3 of Chapter 2). Twenty- four hours after transfection, the coverslips were processed identically for

137 immunofluorescence microscopy using NFLAS. If NFLAS is specific for NFL and doesn’t cross-react with other neuronal intermediate filament polypeptides, one would expect that only some of the cells transfected under the first condition would have NFL immunofluorescence while none of the cells transfected under the second condition would have NFL immunofluorescence.

Considering the large number of plasmids that were transfected under each condition, the lack of NFL immunofluorescence could also be due to low transfection efficiency. Thus, I first verified that each polypeptide was expressed in a significant number of cells that were transfected under conditions as described above by examining the transfection efficiency for each polypeptide under the first transfection condition. To do that, coverslips were fixed twenty-four hours after transfection and processed for single-labeling immunofluorescence microscopy using antibodes NFLAS, RMO270, AB1989, 7C5 and αBB respectively. Each coverslip was randomly scanned by both phase contrast and fluorescence microscopy using a 100x objective lens. For each field of view, I first counted the actual number of cells using phase contrast. Then I switched to fluorescence and counted the number of cells with fluorescence labeling in the same field of view. Normally in total I scanned 20-30 different areas on each coverslip and examined around five hundred cells (usually the cells were not confluent). To calculate the transfection efficiency for each polypeptide, I divided the number of cells with each antibody immunofluorescence by the total number of cells examined. The transfection efficiency for NFL, NFM, NFH, peripherin and α-internexin was 43.63%, 32.75%, 3.62%, 1.94% and 29.37% respectively. Considering that the cells were plated at density of 105/cm2, each polypeptide was expressed in hundreds even tens of thousands of cells on each coverslip even though most of the cells might not express all polypeptides that have been transfected. Since these numbers were estimated using coverslips that had been co-transfected with all five expression vectors, for those co-transfected under the second condition, I expected the transfection efficiency would be even higher for

138 corresponding polypeptide than the number described above. Thus if NFLAS cross-reacted with any other neuronal intermediate polypeptide, the NFLAS immunofluorescence should be easily encountered on the coverslips transfected under the second condition.

After confirming the transfection efficiency for each transfected polypeptide, I compared NFLAS staining pattern of the coverslips transfected under different conditions. To do this, I first scanned the whole coverslips using 10x objective, which took only tens of minutes for each coverslip. If a cell had NFL immunofluorescence, it could be easily detected using this low power objective. I saw a lot of fluorescent cells on the coverslip transfected under the first condition, but not a single fluorescent cell on the coverslip transfected under the second condition, which was then further confirmed using 100x objective. Figure 5.2A shows rabbit polyclonal antibody NFLAS stained-cells, in which plasmid encoding mouse NFL was co-transfected with the other four expression vectors encoding NFM, NFH, peripherin and α-internexin. Figure 5.2B shows NFLAS stained- cells, in which all identical plasmids except the one encoding NFL were co- transfected.

The specificity of other three antibodies (RMO270, 7C5 and αBB) was confirmed in the same way. These three antibodies only stained the cells that were co- transfected with all five expression vectors (Figure 5.2C, E and G) and didn’t stain any of the cells that were transfected with vectors encoding all other polypeptides except the one corresponding to the testing antibody (figure 5.2D, F and H). These data further confirm the specificity of NFLAS, RMO270, 7C5 and αBB for NFL, NFM, peripherin and α-internexin respectively, consistent with the data obtained by immunoblotting.

139

Figure 5.2 Characterization of the specificity of the antibodies using SW13 cl.2 Vim- cells. SW13 cl.2 Vim- cells (A, C, E and G) co-transfected of five expression vectors including pmNFL, pmNFM, pCI-NFH, pBluescript-peripherin and pSRV-α, which encode mouse NFL, mouse NFM, rat NFH, human peripherin and rat α-internexin respectively, were stained with NFLAS (A), RMO270 (C), 7C5 (E) and alphaBB (G). SW13 cl.2 Vim- cells (B, D, F and H) co- transfected with all the expression vectors mentioned earlier except pmNFL (B), pmNFM (D), pBluescript-peripherin (F) and pSRV-α (H), were stained NFLAS (B), RMO270 (D), 7C5 (F) and alphaBB (H). All cells stained with a particular antibody were processed identically and the images were acquired with identical exposures and scaled identically to permit visual comparison of the staining intensities. Note the difference in the staining pattern within each paired images. The antibody only stained these cells that had been transfected with the expression vector encoding the corresponding polypeptide. Scale bar=15µm.

140 5.5 Characterization of the staining quality of the antibodies using SW13 cl.2 Vim- cell system

Even if an antibody is specific for a particular neurofilament polypeptide, it might not stain uniformly along the filaments if the epitope is not uniformly accessible or is masked when the polypeptide is in certain conformations or as a result of formaldehyde fixation. To ensure that the antibody staining reflects the real distribution of polypeptides along neurofilaments, I examined the staining quality of each antibody using SW13 cl. 2 Vim- cell line.

To characterize the staining quality of NFLAS and RMO270, I co-transfected SW13 cl.2 Vim- cells with plasmids pmNFL and pmNFM. Because neither mouse NFL nor NFM can form homopolymers by themselves in vivo (Gill et al., 1990; Wong et al., 1990; Ching and Liem, 1993; Lee et al., 1993), they need to interact with each other to form 10nm-filaments. Thus if the antibody staining quality is good, one would expect that the filaments found in these transfected cells would be decorated continuously by both antibodies. Twenty-four hours after transfection, the cells were fixed and processed for double immunostaining using NFLAS and RMO270. Figure 5.3 (panel A and B) shows the uniform NFLAS and RMO270 staining the filamentous networks formed by exogenous protein NFL and NFM.

Because both peripherin and α-internexin are capable of forming homopolymers in vivo (Ching and Liem, 1993; Cui et al., 1995), I transfected SW13 cl.2 Vim- cells with expression vector pBluescript-peripherin or pSRV-α alone and processed them identically for immunofluorescence microscopy using the corresponding antibody 7C5 or αBB. 7C5 staining of the filamentous networks in the cells expressing human peripherin was very uniform (Figure 5.3C), as was the αBB staining of cells expressing rat internexin (Figure 5.3D). These data demonstrate the epitopes of these antibodies are uniformly accessible along the entire length of these filaments. 141

Figure 5.3 Characterization of the quality of the antibody staining using SW13 cl.2 Vim- cells. A SW13 cl.2 Vim- cell cotransfected with pmNFL and pmNFM and double stained with NFLAS (A), specific for NFL, and RMO270, specific for NFM (B). SW13 cl.2 Vim- cells transfected with pBluescript-peripherin (C) and pSRV-α (D) were stained with 7C5, specific for peripherin, and alphaBB, specific for α-internexin, respectively. Note the uniformity of the staining for each antibody. Scale bar=3µm.

142 5.6 Characterization of NFH antibodies

In total, I have screened eight different antibodies to NFH. To my surprise, I have been unable to find an NFH antibody that met the above criteria for specificity and quality of staining. Normally I tested each NFH antibody in a series of steps. First I examined the quality of the antibody by staining untreated (i.e. not splayed) cultured SCG neurons and splayed cytoskeleton preparations from sister cultures. If the antibody staining along the axon and the splayed filaments was very weak so that it could not be distinguished from the background. I would stop pursuing any further characterization of the antibody since the interpretation of these weak fluorescence signals would be difficult. Antibodies N52, 3G3 and NFHAS exhibited weak staining according to these criteria and were therefore not pursued further (see Table 5.1). Antibody R49 exhibited punctate staining. But on the same dishes, I also found that this antibody stained non-neuronal cells, which could be encountered occasionally in these neuron cultures, with comparable fluorescence intensity as that of the neurons. This suggested that this antibody was not specific and thus I did not carry out any further studies with R49. Antibodies RMO217, AB1989, R14 and CPCA-NFH stained both axons and the splayed filaments strongly and their specificities were then examined in cultured SCG neurons from NFH knock-out mice. The NFH knock-out mice were available in the lab for another project and they were kindly provided by Don W. Cleveland (Rao et al., 1998). In this study, I took advantage of these mice to confirm the specificity of NFH antibodies. To do this, SCG neurons from wild type and NFH knock-out mice were cultured, then fixed five days after plating and processed for immunofluorescence microscopy identically. Surprisingly, in all four cases there was no significant difference in the staining of the wild type and NFH knock out mouse neurons for these antibodies (see table 5.1), which suggests that these antibodies cross react with some other proteins in these neurons.

To further investigate the non-specificity of these antibodies and find out the polypeptide(s) that they might cross-react with, pmNFL was transfected into 143 SW13 cl.1 Vim+ cells (which express vimentin) alone or cotransfected with pmNFM or pmNFM and pCI-NFH into SW13 cl.2 Vim- cells (which do not express vimentin). Antibodies RMO217, R14 and CPCA-NFH not only stained SW13 cl.2 Vim- cells co-transfected with all expression vectors that encode neurofilament triplet proteins respectively but also strongly recognized the filamentous network in cells co-transfected with pmNFL and pmNFM. But they did not recognize the filamentous network in the SW13 cl.1 Vim+ cells that were transfected with pmNFL alone. In these SW13 cl.1 Vim+ cells, NFL coassembled with the endogenous vimentin network. The expression of NFM and NFL in these transfected cells were confirmed by the immunofluorescence of RMO270, specific for NFM, and NFLAS, specific for NFL. These data suggest these three antibodies were cross-reacting with NFM. Figure 5.4 shows an example of the staining results obtained for RMO217.

AB1989 antibody was perhaps the most promising antibody that I encountered, however this antibody also displayed problems. AB1989 staining of SW13 cl.2 Vim- cells transfected with pmNFL and pmNFM was quite similar to that of SW13 cl.2 Vim+ cells transfected pmNFL alone and even untransfected SW13 cl.2 Vim- cells (e.g. Figure 5.5C), but they all were quite different from the uniform staining pattern of those cells transfected with all three neurofilament triplet proteins (Figure 5.5B). In these cells that were not transfected with pCI-NFH, AB1989 stained some filamentous network but the width of the filaments varied along their length. In addition, the AB1989 immunofluorescence intensity along these filaments was comparable to the fluorescence intensity of neurofilament network in SW13 cl.2 Vim- cells that were transfected with neurofilament triplet proteins and stained with AB1989 (compare Figure 5.5B and C). When I compared the AB1989 immunofluorescence in untransfected SW13 cl.2 Vim- cells with their phase contrast images, I noticed a correlation between the AB1989 staining pattern and some tubular structures (presumably membranous) (compare Figure 5.5C and D). In contrast, such correlation was not observed with SW13 cl.2 Vim-

144 cells that were transfected with neurofilament triplet proteins and stained with AB1989 or SW13 cells that were transfected under various conditions and stained with other NFH antibodies. It appeared that AB1989 staining of these tubular structures in SW13 cells only existed in the absence of NFH. One possible explanation for this is that AB1989 affinity for some protein in these tubular structures increased in the absence NFH. In other words, because NFH was expressed in SW13 cl.2 Vim- cells that were transfected with neurofilament triplet proteins, AB1989 affinity for the tubular structures was decreased and thus they only recognized the neurofilament network in these cells formed by neurofilament triplet proteins. Thereby I reasoned that AB1989 staining of these tubular structures was not some usual non-specific staining that might be blocked by using different blocking solution and I did not explore different blocking solutions with this antibody. What AB1989 bound to in these membranous structures is not clear. Nevertheless, these data call into question the specificities of this antibody and its suitability for quantitative immunofluorescence microscopy. It is quite possible that AB1989 staining in the cultured neurons from NFH knock-out mouse might also be due to some membranous proteins.

All eight antibodies that I have screened are either commercially available or widely used. It is quite surprising that none of them met my criteria for this polypeptide composition study and thus I was not able to determine NFH distribution along individual filaments. However in an independent study, NFH hasn’t been able to be detected by western blotting in five-day-old cultured SCG neurons (Gulsen Colakoglu, Brown lab, personal communication). This is consistent with other reports of a delay in onset of NFH expression in optic nerve during development and in the cultured dorsal root ganglion neurons (Shaw and Weber, 1982; Pachter and Liem, 1984; Athlan et al., 1997).

145 Antibody Staining Staining along Staining in SCG Staining in SW13 Staining in SW13 Staining in SW13 name along the splayed neurons from NFH cl.2 Vim- cells co- cl.2 Vim- cells co- cl.1 Vim+ cells axon filaments of knock-out mice transfected with transfected with transfected with of SCG SCG neurons pmNFL, pmNFM and pmNFL, pmNFM pmNFL neurons pCI-NFH N52 Very weak No staining n.d. n.d. n.d. n.d. Staining RMO217 Strong and Strong and Strong and uniform Strong and uniform Strong and uniform No staining uniform uniform staining staining staining staining staining 3G3 Very weak Very weak n.d. n.d. n.d. n.d. Staining staining NFHAS Very weak Very weak n.d. No staining n.d. No staining Staining staining AB1989 Punctate Punctate Punctate staining Strong and uniform Strong and non- Strong and non- 146 (R12) Staining staining staining uniform staining uniform staining R49 Punctate Punctate n.d. n.d. n.d. n.d. Staining staining R14 Strong and Strong and Strong and uniform Strong and uniform Strong and uniform No staining uniform uniform staining staining staining staining staining CPCA- Strong and Strong and Strong and uniform Strong and uniform Strong and uniform No staining NF-H uniform uniform staining staining staining staining staining

n.d. = not determined

Table 5.1 Characterization of NFH antibodies

Figure 5.4 Characterization of NFH antibody RMO217. (A) Characterization of RMO217 by immunoblotting. The lane marked “M” contained the molecular weight marker. The protein preparation was obtained from rat spinal cord and the protein staining is shown in the lane marked “CBB” (Coomassie Brilliant Blue). The proteins were transferred onto PVDF membrane and stained with RMO217. The lane marked “RMO217” shows that this antibody is very specific for NFH (apparent molecular weight 190kD) by immunoblotting. This image is reproduced from Uchida et al. (1999). Axons in cultured SCG neurons from wild type (B) and NFH knock-out mice (C) were stained with RMO270. The axons were treated with high concentration of detergent (0.5% Triton X-100) before being fixed and neurofilament splaying could be observed along the axon, indicated by the arrow heads. Note there was significant AB1989 staining along the axons in both B and C. SW13 cl.2 Vim- cells co-transfected with pmNFL, pmNFM and pCI-NFH (D) or with pmNFL and pmNFM (E) were stained with RMO217. SW13 cl.1 Vim+ cells transfected with pmNFL alone and stained with RMO217 (F). Note the filament networks were stained by RMO217 in D and E but not in F, in which the NFL expression was confirmed by NFLAS staining that was not shown here. Scale bar=12µm.

147

Figure 5.5 Characterization of NFH antibody AB1989. (A, B) A SW13 cl.2 Vim- cell from transient co-transfection of pmNFL, pmNFM and pCI-NFH was double stained with RMO270 and AB1989. Note that AB1989 stained strongly and uniformly long the filaments. (C) An untransfected SW13 cl.2 Vim- cell was stained with AB1989. Note that AB1989 staining in C is quite different from that in B. (D) Phase contrast image of the cell in C. Arrow heads in C and D point to two examples of tubular structures, presumably membranous in nature that AB1989 recognized. AB1989 staining patterns of SW13 cl.2 Vim- cells transfected with pmNFL and pmNFM, and SW13 cl.1 Vim+ cells transfected with pmNFL alone were quite similar to (C). Scale bar=4µm.

148 5.7 NFM incorporation along the filaments

In this section and following sections, I will discuss my results for each of the four polypeptides examined (NFM, peripherin, α-internexin and NFL) with neurofilaments from cultured SCG neurons. To visualize and capture neurofilaments, I transfected the neurons with GFP-tagged NFM fusion proteins. To determine the contribution of the exogenous protein GFP-NFM to the NFM distribution along individual filaments, I compared the NFM distribution along splayed filaments from both transfected and untransfected cells. To do that, axonal neurofilaments were induced to splay apart from each other by permeabilization of neurons with 0.5% Triton X-100 in the presence of 0.19M NaCl 3 days after they were transfected by nuclear injection with the GFP-NFM expression vector or 5 days after plating without transfection. Since the neurons were transfected two days after plating, 3 days after transfection is equal to 5 days after plating. Then the neurofilaments from both transfected and untransfected cells were processed for double labeling immunofluorescence microscopy using the antibody NFLAS, specific for NFL, and antibody RMO270, specific for NFM (see section 2.6.1 of Chapter 2). The average proportion of the splayed filaments labeled with NFM immunofluorescence was 99.7% for transfected neurons (minimum=94.4%, maximum=100.0%; n=150), while for those splayed filaments from untransfected neurons, the average proportion was 99.2% (minimum=90.0%, maximum=100.0%; n=150). Figure 5.6F summarized the data for 150 splayed neurofilaments from untransfected cells processed in three independent experiments. These data suggest that on average exogenous GFP-NFM only increase the NFM immunofluorescence along the filament by 0.5% of the filament length, which is insignificant compared with average 99.2% of filament length that contained endogenous NFM.

To characterize the NFM incorporation along the captured filaments, the coverslips bearing the captured filaments were processed for double-labeling immunofluorescence microscopy using antibodies NFLAS and RMO270. Both 149 antibodies stained uniformly along the whole length of the captured filaments (Figure 5.6 A-C and A’-C’). The quantification of NFM incorporation along individual filaments was carried out as described in section 2.8.3 of Chapter 2. In total, I analyzed ten captured filaments, six of which were moving anterogradely before the movements stopped and four were moving retrogradely. Two of the analyzed captured filaments didn’t move completely into the gap before they stopped so it was impossible to measure their length, though they were longer than twenty microns. The parts of the filament that was overlapped with the edge of the gap were excluded from the analysis of the NFM distribution. The average length for the other eight filaments was 12.0µm (minimum=3.7µm, maximum=21.1µm; n=8). On average, the captured filaments had NFM incorporated along 99.8% (minimum=96.4%, maximum=100%; n=10, Figure 5.6E) of their length, which includes both endogenous and exogenous NFM.

5.8 Peripherin incorporation along the filaments

For the quantification of peripherin incorporation along the captured filaments, initially I processed the coverslips for immunofluorescence microscopy using NFLAS, specific for NFL, and 7C5, specific for peripherin, as described above. 7C5 staining along the captured filaments was very weak. Occasionally there were some splayed filaments elsewhere on these same coverslips since splaying can sometimes occur even in 0.02% saponin, but the 7C5 staining along these filaments was also very faint. This weak staining was different from the strong staining pattern that I observed with splayed preparations processed in the open dishes for immunofluorescence microscopy. What caused the difference is not clear. The coverslips recovered from the perfusion chamber used for capturing filaments and the coverslips from the open dishes used for splaying were processed identically for immunofluorescence microscopy except for the fixation. Cultures observed in the perfusion chamber were fixed by flowing 2-3ml fixative

150 into the chamber while the cultures in the open dish were fixed by adding 2-3ml fixative into the dish. So the different fixation condition could be the reason for the difference in the staining pattern of the filaments.

To compare the fixation conditions between open dishes and the perfusion chamber, sister coverslips from the same cultures were divided into four groups. Group one coverslips were treated with 0.5% Triton X-100 in open dishes to cause neurofilament splaying (see section 2.5.3 of Chapter 2) and then fixed for twenty minutes using 4% (w/v) formaldehyde. Group two coverslips were assembled into the perfusion chamber. Then 0.5% Triton X-100 was flowed into the chamber under gravity and the coverslips were fixed by perfusing 2ml 4% (w/v) formaldehyde to mimic a capture experiment. Fifteen minutes later, the chamber was disassembled and the coverslips were placed in fresh fixative for another 15-20 minutes. This condition was what I usually used to fix the coverslips in the perfusion chambers for immunofluorescence microscopy. Group three coverslips were processed in the same way as the group two except that there was no post-fixation after disassembly of the chamber. Group four coverslips were processed similarly to group three except that the chamber was disassembled five minutes after the flow of fixative and there was no post- fixation. Then I processed all the coverslips from these four groups identically for double labeling immunofluorescence microscopy using NFLAS and 7C5 antibody. I found that 7C5 staining of the coverslips from first group was consistently strong, but the staining of the coverslips from second and third groups was very weak. 7C5 stained the coverslips in fourth group much stronger than group two and three but weaker than group one. These data suggest that weak staining in the chamber system was caused by over-fixation instead of poor fixation of 7C5 epitopes.

To find a fixation condition in the chamber system that is comparable to what is normally used in open dishes (as described above, please also see section 2.5.3

151 of Chapter 2), which gives strong staining for the 7C5 antibody, I fixed splayed preparations in the perfusion chamber by perfusing 2ml of different concentrations (3%, 2% and 1% (w/v)) of formaldehyde as described above. Fifteen minutes later the chamber was disassembled, the coverslips were recovered and processed for immunofluorescence microscopy identically with those fixed in the open dishes without any post-fixation as described above. I chose fifteen minutes for fixation, because it was the time that it normally took for me to mark the axon that contained the captured filaments using the diamond- scoring objective marker before disassembly of the chamber. Comparing the 7C5 staining under these different conditions by visual inspection, the coverslips that were fixed in the perfusion chamber using 2% (w/v) formaldehyde for fifteen minutes were quite similar to those processed in open dishes using 4% (w/v) formaldehyde for twenty minutes. All these observations suggest that 7C5 epitope was sensitive to formaldehyde fixation and over fixation tended to destroy the epitope. Fixation in the perfusion chamber appeared to be more efficient than in open dishes using the same conditions, which is presumably due to thorough change of solution during the perfusion of fixative in the chamber. Since in this study I was interested in the relative peripherin incorporation along the filaments rather than its absolute amount, I didn’t make a quantitative comparison of 7C5 immunofluorescence intensity along individual filaments among the different fixation conditions.

To quantify the peripherin distribution along moving filaments using the optimal fixation condition, the neurofilaments were captured as described above and then they were fixed by flowing 2ml 2% (w/v) formaldehyde into the chamber. Fifteen minutes later, the coverslips recovered from the disassembled chamber were placed in PBS and later processed for double-labeling immuofluorescence microscopy as described above using NFLAS and 7C5 antibodies. Compared to the uniform staining pattern of NFLAS along these captured filaments (Figure 5.6 G-I), the 7C5 staining was slight less uniform but nevertheless largely continuous

152 (Figure 5.6 G’-I’). I quantified the peripherin incorporation (see details in section 2.8.3 of Chapter 2) along fifteen captured moving filaments, among which eight filaments were moving anterogradely and seven were moving retrogradely. The average length for twelve of the captured filaments was 8.2µm (maximum=21.2µm, minimum=4.1µm; n=12). It is impossible to measure the length of other three filaments because in each case one of their ends was overlapping with the edge of the gaps or folded on itself during the capture. All the folding and overlapping parts of the filaments were excluded from the analysis of peripherin incorporation. On average, all fifteen captured neurofilaments had peripherin incorporated along 96.8% of their length (maximum=100%, minimum=92.6%; n=15, Figure 5.6 K).

To make the comparison of the peripherin distribution between captured and splayed neurofilaments, I induced the axonal neurofilaments in the untransfected cultures to splay apart from each other in open dishes as described earlier and processed them for double labeling immunofluorescence microscopy using NFLAS and 7C5 antibodies (Figure 5.6 J). 7C5 staining along the splayed filaments was quite similar to the captured ones (compare Figure 5.6 J with Figure 5.6 G’-I’). In total I analyzed 150 unpaired splayed filaments from three different experiments and the average proportion of the filaments that contained 7C5 immunofluorescence was 97.8% (maximum=100%, minimum=86.3%; n=150, Figure 5.6 L). These data suggest that peripherin is incorporated along most of the length of individual moving and pausing neurofilaments from cultured SCG neurons.

5.9 Alpha-internexin incorporation along the filaments

Since αBB, the antibody specific for α-internexin, is a rabbit polyclonal antibody, it could not be paired with NFLAS for double labeling immunofluorescence microscopy. Given that neurofilaments in these cultured neurons contain NFM 153 along an average of 99% of their length (see above), NFM immunofluorescence was used for the measurement of neurofilament length in the analysis of α- internexin incorporation along captured and splayed filaments. The captured moving filaments and splayed filament preparations were obtained as described above and processed for double-labeling immunofluorescence microscopy using RMO270 and αBB antibodies. Both antibodies stained all filaments very uniformly along their length (Figure 5.6 M-P and M’-P’). Single filaments from splayed preparations were identified based on their NFM immunofluorescence and the proportion of each filament containing αBB staining was measured as described above. In total I analyzed 14 captured filaments and eight of them were moving anterogradely and six were moving retrogradely. The average length for nine of the captured filaments was 9.8µm (maximum=14.9µm, minimum=3.7µm; n=9) and the other five were not measured due to overlap of the filament with the edge of the gap or folding of the filament during the capture. These folding or overlapping parts of the filaments were not included for the analysis of α-internexin distribution along the filaments. On average, α-internexin was incorporated along 99.7% of the length for both captured and splayed neurofilaments (captured neurofilaments: maximum=100%, minimum=97.9%, n=14; splayed neurofilaments: maximum=100%, minimum=94.6%, n=150). These data are summarized in the histograms of figure 5.6 Q and R.

5.10 NFL incorporation along the filaments

Several studies have demonstrated earlier that NFL is present in every neurofilament (Ching and Liem, 1993; Lee et al., 1993; see details in section 1.2.2 of Chapter 1). In this study, I found that every filament labeled with anti- NFM or peripherin immunofluorescence had NFL incorporated along its whole length. Since every filament labeled with anti α-internexin also has anti-NFM staining along the whole length, it is clear that every filament in these neurons has NFL distributed along 100% of its length. 154

Figure 5.6 Characterization of the polypeptide distribution along captured and splayed neurofilaments. Captured neurofilaments in gaps double-stained respectively with NFLAS (A-C) and RMO270 (A’-C’), NFLAS (G-I) and 7C5 (G’- I’), and RMO270 (M-O) and alphaBB (M’-O’). Splayed neurofilaments double- stained with NFLAS (D) and RMO270 (D’), NFLAS (J) and 7C5 (J’), and RMO270 (P) and alphaBB (P’) as pointed by the arrow heads. (E, F) Histograms of NFM distribution along captured (E) and splayed filaments (F). (K, L) Histograms of peripherin distribution along captured (K) and splayed filaments (L). (Q, R) Histograms of α-internexin distribution along captured (Q) and splayed filaments (R). All filament contain all polypeptides along all or most of their length. Scale bar=3µm.

155 5.11 Summary

In this chapter, I demonstrated that neurofilaments in cultured SCG neurons contained at least four polypeptides: NFL, NFM, peripherin and α-internexin. On average each was present along >95% of the neurofilament length. Apparently there was no difference between moving filaments and the overall neurofilament population, most of which are pausing at any point in time. Thus these data show no evidence that moving and pausing neurofilaments differ in their polypeptide composition.

156

CHAPTER 6

PHOSPHORYLATION STATE OF MOVING NEUROFILAMENTS

6.1 Introduction

Although increased phosphorylation of neurofilament proteins has been correlated with slowing neurofilament transport in many studies, only two phosphoepitopes have been related directly or indirectly with regulation of neurofilament transport: RT97 and SMI36 (Ackerley et al., 2000; Sanchez et al., 2000; Ackerley et al., 2003). Both RT97 and SMI36 are located at the carboxyl terminal domains of NFH and NFM (Sternberger and Sternberger, 1983; Sanchez et al., 2000). The RT97 epitope is phosphorylated by Cdk5/p35 (Bajaj and Miller, 1997; Ackerley et al., 2003). Phosphorylation at this epitope has been correlated with a slowing of the axonal transport of neurofilament proteins in both optic nerve and cultured neurons (Jung and Shea, 1999; Jung et al., 2000; Sanchez et al., 2000; Ackerley et al., 2003, see details in section 1.5.4)). In the Ackerley’s study, they showed that this was due to a change in pause time. This suggests that phosphorylation at these epitopes regulates neurofilament pausing behavior. The SMI36 epitope may be phosphorylated by p42/p44 MAPK and SapK1b MAP kinase (Ackerley et al, 2000) and phosphorylation at this epitope has been suggested to be correlated with a slowing of neurofilament transport in cultured neurons by glutamate (Ackerley et al., 2000). In this study, I compared moving neurofilaments with the overall neurofilament population in cultured SCG neurons, most of which are pausing at any point in time, by immunofluorescence

157 microscopy using RT97 and SMI36 antibodies to see whether moving neurofilaments differ from the overall population in their phosphorylation state at these phosphoepitopes.

6.2 Strategy

To test the hypothesis that phosphorylation at the RT97 and SMI36 epitopes regulates neurofilament pausing behavior, I characterized the distribution of these two epitopes along the moving filaments and compared them with the splayed neurofilaments by immunofluorescence microscopy to see whether captured neurofilaments are hyperphosphorylated compared with the whole population. Since RT97 and SMI36 epitopes can also be generated by phosphorylation of microtubule associated proteins (MAPs), I first determined the contribution of MAPs to the RT97 and SMI36 staining along axons under conditions that were used for immunofluorescence microscopy in the present studies. To visualize and capture moving neurofilaments, I transfected cultured SCG neurons with GFP-NFM fusion protein. To determine whether overexpressing GFP-NFM perturbed the distribution of RT97 and SMI36 phosphoepitope distributions on neurofilaments in these transfected neurons, I compared these antibody stainings along single neurofilaments from transfected and untransfected neurons. To quantify the distribution of each phosphoepitope along the moving neurofilament and the overall neurofilament population, the moving filaments were captured and the axonal filaments were induced to splay apart from each other as described earlier (see details in Chapter 5). Then they were fixed and processed for double labeling fluorescence microscopy using polyclonal antibody NFLAS, specific for NFL, and RT97 or SMI36 antibodies. I found that SMI36 stained both captured and splayed filaments very weakly and there was a lot of ambiguity in interpreting the data. Thus I chose to focus on the RT97 epitope in the quantitative comparison of fluorescence intensity between captured and splayed neurofilaments.

158 6.3 Naturally occurring gaps in the neurofilament array lack RT97 and the SMI36 staining

In addition to neurofilament proteins, RT97 phosphoepitope has also been demonstrated to be present on tau (Brion et al., 1993) and microtubule- associated protein 1B (MAP1B) (Johnstone et al., 1997). SMI antibodies including SMI31, SMI33, SMI34, SMI35 and SMI310 also recognize both neurofilament and tau proteins (Lichtenberg-Kraag et al., 1992). To determine whether the RT97 and SMI36 antibodies recognize epitopes on microtubule associated proteins (MAPs) under the conditions that I used for immunofluorescence microscopy in the present study, I performed double labeling immunofluorescence microscopy using antibodies NFLAS, specific for NFL, and either RT97 or SMI36 for neurons that were fixed with 4% (w/v) formaldehyde directly without prior-permeabilization. Even though formaldehyde is not the optimal fixative for microtubules, one wouldn’t expect any loss of microtubule associated proteins from these neurons because the cells were not permeabilized. Thus the RT97 and SMI36 staining pattern observed in these neurons should represent the staining pattern of neurofilaments and microtubule associated proteins.

To differentiate staining pattern of microtubule associate proteins from those of neurofilaments, I took advantage of naturally occurring gaps in the neurofilament array of these cultured SCG neurons. It has been demonstrated previously the microtubule arrays are continuous across these neurofilament gaps (Wang et al., 2000). If RT97 and SMI36 staining of MAPs are significant in these axons, then I would expect to observe staining in these gaps. However, I observed no staining (Figure 6.1), which suggests MAPs conttribute little to the overall RT97 and SMI36 fluorescence in these axons. Perhaps this reflects the relative abundance of neurofilament phosphoepitopes compared with those of MAPs in these axons.

159

Figure 6.1 Comparison of RT97 and SMI36 staining of gaps in the axonal neurofilament array axons. Axons from cultured SCG neurons were fixed using 4% (w/v) formaldehyde without prior-permeabilization and processed for double immunofluorescence microscopy using antibodies NFLAS (A) and RT97 (B) or NFLAS (C) and SMI36 (D). Note that neither RT97 nor SMI36 stained the naturally occurring gaps in the axons, which indicates that the contributions of MAPs to both RT97 and SMI36 stainings along these axons were insignificant. Scale bar= 6µm.

6.4 RT97 and SMI36 staining along splayed filaments from transfected and untransfected neurons

To visualize and capture moving neurofilaments, I had to transfect cultured SCG neurons with GFP-NFM fusion protein. Since both RT97 and SMI36 epitopes can be generated by phosphorylation of NFM, I analyzed RT97 and SMI36 staining along splayed filaments from neurons transfected with GFP-NFM and compared this to untransfected neurons to examine whether overexpressing GFP-NFM fusion protein could change the distribution of these two phosphoepitopes along individual filaments. In other words, does the expression of GFP-NFM perturb the phosphoepitope distribution on neurofilaments in these neurons? To compare the RT97 distribution, some sister dishes from the same batch of cultures were transfected with GFP-NFM fusion protein by nuclear injection two days after plating and other were not transfected. Three days after microinjection, all these

160 dishes were treated with high concentration of detergent to induce neurofilament splaying and then fixed using formaldehyde as described in section 2.5.3 of Chapter 2. All dishes were processed identically for double-labeling immunofluorescence microscopy using NFLAS antibody, which is specific for NFL, and RT97. Images of the NFL and RT97 immunofluorescence and the GFP intrinsic fluorescence were acquired at random locations.

Single neurofilaments were identified based on their NFL immunofluorescence and the proportion of their length that stained with RT97 was measured as described in section 2.8.3 of Chapter 2. RT97 staining was quite punctate along individual neurofilaments in all dishes. A very small portion of neurofilaments were unlabeled along their entire visible length, some contained one or more contiguous labeled and unlabeled segments with discrete transitions between them (Figure 6.2D), or were labeled along the entire length with occasional interruption by unlabeled regions (Figure 6.2B). The histograms in Figure 6.2E summarized the data quantitatively. On average, splayed neurofilaments from transfected neurons stained with RT97 antibody along 84.1% of their length (maximum=100%, minimum=0; n=267) whereas splayed neurofilaments from untransfected neurons stained with RT97 along 76.5% of their length (maximum=100%, minimum=0; n=267).

The comparison of SMI36 epitope distribution along splayed filaments from transfected and untransfected neurons was carried out in the same way as described above for RT97 except that the cultures were processed for immunofluorescence microscopy using NFLAS and SMI36 antibodies. A significant proportion of the neurofilaments in all dishes did not stain with SMI36 or stained very weakly, barely distinguishable from the background (Figure 6.2I). Only a small portion of the neurofilaments were labeled continuously (Figure 6.2G). On average, splayed neurofilaments from transfected neurons had SMI36 immuofluorescence along 62.4% of their length (maximum=100%, minimum=0;

161 n=177) whereas those filaments from untransfected neurons had SMI36 immunofluorescence along 47.8% of their length (maximum=99.3%, minimum=0, n=188), which is summarized in the histograms of Figure 6.2J.

These data indicated that overexpressing GFP-NFM in these neurons did cause an 8% and 15% increase respectively in the average proportions of individual filaments that were labeled with RT97 and SMI36. This is presumably due to an increase in NFM content along those filaments. However there is still a broad distribution pattern for both epitopes. There are still a significant number of filaments in the transfected neurons that have either of these two phosphoepitopes along small portions of their length and a significant number with these epitopes along a high proportion of their length. Thus even though there were a slight increase (8% for RT97 and 15% for SMI36) in the percentage of the filament length stained with each antibody, I could still determine whether there is a correlation between these phosphoepitope distributions and neurofilament moving and pausing behaviors.

6.5 RT97 and SMI36 staining of the captured filaments

To characterize the distribution of RT97 epitope along moving neurofilaments, I processed the captured filaments for immunofluorescence microscopy using NFLAS and RT97 antibodies. The staining with RT97 was very weak along the captured filaments. This was not unique to the captured filaments because I was able to find some splayed filaments from other axons on these same coverslips and they, too, stained poorly. It turned out that RT97 epitope, like 7C5 epitope described earlier, was sensitive to formaldehyde fixation and fixation using 4% (w/v) formaldehyde in the chamber appeared to over-fix and destroy the epitope. Thus I fixed the captured filaments by perfusing 2% (w/v) formaldehyde into the chamber system. Fifteen minutes later I disassembled the chamber and placed the coverslips in PBS without any post-fixation, exactly as I did for the

162

Figure 6.2 Comparison of RT97 and SMI36 phosphoepitope distribution along splayed filaments from transfected and untransfected cells. Axonal neurofilaments were induced to splay apart from each other by permeabilization of neurons with 0.5% Triton X-100 in the presence of 0.19M NaCl three days after they were transfected by nuclear injection with the GFP-NFM expression vector (equal to five days after plating) or five days after plating without nuclear injection. Then the neurofilaments from both transfected and untransfected cells were processed for double labeling immunofluorescence microscopy using NFLAS and RT97 antibodies (A-D) or NFLAS and SMI36 antibodies (F-I). RT97 and SMI36 phosphoepitopes are distributed more continuously along some filaments (B and G) than others (D and I). Histograms of the proportion of each filament from transfected and untransfected cells that contained RT97 (E) or SMI36 phosphoepitope (J). Scale bar=4µm.

163 coverslips processed for immunofluorescence microscopy using 7C5 antibody. When the coverslips were fixed this way in the perfusion chamber, the RT97 immunostaining pattern along the splayed filaments was quite similar to those processed in open dishes. RT97 staining along captured filaments was quite strong and continuous. Figure 6.3 shows two examples of captured filaments stained with NFLAS (Figure 6.3 A and B) and RT97 (Figure 6.3 A’ and B’). In total, I analyzed RT97 epitope distribution along eleven captured filaments. Nine of the captured filaments were moving anterogradely and two of them were moving retrogradely. The length of filaments ranged from 4.1µm to >28.3µm. I was unable to estimate the actual length of two of the filaments because both of their ends were not visible in the gaps. The average length for the other nine captured filaments was 7.5µm. The proportion of the eleven captured filaments that labeled with RT97 varied between 94.9% and 100.0% (mean=98.5%, n=11). These data indicate that RT97 epitope is present along almost the entire length of these captured moving neurofilaments.

I analyzed SMI36 epitope distribution along nine captured filaments. Seven of the captured filaments were moving anterogradely and two were moving retrogradely. The length of the filaments ranged from 5.5µm to 16.5µm (mean=9.6µm, n=9). On average, these captured filaments had 95.7% of their length stained with SMI36 (maximum=100.0%, minimum=71.7%; n=11, Figure 6.3F). But the staining along most captured filaments was very weak with occasional bright puncta (Figure 6.3 D’ and E’). Because similar fluorescent dots were also present in the background, it was hard to tell whether these bright dots reflected real pattern of SMI36 labeling along the captured filaments or they were just non-specific staining. I included all the fluorescent dots that were aligned perfectly with the filament course in the analysis. As described earlier, this weak staining of SMI36 was also observed along splayed filaments and it made the quantitative study problematic. Thus I did not pursue the characterization of SMI36 epitope any further.

164

Figure 6.3 Characterization of RT97 and SMI36 phosphoepitope distribution along captured filaments. Captured filaments were double stained with NFLAS (A, B) and RT97 (A’, B’), or NFLAS (D, E) and SMI36 (D’, E’). Histograms of RT97 (C) and SMI36 (F) phosphoepitope distribution for 11 and 9 captured neurofilament respectively. Comparing to the histograms for the distributions of both phosphoepitopes along splayed filament in Figure 6.2, there was no broad distribution for either epitope along captured filaments at all. Scale bar= 4µm.

165 6.6 RMO55 and FNP7 staining of the captured filaments

In addition to RT97 and SMI36, I also explored the suitability of using the distribution of RMO55 and FNP7 epitopes along individual filaments to differentiate moving and pausing neurofilaments. RMO55 is a phosphoepitope located in the tail domain of NFM while FNP7 is a non-phosphorylated epitope present on both NFM and NFH. Previous studies have demonstrated a bimodal distribution pattern for RMO55 and FNP7 staining along individual filaments, which indicated that neurofilaments in these axons are more likely to be phosphorylated at the RMO55 and FNP7 epitopes along either all or none of their length than along part of their length (Brown, 1998). Thus, if phosphorylation at the RMO55 or FNP7 epitopes was involved in the regulation of neurofilament transport, one would expect that captured neurofilaments would more likely be from those with a small portion of their length labeled with RMO55 or a major portion of the length labeled with FNP7.

I processed six captured filaments for double labeling immunofluorescence microscopy using NFLAS and RMO55 antibodies. Four filaments were moving anterogradely and two were moving retrogradely. The filaments length ranged from 5.9µm to 21.2µm (mean=10.6µm, n=6). All of them stained very strongly and uniformly with RMO55 (Figure 6.4 A’ and B’). I also noticed strong punctate staining in some neurofilament gaps, which suggested that RMO55 was also recognizing something in the axon other than neurofilaments. RMO55 antibody has been shown to be very specific to NFM using western blotting in two published studies (Lee et al., 1987; Brown, 1998) and as well as in the present study. This further suggests that the antibody specificity against denatured proteins as examined by western blotting doesn’t reflect the antibody reactivity to proteins in their native state. The captured filaments contained RMO55 staining along >97.7% of their length, but the interpretation of these data is hard because the specificity of RMO55 is questionable.

166 In total I analyzed FNP7 epitope distribution along three captured filaments. Two filaments were moving anterogradely and one was moving retrogradely. The filament length varied between 6.9µm and 12.9µm (mean=10.4µm, n=3). On average, FNP7 labeled an average 99.7% of their length (maximum=100.0%, minimum=99.5%; n=3). However the staining of FNP7 was very weak along individual filaments including these three captured filaments (Figure 6.4 C’ and D’) and some splayed filaments encountered occasionally on the same dishes. Since FNP7 recognizes non-phosphorylated epitopes on NFM and NFH, this data suggest that these filaments might be phosphorylated at these epitopes. Because of the non-specific staining of RMO55 and weak staining of FNP7, there was a lot of ambiguity in interpreting the data and thus I did not carry out any further studies with these two antibodies.

Figure 6.4 Characterization of RMO55 and FNP7 phosphoepitope distribution along captured filaments. Captured filaments were double stained with NFLAS (A, B) and RMO55 (A’, B’), or NFLAS (C, D) and FNP7 (C’, D’). Scale bar= 4µm.

167 6.7 Quantification of RT97 staining intensity

Different from the broad distribution for RT97 epitope on the splayed filaments, on average captured filaments contained the epitope along >98% of their length. However my measurements of the distribution of fluorescence along filaments (expressed as a percentage of filament length) do not take into account possible difference in overall intensity of that fluorescence. For example, it could be that two filaments can have the same proportion of their length containing RT97 immunofluorescence but each can have different fluorescence intensity. To further investigate the phosphorylation state of the captured filaments, I characterized the immunofluorescence intensity quantitatively for RT97 epitope. For each experiment, I typically captured 1-3 filaments (one captured filament per coverslip). Thus it was necessary to compare data from several experiments. However the data from different experiments can not be compared together. As described in chapter 3, when the single filaments from detergent-splayed cytoskeletal preparations were stained under identical conditions but in separate experiments using polyclonal antibody NFLAS, specific for NFL, which is present in every neurofilament, the fluorescence intensity of individual filaments from the same experiments were comparable to each other but there were a lot of variations among those from different experiments.

My approach was to compare the RT97 immunofluorescence intensity of captured neurofilaments with that of the splayed filaments in the same experiment. Captured filaments and splayed filaments from neurons of sister dishes that expressed GFP-tagged NFM fusion proteins were processed identically for double labeling immunofluorescence microscopy using the NFLAS and RT97 antibodies. For each experiment, I quantified RT97 immunofluorescence intensity along at least 300 splayed filaments, which provided enough data to describe the distribution pattern of phosphorylation state among individual filaments at the RT97 epitope. Although in this case statistical comparison between the captured and splayed filaments was impossible due to 168 the small number of captured filaments that I could obtain from each experiment, I was still able to conclude whether these captured filaments were hypophosphorylated or not relative to splayed filaments.

Because this approach involved a large amount of quantitative analysis of fluorescence intensity for each experiment, I divided RT97 immunofluorescence intensity of each filament by its NFLAS immunofluorescence intensity for the analysis and comparison. I have shown in Chapter 3 that NFL is present throughout all neurofilaments and that NFL immuofluorescence is proportional to the filament length when stained using the polyclonal antibody NFLAS, specific for NFL. By using the RT97/NFLAS immunofluorescence intensity ratio, I could analyze fluorescence intensity using the raw image without any flat field correction and without creating segmented masks for each filament (see details in section 2.8.5 of Chapter 2) and without measuring the neurofilament length, which saved a lot of time.

6.8 Quantitative analysis of the influence of the fixation conditions on RT97 staining

As mentioned earlier, to achieve similar RT97 staining along individual neurofilaments, coverslips processed in the perfusion chamber, which is used to capture moving filaments, and those processed in open dishes were fixed using different formaldehyde concentrations. The coverslips in the perfusion chamber were fixed by flowing 2ml 2% (w/v) formaldehyde into the chamber. Fifteen minutes later, the chamber was disassembled and the coverslips were placed in PBS without any post-fixation. The coverslip processed in open dishes was fixed using 4% (w/v) formaldehyde for twenty minutes. To confirm that the fixation condition for RT97 epitope in the perfusion chamber was equal to that in open dishes in a quantitative way, I processed four sister coverslips from the same culture batch for splaying in the perfusion chamber and another four sister

169 coverslips for splaying in open dishes as described above. To permit quantitative comparison of the fluorescence intensity of neurofilaments fixed under these different conditions, I processed all the coverslips side by side for immunostaining using RT97 and NFLAS antibodies. I analyzed RT97 and NFLAS immunofluorescence intensity along 306 splayed neurofilaments from the coverslips splayed and fixed in the perfusion chamber and 305 filaments from those processed in the open dishes. For each filament, I divided its RT97 immunofluorescence intensity by its NFLAS fluorescence intensity. Figure 6.5 summarizes the RT97/NFLAs immunofluorescence intensity ratio of all analyzed filaments. Statistical comparison of two data sets showed that they were not significantly different from each other (p>0.05, Kolmogorov-Smirnov test). These data suggest the RT97 immunofluorescence of neurofilaments fixed in the perfusion chamber are quantitatively comparable to those fixed in the open dishes.

6.9 Quantitative comparison of RT97 staining along captured and splayed neurofilaments

For each experiment, I established sixteen culture dishes of SCG neurons and transfected all dishes with GFP-tagged NFM fusion proteins by nuclear injection two days after plating. Three days later, I processed eight coverslips for capturing and I typically captured 1-3 filaments in each experiment. The other eight coverslips were processed for splaying. To permit quantitative comparison of the fluorescence intensities of filaments from different sister culture dishes within each experiment, I processed the coverslips side-by-side for double-labeling fluorescence microscopy using antibodies NFLAS and RT97 to ensure identical treatments. In total I did three experiments for this analysis. For each experiment, I analyzed around three hundred splayed neurofilaments from detergent-splayed cytoskeletal preparations. For each filament, RT97 and NFLAS immunofluorescence intensity was quantified along the filament length.

170

Figure 6.5 Comparison of RT97/NFLAS immunofluorescence intensity ratio along splayed neurofilaments processed in the open dishes and those processed in open dishes. RT97/NFLAS immunofluorescence intensity ratio for 305 neurofilaments that were splayed and fixed in open dishes and for 306 neurofilaments that were splayed and fixed in the perfusion chamber. To obtain the ratio for each filament, the RT97 immunofluorescence intensity of the filament was divided by its NFLAS fluorescence intensity. These two distributions are not significant different from each other (P>0.05, Kolmogorov-Smirnov test).

171 To get the RT97/NFLAS immunofluorescence intensity ratio, RT97 fluorescence intensity of each filament was divided by its NFLAS fluorescence intensity. Figure 6.6A shows the summarized data for each experiment. Statistical comparison of the three data sets demonstrated that they were not significantly different from each other (p>0.05, Kruskal-Wallis test), which suggests that the immunostaining conditions for these three experiments were comparable to each other and I can combine the data. Figure 6.6 B showed the frequency distribution of the combined data set. The RT97/NFLAS immunofluorescence intensity ratio for all the analyzed splayed filaments varied between 0 and 1.0(average=0.4, n=912).

In total I analyzed nine captured filaments from three experiments exactly as described for splayed filaments. Four of the captured filaments were moving anterogradely and five were moving retrogradely. Average filament length is 8.8µm (minimum=4.9µm, maximum=13.2µm; n=9). The RT97/NFLAS immunofluorescence intensity ratio for these captured filaments varied between 0.2 and 0.70 (average=0.4, n=9). Comparison between the captured and splayed filaments (Figure 6.6B) shows that the phosphorylation state of the captured filaments at the RT97 epitope varied a lot but overall they were not hypophosphorylated at this epitope in these cultured neurons compared to the population as a whole. In other words, these data provide no indication of any correlation between the extent of phosphorylation at the RT97 epitope and whether a neurofilament is moving or pausing in these axons.

172

Figure 6.6 RT97 immunofluorescence intensity analysis and comparison. (A) Cumulative frequency distributions of RT97/NFLAS immunofluorescence intensity ratio for splayed neurofilaments from three independent experiments and they are not significant from each other (P>0.05, Kruskal-Wallis test). (B) Cumulative frequency distribution of combined data from three experiments in (A) and relative positions for the nine captured filaments, which is superimposed over the cumulative frequency distribution curve. Note that the RT97/NFLAS immunofluorescence intensity ratios for captured filaments were scattered the entire spectrum of the splayed filaments, which indicates that there is no correlation between the extent of phosphorylation at the RT97 epitope and neurofilament moving and pausing behavior in these axons.

173 6.10 Summary

In this chapter, I characterized the phosphorylation state of neurofilaments in cultured SCG neurons at several phosphoepitopes: RT97, SMI36, RMO55 and FNP7. Because of non-specific staining of antibody RMO55 and weak staining of antibodies SMI36 and FNP7, I found these three antibodies were not useful in this study. Thus I chose to focus on RT97 epitope in the quantitative comparison between the captured neurofilaments and the overall neurofilament population. RT97 epitope was distributed along almost the entire length of the captured filaments and quantitative RT97 immunofluorescence intensity analysis further confirmed that the captured neurofilaments were not hypophosphorylated at this epitope relative to the whole population. Thereby I concluded that phosphorylation at the RT97 epitope is not a key regulator of neurofilament transport in cultured SCG neurons.

174

CHAPTER 7

DISCUSSION

7.1 The capturing technique

A significant challenge to the study of axonal transport of neurofilaments is that their movement is infrequent. At any point in time, most neurofilaments appear to be pausing. For example, during the observations of naturally occurring gaps in cultured neurons typically there is about one moving filament per five minutes of observation time (Uchida and Brown, 2004). Since it takes a filament just tens of seconds to traverse the entire gap, the time window to study moving filaments is very narrow. To address these problems, I developed a rapid perfusion chamber system so that I could capture moving filaments in gaps where there are no other neurofilaments. By rapidly permeabilizing axons, I was able to stop filaments while they were moving through gaps and before they exited the gaps. This chamber system certainly provides a simple model to study these moving filaments directly and a completely new approach to probe the mechanism of slow axonal transport.

The underlying mechanism of the capture is not clear. I tried to reactivate the movement by capturing moving filaments using low concentrations of saponin in the presence or absence of taxol, phalloidin and AMP-PNP, which could potentially have stabilized the transport machinery during the permeabilization.

175 After the filaments stopped in the gap, I perfused reactivation solution containing ATP into the chamber but none of these reactivation attempts were successful. Even in the presence of high concentrations of ATP, the moving filaments stopped as usual after the perfusion of permeabilization solution had started. Reactivation has been very successful in the study of intracellular organelle transport using similar experimental approaches (Clark and Rosenbaum, 1982; Stearns and Ochs, 1982; Koonce and Schliwa, 1986). What makes the difference is not clear. One possible explanation is that these other studies were carried out with cells from non-mammalian species. For example, Clark and Rosenbaum used melanophores obtained from Fundulus heteroclitus for their study and Koonce and Schliwa used the giant freshwater ameba Reticulomyxa. To the best of my knowledge, no reactivation experiments of transport of any kind have been reported using permeabilized mammalian nerve cells.

Even though I did not succeed in reactivating the movements of neurofilaments in these permeabilized axons, it may still be possible with future effort. If the movements could be reactivated somehow in the future, this permeabilization- reactivation model system would prove to be very powerful in the study of neurofilament transport. One could make great progress in understanding the molecular mechanism of neurofilament transport and its regulation by identifying the essential factors for the transport. For example, one could identify the motors that power neurofilament movement in these cultured neurons by applying function blocking antibodies against potential motor proteins into the reactivation system. If the movements could still be reactivated in the presence of some control antibodies but not in the function blocking antibody, it would be strong evidence indicating that the corresponding motor is involved in neurofilament transport. Pharmacological approaches could also be very fruitful.

176 7.2 Neurofilament polymer transport in axons

7.2.1 Validation of the polymer transport hypothesis

The form in which cytoskeletal proteins are transported along axons has been debated for decades between those who believed in the movements of assembled polymers and those who believed in movements of unassembled subunits (Baas and Brown, 1997; Hirokawa et al., 1997). This long-lasting controversy is largely due to the failure of numerous attempts to detect the movements of cytoskeletal proteins directly in living cells (Brown, 2000, see details in section 1.5.2 and 1.5.3 of Chapter 1). Both neurofilament and microtubule transport have now been demonstrated in cultured nerve cells by live-cell imaging studies (Wang et al., 2000; Roy et al., 2000; Wang and Brown, 2001; Wang and Brown, 2002; Ackerley et al., 2003; Uchida and Brown, 2004). The moving structures are filamentous and appear to be diffraction-limited in width. Because of the resolution limit of light microscopy, Hirokawa and his colleagues have argued these moving structures are not necessarily assembled polymers (Hirokawa and Terada, 2000; Terada, 2003). In this study, I captured moving filaments in naturally occurring gaps along the axonal neurofilament array. Because there are no other neurofilaments in these gaps, I could identify the captured filaments unambiguously and examine their structure by both immunofluorescence microscopy and correlative light and electron microscopy.

I found that captured filaments are resistant to extraction with non-ionic detergent and they stained uniformly using NFL antibody. Statistical comparison of the fluorescence intensities of these captured filaments indicated that they have NFL contents not significantly different from single neurofilaments in detergent- splayed axonal cytoskeletons, which supports the hypothesis that neurofilament proteins move as single neurofilament polymers. To prove this definitively, I performed correlative light and electron microscopy and demonstrated that moving neurofilament proteins as observed by fluorescence microscopy were 177 indeed 10-nm-diameter single neurofilament polymers. Since previous studies have demonstrated that filaments represent ≥97% of the total moving neurofilament proteins in these cultured SCG neurons (Wang and Brown, 2001; Uchida and Brown, 2004), I conclude that neurofilament polymers are the predominant transport form in these cells.

Though filamentous structures are the predominant transport form for neurofilament proteins in cultured SCG neurons, some studies in other cell types have reported movement in a predominantly punctate form (see details in section 1.5.2 and 1.5.4 of Chapter 1). These data have been cited as evidence for the subunit transport hypothesis, though they are not conclusive because of technical concerns associated with each study. Nevertheless, the movement of neurofilament proteins in the form of subunits under certain conditions is not necessarily inconsistent with their movement in a filamentous form under other conditions. In the cultured SCG neurons, some of the total fluorescent neurofilament proteins were observed to move in the form of punctate blobs or dots (Wang and Brown, 2001; Uchida and Brown, 2004). However what these moving structures represented is not clear. They might be unassembled neurofilament proteins or they might represent short filaments. Because neurofilament proteins were observed to move predominantly in the form of puncta shortly after neurite initiation in PC12 neuronal cell lines and filamentous moving structures were not observed until some time later (Yabe et al., 2001; Helfand et al., 2003a), Shea and colleagues have proposed that neurofilament proteins transported in unpolymerized form may represent motile precursors of neurofilament assembly (Yabe et al., 1999; Yabe et a., 2001). Based on these studies and hypothesis, it appears that there might be transport machineries for both polymers and subunits and the predominant form observed to move could depend on what is the predominant form of the proteins in the axon. For example, neurofilament proteins might move predominantly as precursors of neurofilament assembly in axons immediately after neurite initiation as observed

178 in studies with neuronal cell lines, whereas they might move in a predominantly filamentous form at later times, as observed in SCG neurons. However there are a lot of questions that need to be addressed in order to accommodate these different results. Do the polymer transport and subunit transport machinery serve a different purpose? Do different neurofilament transport forms share the same regulatory machinery? How do they relate to neurofilament assembly regulation? Are they dependent on each other or independent of each other? How does the cell regulate the balance between these two transport forms?

7.2.2 Neurofilament polymers as carrier structures

Transported along with cytoskeletal protein in the slow component of axonal transport are hundreds of cytosolic proteins, which represent the entire spectrum of proteins that comprise the axoplasm. How these proteins all move together down the axon for days or weeks is unknown. Two decades ago when the polymer transport model was articulated for the first time, Lasek and his colleagues proposed that cytosolic proteins might be transported along the axon by riding piggyback on the moving polymers (Lasek et al., 1984; Lasek, 1986). Now movements of cytoskeletal proteins in filamentous form have been observed in cultured neurons (Wang et al., 2000; Roy et al., 2000; Wang and Brown, 2001; Wang and Brown, 2002; Ackerley et al., 2003) and in this study I demonstrated that the moving GFP tagged neurofilament proteins are single neurofilament polymers by immunofluorescence and electron microscopy. Many cytosolic proteins conveyed by slow axonal transport have also been shown to bind, directly or indirectly to cytoskeletal polymers. For example, spectrin, which moves in slow component a, has been shown to interact indirectly with neurofilaments in live cells (Macioce et al., 1999). Perhaps spectrin and other proteins moving in the slow component can associate with moving neurofilaments and perhaps this might provide a mechanism for how these proteins move along axons.

179 For any proteins that are transported by being associated with cytoskeletal polymers, it will be fascinating to address questions like: what is mechanism that regulates when and where these proteins attach with or detach from the polymers? Could the pausing behaviors of the neurofilament be these “loading” and “unloading” statuses for these associated proteins? How do these regulatory machineries interact with that of neurofilament transport? Would some of the proteins associated with the neurofilaments polymers during the transport be the part of neurofilament transport machinery?

Recently Brown proposed that these cytosolic proteins might associate with each other and get transported along axons as functional complexes by interacting directly with transport machinery (Brown, 2003b). Forming functional complexes now appears to be the universal idea for how cells analyze the large information flow and regulate the synchronized events insides the cell. Thus the identification and characterization of the interaction between the cytosolic proteins and cytoskeletal polymers, and the various protein complexes may provide fundamental insights into the mechanism that organize the cytosolic compartment of cytoplasm (Brown, 2003b).

7.2.3 Mechanism of neurofilament polymer transport

The rapid rate of neurofilament movement along axons suggests the transport is powered by fast motors similar to those for membranous organelles (Brown, 2003b). In support of this, various studies have demonstrated the association of microtubule and microfilament associated motor proteins with neurofilaments (see details in section 1.5.4 of Chapter 1). Disruptions of cytoplasmic dynein and KIF5A lead to impaired neurofilament transport (Xia et al., 2003; He et al., 2005).

It has been proposed that neurofilament and membranous organelle transport may share the same underlying mechanism (Brown, 2003b). But unlike

180 membranous organelles, neurofilaments are highly flexible elongated structures and moving neurofilaments can be more than thirty micrometers long, which raises the question of how motor proteins interact with neurofilaments and power their transport. In the present study, when I analyzed the structure of moving neurofilaments by immunofluorescence microscopy using NFL antibody, the staining along all seventeen analyzed captured filaments was comparably uniform to unpaired filaments from splayed cytoskeletons. I didn’t observe a big variation in the fluorescence intensity along the length of any captured filament, which suggested that there were no significant folds or bendings in the captured moving neurofilaments. The data from correlated light and electron microscopy studies of five captured filaments further confirmed that neurofilament polymers were transported along the axon in an extended from. How a long flexible structure like a neurofilament can move along an axon without folding is not clear. The simplest explanation would be that a motor protein is associated with the leading edge of the moving filaments and pulls the filament along the axon. If there are no motors at the leading edge of the filament, one would expect that the filament could easily fold back on itself. It will also be interesting to know how many motors are required to transport a single neurofilament and how these motors coordinate with each other so that the filament won’t fold on itself during the transport. If there are motors associated specifically at or near the tip of the filament, it would be interesting to know what mechanism enables this.

7.2.4 Cut and run model for cytoskeletal polymer transport

The longest neurofilament that has been observed to move along axons is around 39µm (Roy et al., 2000). The average length of moving neurofilaments observed by live-cell imaging in cultured SCG neurons varied from 4.1µm to 9.8µm depending the particular study (Roy et al., 2000; Wang et al., 2000; Wang and Brown, 2001; Uchida and Brown, 2005). Although I haven’t measured the length of neurofilaments from detergent-splayed preparations, it is my impression 181 that they are much longer than the moving neurofilaments. In bullfrog olfactory axons, the average length for neurofilaments was calculated to be about 118µm by tracking the filaments through serial sections of axons in cross section (Burton and Wentz, 1992). Individual neurofilaments as long as 183µm have also been observed in splayed cytoskeletal preparations using cultured rat DRG neurons (Brown, 1997).

Similar observations have also been made for microtubule transport. In the axons of cultured SCG neurons, the average length of microtubules that were observed to move was around 2.7µm with the maximum length around 5µm (Wang and Brown, 2002), while the length of microtubules obtained from splayed preparations using same cultures ranged from 30 to 128µm (Brown et al., 1993). The average length of microtubules in cultured rat sensory neurons was calculated to be 108µm (Bray and Bunge, 1981). Recently Baas and his colleagues proposed a model called “cut and run”, in which long microtubules are stationary, but relative short microtubules are mobile, to explain microtubule organization inside the cell (Baas et al., 2005). In this model, longer microtubules are less likely to move due to the interaction with other components within the cytoplasm, while the transport can be set forth by severing events, which renders the microtubules short enough to move. Two enzymes that sever microtubules have been identified in neurons: katanin and spastin (Baas et al., 2005). Live cell imaging studies on cortical neurons have also demonstrated that short microtubules are generated by severing of looped bundles of long microtubules in the growth cones, followed by transport of the short microtubules into filopodia (Dent et al., 1999). In this view, microtubule severing and subsequent transport would be an efficient way to provide microtubules for newly forming axonal branches, which might be hundreds of microns away from the cell body.

Could this “cut and run” model be the mechanism for the cell to supply neurofilament proteins efficiently to different regions of axons, especially the

182 newly formed axons? It could explain why moving neurofilaments appear to be much shorter in length than the axonal neurofilaments as a whole. Unfortunately, little is known about the regulatory machinery for neurofilament stability in axons, not to mention whether there are enzymes that are associated with neurofilament severing and redistribution. In the “cut and run’ model for microtubules, the capacity of a microtubule to move is inversely proportional to its length, with only the shortest microtubules displaying rapid concerted motion, while studies on neurofilament transport have revealed no difference for the motile behavior of filaments with lengths varying from 0.7µm to 39µm. It has also been proposed in the model that the threshold length for microtubule to move might possibly vary depending on the degree to which the microtubules are crosslinked with other structures (Baas et al., 2005). There may also be a threshold length for neurofilaments but it may be more variable than for microtubules. One mechanism for this variability is that the degree to which neurofilaments are crosslinked with other structures might vary a lot along axons. For instance, the interaction of neurofilaments with neurofilaments or other components of axons has been proposed be mediated by side arms of NFM and NFH, which is proposed to be regulated by phosphorylation of the tail domains of neurofilament proteins (see details in section 1.1.3 of Chapter 1). The extent of neurofilament phosphorylation varies both spatially and temporally and this suggests that the neurofilament interaction might also vary a lot along the axon, which according to the “cut and run” model predicts that neurofilaments could have different threshold length that can be transported.

The “cut and run” model raises the possibility that there might be some enzymes that sever long neurofilaments and produce relatively short filaments that interact less with other neurofilaments or axonal components and have a higher affinity for the transport machinery. According to this model, axonal neurofilaments could be divided into two distinct populations: relatively long filaments, which tend to pause for prolonged times and relatively short filaments, which tend to pause for

183 short periods of time and move more frequently along the axon. Validation of this model for neurofilament transport requires more and detailed studies on neurofilament length and assembly and identifying the proteins that might be involved in their regulation.

7.3 Polypeptide composition of moving neurofilaments

7.3.1 Both moving and pausing neurofilaments in cultured SCG neurons are complex heteropolymers

I have demonstrated unequivocally that neurofilament polymers are transported along axons. To understand the mechanism and regulation of this movement, it is important to know the polypeptide composition of these moving polymers. This would be an important first step towards determining the polypeptide or polypeptides that might mediate the interaction between moving neurofilaments and the transport motors, and identifying modifications on the polypeptides that might be involved in the regulation of neurofilament transport. In contrast to microfilaments and microtubules, neurofilaments have no fixed subunit compositions or stoichiometry (see section 1.3 of Chapter 1). In fact the polypeptide composition of neurofilaments in cultured SCG neurons has not been studied previously. To address this, I characterized the polypeptide composition of neurofilaments in these neurons. To do this, I induced axonal neurofilaments to splay apart from each other using a high concentration of non-ionic detergent and processed them for immunofluorescence microscopy. To determine whether moving and pausing neurofilaments differ in their polypeptide composition, I also characterized the polypeptide composition of moving neurofilaments. To do this, I took advantage of the capturing technique that I developed earlier to stop moving filaments within gaps in the neurofilament array and analyzed them by immunofluorescence microscopy.

184 I found that both moving and pausing neurofilaments from cultured SCG neurons were composed of at least four polypeptides, which included NFL, NFM, peripherin and α-internexin and all four polypeptides were distributed along >95% of all neurofilament length. I was not able to determine NFH incorporation along these filaments because I could not find an antibody that had good staining quality and is specific for NFH. Surprisingly, all the NFH antibodies that I tried are either commercially available or widely used but I found that all of them were either not specific for NFH or stained poorly in my immunofluorescence protocol. Nevertheless, it is likely that the NFH content of these neurofilaments is low because others in the Brown lab have observed very low levels of NFH expression as judged by western blotting (Gulsen Colakoglu, Brown lab, personal communication).

α-internexin is the first neuronal intermediate filament protein that is expressed in postmitotic neurons in the developing nervous system. In the adult, its expression is restricted to mature neurons in the CNS. Thus α-internexin in these cultured SCG neurons might be involved in neurite outgrowth and elongation as reported in differentiated neural cell lines (Shea and Beermann, 1999). The co-expression of α-internexin and peripherin in these cultured neurons is consistent with other observations that during development, the expression of α-internexin and peripherin overlaps with neurofilament triplet proteins and this continues in a subset of adult neurons (Nixon and Shea, 1992). More recently, studies with cultured dorsal root ganglion neurons demonstrated that all five neurofilament polypeptides were expressed at various times with considerable overlap with each other in these cells, which resembled their developmental expression pattern in some aspects (Athlan et al., 1997).

Surprisingly, even though both α-internexin and peripherin can self-assemble into 10-nm filaments in vivo (Ching and Liem, 1993; Cui et al., 1998), I didn’t find any neurofilaments from these cultured SCG neurons that contained only α-internexin

185 or only peripherin in this study. All four polypeptides that I analyzed coassembled with each other and formed hetero-polymers in these neurons. In contrast, studies with PC12 cells have revealed three different subpopulations of filaments composed of only peripherin, or only neurofilament triplet proteins, or both peripherin and neurofilament triplet proteins (Parysek et al., 1993). The difference in these observations might be related to the relative expression level of peripherin and other neurofilament proteins in these neurons. It is possible that in cultured SCG neurons the ratio of peripherin to other neurofilament proteins is relatively low and thus the probability of co-assembly of peripherin with other neuronal intermediate filaments is much higher than the probability of self- assembly. On the other hand, in PC12 cells the expression level of other neurofilament proteins is relatively low thus peripherin self-assembly is more likely.

7.3.2 Neurofilament polypeptides and the transport machinery

The complex composition of moving neurofilaments raises the question which polypeptide(s) is/are the binding partner of motor proteins or binding partner of adaptors for the motor proteins during neurofilament transport. Studies with NFM, NFH, peripherin and α-internexin in knock out mice have demonstrated that neurofilaments are distributed along axons (Elder et al., 1998; Zhu et al., 1998; Rao et al., 1998; Elder et al., 1999a; Levavasseur et al., 1999; Kriz et al., 2000; Lariviere et al., 2002), which suggests that neurofilaments are transported into the axons in these mice and thus that none of these four polypeptides is critical for neurofilament transport. Recently, Nixon and his colleagues revealed that neither NFL nor NFH is required for NFM transport in axons (Yuan et al., 2003). These data collectively suggest that none of the neurofilament polypeptides are critical for neurofilament transport. One possible explanation for this is that the

186 neurofilament transport machinery can interact with multiple neurofilament polypeptides and thus neurofilament transport won’t be disrupted in the absence of one particular polypeptide.

In line with the above reasoning, studies from different labs suggest that several neurofilament polypeptides can mediate the interaction between neurofilament and the transport machinery. For example, the fact that α-internexin is the only neuronal intermediate filament protein in many small interneurons in the cerebrum and granule cells in the cerebellum suggests that α-internexin forms homopolymers in these cells (Chiu et al., 1989; Kaplan et al., 1990; Chien et al., 1996). In order to transport into the axon, α-internexin must be able to interact directly or indirectly with transport machinery, but how α-internexin associates with motor proteins is not clear.

In the cerebellum a sub-population of climbing fibers and mossy fibers don’t have other neurofilament proteins except peripherin (Errante et al., 1998), which indicates that peripherin can be transported along the axon by itself and so it must also be capable of interaction, directly or indirectly, with the transport machinery. Recently, live cell imaging studies in PC12 cells have revealed rapid transport of peripherin throughout the cell bodies, neurites and growth cones (Helfand et al., 2003a). Short peripherin filaments and punctates were shown to co-localize with microtubules and motor proteins including kinesin and dynein by immunofluorescence microscopy (Helfand et al., 2003a). The association of peripherin punctates with the motor proteins and microtubules was further demonstrated by platinum replica immunogold electron microscopy (Helfand et al., 2003a).

NFM has been shown to be co-immunoprecipitated with kinesin, but NFL was not (Yabe et al., 2000). To further study the roles of NFL and NFM separately in the neurofilament transport, Nixon and his colleagues used NFM and NFH, or NFL

187 and NFH double knock out mice (Yuan et al., 2003). They found that in the NFM and NFH double knock out mouse the amount of NFL that was transported into axons of optic nerve was very low and they concluded that NFL alone was incapable of efficient transport. On the other hand, NFM could be transported into axons in the absence of other neurofilament triplet proteins. But in the α- internexin and NFL double knock out mice, the NFM transport was abolished. Thus the author proposed that neurofilament protein transport requires hetero- oligomer formation for efficient transport, and NFM is most likely to be the critical subunit that interacts with the putative motor proteins (Yuan et al., 2003). However, the overall NFL protein levels in those NFM and NFH double knock-out mice was very low (NFM appears to regulate the level of NFL expression, Elder et al., 1998; Elder et al., 1999a) and thus it is hard to say whether non-detectable NFL transport was due to inefficient transport or low expression level of the protein. More recently, an association between dynein motor complex and neurofilaments has been demonstrated by fluid tapping mode atomic force microscopy (Wagner et al., 2004). Yeast two-hybrid and affinity chromatography assays have further revealed that NFM binds directly to dynein intermediate chain and antibodies directed against either of them can disrupt the interaction, which supports the hypothesis that NFM is at least one of the neurofilament polypeptides that can interact with the transport motors.

All of these data suggest that NFM, α-internexin and peripherin can act as binding partners for the motor proteins that are involved in neurofilament transport in these cultured SCG neurons. Since all of them share a common α- helical coilded coil rod domain, this may be the site of motor binding. Alternatively, there may be specific adaptor proteins for each neurofilament polypeptide. It will be very interesting to know if motor proteins have preferential binding partners among all the polypeptides that can interact with the transport machinery.

188 7.4 Phosphorylation of neurofilament proteins and neurofilament transport

7.4.1 Regulation of neurofilament pausing behaviors

Live-cell imaging studies on the axonal transport of GFP-tagged neurofilament proteins suggest that the overall slow rate of neurofilament transport is a result of rapid movements interrupted by prolonged pauses, but little is known about the regulation of the moving and pausing behavior. Recently Brown and his colleagues proposed that neurofilaments can exist in two states that differ in their pausing behavior, which were referred to as on track and off track (Brown et al., 2005). Neurofilaments can alternate between short bouts of rapid movement interrupted by short pauses (on track state) and prolonged pauses (off track state) during which the neurofilament temporarily disengages from the tracks. The idea that axonal neurofilaments have two distinct populations: moving and stationary, was first proposed by Nixon and Logvinenko (Nixon and Logvinenko, 1986). However, their data were not incisive since in that study they didn’t differentiate neurofilament proteins from the proteins that might co-migrate with neurofilaments on one-dimensional SDS-PAGE. But computation modeling studies suggest that basic idea of stationary neurofilaments does appear to be correct (Brown et al., 2005; Cracium et al., 2005). These two different pausing behaviors can also be distinguished in the live cell imaging studies. For example, most of the neurofilaments at the edge of gaps in the neurofilament array didn’t move during the time of observations, which suggested that those could be filaments that were in the state of prolonged pauses.

In the simulation studies of Brown et al (2005), it was predicted that neurofilaments spend ~8% of their time on track and 97% of the time pausing during their journey along the axon. In other words, neurofilaments spend 92% of their time in prolonged pauses and 5% of their time in short pauses. Based on these numbers, one would expect that variations in the short pauses wouldn’t 189 change the overall rate of neurofilament transport very much while the prolonged pauses should be the key factor that regulates neurofilament movement frequency and the overall rate. What could regulate the prolonged pauses? There are several factors that might be involved, but two of them sound most appealing. First, it could be the phosphorylation of tail domains of neurofilament proteins. As described in section 1.5.5 of Chapter 1, when the KSP sites on the tail domains of NFH were mutated to mimic permanent phosphorylation, the pause durations increased for these mutants during rapid movements. Unfortunately, in those studies, there was no report concerning the changes in movement frequency for those mutants compared with mutants that mimic the permanently de-phosphorylated form, which would be an indication of changes in prolonged pauses mediated by phosphorylation alone as oppose to some other effect of the site-directed mutogenesis. Also, those studies examined only the short (presumed on track) pauses, which may not give an accurate picture of the overall pausing behavior.

Second, the availability of tracks for neurofilament movement may contribute to the prolonged pauses. Two factors could regulate the availability of tracks. One factor could be the proximity of the filament to the track along which they move (Brown et al., 2005) and the other one could be the “clearness” of the track. Several studies have suggested microtubules as the principle tracks for neurofilament transport (Shah et al., 2000; Francis et al., 2005). But in most axons, neurofilaments usually outnumber microtubules. For example, in the sciatic nerve of 14-wk-old mice, the ratio of neurofilaments to microtubules ranges from 7:1 to 16:1 (Reles and Friede, 1991). In cultured neurons, neurofilaments also outnumber microtubules in some axons (see Figure 4.1A in Chapter 4). These data suggest that a lot of neurofilaments are not proximal to microtubules at any point in time and thus it is possible that they cannot be transported and this could cause them to pause for prolonged times. In addition, microtubule associated proteins might also be involved in the regulation

190 neurofilament transport. Microtubule associated proteins regulate microtubule stability and dynamics, and help maintain cell shape or neurite outgrowth. They are also potential competitors of microtubule associated motor proteins for binding sites. For example, over-expressing tau can lead to reduced attachment of kinesin and disturbed axonal transport (Seits et al., 2002). These microtubule associated proteins bind to microtubules in a de-phosphorylation dependent manner. They can be phosphorylated by various kinases including microtubule affinity regulating kinase (MARK)/Par1kinase and detach from microtubules, thus facilitating axonal transport by clearing the microtubule walls (Mandelkow et al., 2004). So hypophosphorylated MAPs decorated along microtubules could also lead to prolonged neurofilament pausing behavior.

7.4.2 Moving neurofilaments are not hypophosphorylated at the RT97 epitope compared with pausing neurofilaments

To further investigate the role of phosphorylation in neurofilament transport, I initiated a characterization of the phosphorylation state of moving neurofilaments at several phosphoepitopes that have been implicated in the regulation of neurofilament transport. To do that, I analyzed the distribution of RT97, SMI36, RMO55 and FNP7 epitopes along captured neurofilaments and compared them with the total neurofilament population in these cells, most of which are pausing at any point in time. All these phosphoepitopes were distributed along most of the length of the captured moving neurofilaments. However, I found that RMO55 is not specific for immunostaining and that SMI36 and FNP7 stain very weakly along most captured filaments, which caused a lot ambiguity about the data. For this reason, I chose to focus on characterization of RT97 epitope for my quantitative measurements. As described in section 1.5.6 of Chapter 1, phosphorylation at RT97 epitopes has been correlated with regional accumulation of neurofilament proteins and pausing behavior of individual neurofilaments as revealed by studies from both optic nerve and cultured

191 neurons (Sanchez et al., 2000; Ackerley et al., 2003). This has lead to the hypothesis that moving neurofilaments may be hypophosphorylated at the RT97 epitope compared with pausing neurofilaments. I found that in cultured SCG neurons there was a lot of variation in the distribution of the RT97 epitope along the neurofilaments: some filaments contained the epitope along their whole length, some parts of the length and some not at all. In contrast, the epitope was distributed along almost entire length of all the captured filaments. Further quantitative analysis and comparison of the RT97 immunofluorescence intensity along the captured and splayed filaments confirmed that the captured neurofilaments did not differ from the neurofilament population as a whole.

These data indicate that phosphorylation at the RT97 epitope is not a regulator of neurofilament transport in these cultured SCG neurons. This conclusion appears to contradict other studies that have suggested that phosphorylation at this epitope is a key regulator of neurofilament transport. However there are several caveats associated with my study that preclude the extrapolation that phosphorylation at RT97 epitope is not important for regulation of neurofilament transport.

The first caveat is that the model system I used in my study is five-day-old SCG neuron cultures, in which the expression level of NFH was still very low. It has been previously demonstrated that the generation of the RT97 epitope on NFM was selectively increased in NFH knock out mice (Sanchez et al., 2000). Thus it is quite possible the RT97 epitope that was analyzed in my study was mainly located on NFM, while previous studies on this epitope were exclusively focused on NFH (see details in section 1.5.6 of Chapter 1). This might suggest that the RT97 epitopes on NFM and NFH might have different roles. To address this problem, one could express site directed mutants of GFP-tagged NFM and GFP- tagged NFH separately in sister cultures from NFM and NFH double knock out

192 mice. By comparing the neurofilament transport under these two different conditions, one can conclude whether or not NFM and NFH have different regulatory effects on neurofilament transport.

A second caveat is that only a subset of RT97 sites might be involved in the regulation of neurofilament transport. In fact, the RT97 epitope has not been mapped on neurofilaments. It has been suggested that RT97 epitope is generated through phosphorylation by the kinase Cdk5, which preferentially phosphorylates serine in KSPXK motifs, where X is not an acidic amino acid). In the carboxyl tail domain of mouse NFM and NFH, there are two and four KSPXK sites (X is either valine or methionine) respectively and each one of them could, in principle, be phosphorylated by Cdk5. However it is not clear whether each one of these phosphorylated KSPXK sites can generate the RT97 epitope or whether it requires a combination of different phosphorylated KSPXK sites to form the epitope. It is also quite possible that the RT97 epitope could be generated by different KSP sites that might have different roles in neurofilament transport and that the RT97 antibody might not be able to differentiate the epitopes generated by these different KSPXK sites. To further investigate the role of each RT97 epitope KSP site on NFM and NFH in neurofilament transport, one could extend the studies from the Miller lab (Ackerley et al., 2003; see details in section 1.5.6 of Chapter 1), in which all RT97 epitope KSP sites on NFH were mutated together to mimic either permanently dephosphorylated or permanently phosphorylated RT97 epitope. One could make two different point mutations at each site separately either to mimic dephosphorylated KSP sites or phosphorylated KSP sites and then transfect different mutants into cultured neurons from NFM and NFH double knock out mice. If one particular KSP site were critical for regulation of neurofilament transport, one would expect less frequent neurofilament transport in the cells transfected with the mutant mimicking phosphorylation when compared with the mutant mimicking dephosphorylation.

193 A third caveat is that different axonal locations of the moving and pausing neurofilaments that I analyzed might have obscured differences in their phosphorylation state at the RT97 epitope. As described in Chapter 6, to analyze the phosphorylation state of moving neurofilaments I took advantage of naturally occurring gaps in the neurofilament array and captured the moving neurofilaments in these gaps by permeabilizing the axons. The gaps were usually found in thin axons that have relative few filaments and are located quite far away from the cell body. On the other hand, splayed neurofilaments were randomly selected from detergent-splayed cytoskeletal preparations along the entire axon. It has been demonstrated by several labs that the phosphorylation state of neurofilaments at certain epitopes varies spatially in a very defined manner in both culture neurons and in vivo (see details in section 1.4.1 of Chapter 1). Overall there is a proximal-to-distal gradient of increasing phosphorylation for neurofilament proteins along the axon for some phosphoepitopes. However, whether there is a similar proximal-to-distal gradient in the phosphorylation state of the RT97 epitope on neurofilament proteins along the axon is largely unknown. To explore this possibility, one could compare the phosphorylation state of the RT97 epitope on neurofilaments from different regions of axons, such as the proximal, distal and middle compartments. If there is a difference in RT97 epitope distribution on neurofilaments from different regions, the next question to ask would be: if the RT97 epitope is a key regulator of neurofilament transport, would the neurofilament movement frequency be different in different regions of the axon? For example, if the neurofilaments from proximal regions of axons were relatively hypophosphorylated at RT97 epitope, one might expect that neurofilaments would move more frequently.

In conclusion, my data suggest that the expression of one particular important phosphoepitope, RT97, might not be sufficient to regulate neurofilament transport. However, all the caveats associated with this study mean that my data are not incisive. Nevertheless, it is interesting to note that studies from two

194 prominent labs have demonstrated that loss of the tail domain of either NFM or NFH and all of its associated phosphorylation sites didn’t affect the rate of neurofilament transport in optic nerve as revealed by radioisotopic labeling approach (Rao et al., 2002; Rao et al., 2003). Recently, Nixon and his colleagues proposed in a poster abstract that the phosphorylated tail domains of NFM and NFH together regulated neurofilament content but not the rate of slow axonal transport of neurofilaments (Rao et al., 2004).

7.5 Future directions

In the last several years, our understanding of the slow axonal transport of neurofilament proteins has been advanced greatly. However defining how neurofilament transport interacts with the transport of other cytosolic or cytoskeletal proteins that also move in the slow components of axonal transport and how neurofilament transport is regulated is largely incomplete. Identifying the proteins that associate with neurofilament proteins in axons will be important in understanding the neurofilament transport machinery and the associated regulatory mechanism. Most current studies are concentrated on in vitro cell culture models, which have limited definition of slow axonal transport. In vivo imaging of axonal transport could provide more detailed information for neurofilament transport regulatory machinery. For example, how is neurofilament transport regulated by myelination? How is the transport related to neuronal activity? Is the transport bidirectional in vivo?

195

LIST OF REFERENCES

Aebi U, Fowler WE, Rew P, Sun TT. (1983) The fibrillar substructure of keratin filaments unraveled. J. Cell. Biol. 97:1131-1143.

Aebi U, Cohn J, Buhle L, Gerace L. (1986) The nuclear lamina is a meshwork of intermediate-type filaments. Nature 323:560-564.

Ackerley S, Grierson AJ, Brownlees J, Thornhill P, Anderton BH, Leigh PN, Shaw CE, Miller CC. (2000) Glutamate slows axonal transport of neurofilaments in transfected neurons. J. Cell Biol. 150:165-176.

Ackerley S, Thornhill P, Grierson AJ, Brownlees J, Anderton BH, Leigh PN, Shaw CE, Miller CC. (2003) Neurofilament heavy chain side arm phosphorylation regulates axonal transport of neurofilaments. J. Cell Biol. 161:489-495.

Alberts B. (1998) The cell as a collection of protein machines;preparing the next generation of molecular biologists. Cell 92: 291–294.

Al-Chalabi A, Andersen PM, Nilsson P, Chioza B, Andersson JL, Russ C, Shaw CE, Powell JF, Leigh PN. (1999) Deletions of the heavy neurofilament subunit tail in amyotrophic lateral sclerosis. Hum. Mol. Genet. 8:157–164.

Al-Chalabi A, Miller CC. (2003) Neurofilaments and neurological disease. Bioessays. 25:346-55.

Altman J, Bayer SA. (1982) Development of the cranial nerve ganglia and related nuclei in rat. (Advances in Anatomy, Embryology and Cell Biology, vol. 74), Springer-Verlag, Berlin. 196 Altman J, Bayer SA. (1984) The Development of rat spinal cord. (Advances in Anatomy, Embryology and Cell Biology, vol. 85), Springer-Verlag, Berlin.

Altman J, Bayer SA (1986) The Development of rat hypothalamus. (Advances in Anatomy, Embryology and Cell Biology, vol. 100), Springer-Verlag, Berlin.

Athlan ES, Sacher MG, Mushynski WE. (1997) Association between intermediate filament protein expression in cultured dorsal root ganglion neurons. J. Neurosci. Res. 47: 300-310.

Archer DR, Watson DF, Griffin JW. (1994) Phosphorylation-dependent immunoreactivity of neurofilaments and the rate of slow axonal transport in the central and peripheral axons of the dorsal root ganglion. J. Neurochem. 62:1119- 1125.

Baas PW, and Brown A. (1997) Slow axonal transport: the polymer transport model. Trends Cell Biol. 7: 380–384.

Baas PW, Karabay A, Qiang L. (2005) Microtubules cut and run. Trends Cell Biol. 15:518-524.

Bajaj NP, Miller CC. (1997) Phosphorylation of neurofilament heavy-chain sidearm fragments by cyclin-dependent kinase-5 and glycogen synthase kinase- 3alpha in transfected cells. J. Neurochem. 69:737-743.

Bamburg JR, Bray D, Chapman K. (1986) Assembly of microtubules at the tip of growing axons. Nature. 321:788-790.

Beaulieu JM, Nguyen MD, Julien JP. (1999a) Late onset death of motor neurons in mice overexpressing wild-type peripherin. J Cell Biol 147:531–544.

Beaulieu JM, Robertson J, Julien JP. (1999b) Interactions between peripherin and neurofilaments in cultured cells: disruption of peripherin assembly by the NF- M and NF-H subunits. Biochem Cell Biol 77:41– 45.

197 Beaulieu JM, Jacomy H, Julien JP. (2000) Formation of intermediate filament protein aggregates with disparate effects in two transgenic mouse models lacking the neurofilament light subunit. J Neurosci 20:5321–5328.

Beaulieu JM, Kriz J, Julien JP. (2002) Induction of peripherin expression in subsets of brain neurons after lesion injury or cerebral ischemia. Brain Res 946:153–161.

Beaulieu JM, Julien JP. (2003) Peripherin-mediated death of motor neurons rescued by overexpression of neurofilament NF-H proteins. J Neurochem 85:248 –256.

Bennett GS, Tapscott SJ, DiLullo C, Holtzer H. (1984) Differential binding of antibodies against the neurofilament triplet proteins in different avian neurons. Brain Res. 304:291-302.

Bennett GS, Quintana R. (1997) Identification of Ser-Pro and Thr-Pro phosphorylation sites in chicken neurofilament-M tail domain. J. Neurochem. 68: 534– 543.

Benson DL, Mandell JW, Shaw G, Banker G. (1996) Compartmentation of alpha- internexin and neurofilament triplet proteins in cultured hippocampal neurons. J. Neurocytol. 25:181-196.

Bergeron C, Beric-Maskarel K, Muntasser S, Weyer L, Somerville MJ, Percy ME. 1994. Neurofilament light and polyadenylated mRNA levels are decreased in amyotrophic lateral sclerosis motor neurons. J Neuropathol Exp Neurol 53:221– 230.

Bignami A, Raju T, Dahl D. (1982) Localization of vimentin, the nonspecific intermediate filament protein, in embryonal glia and in early differentiating neuron, In vivo and in vitro immunofluorescence study of the rat embryo with vimentin and neurofilaments antisera, Dev. Biol. 91: 286-295.

Black MM, Lasek RJ. (1979) Axonal transport of actin: slow component b is the principal source of actin for the axon. Brain Res. 10:401-13.

198 Black MM, Cochran JM, Kurdyla JT. (1984) Solubility properties of neuronal tubulin: evidence for labile and stable microtubules. Brain Res. 295:255-263.

Boyne LJ, Fischer I, Shea TB. (1996) Role of Vimentin in early stages of neuritogenesis in cultured hippocampal neurons. Int. J. Dev. Neurosci. 14: 739- 748.

Bray D, Bunge MB. (1981) Serial analysis of microtubules in cultured rat sensory axons.cJ Neurocytol. 10:589-605.

Bray D. (1997) The riddle of slow transport – an introduction. Trends Cell Biol. 7: 379.

Brion JP, Couck AM, Robertson J, Loviny TL, Anderton BH. (1993) Neurofilament monoclonal antibodies RT97 and 8D8 recognize different modified epitopes in paired helical filament-tau in Alzheimer's disease. J. Neurochem 60:1372-1382.

Brody BA, Ley CA, Parysek LM. (1989) Selective distribution fo the 57 kDa neural intermediate filament protein in the rat CNS. J. Neurosci. 9:2391-2401.

Brown A, Slaughter T, Black MM. (1992) Newly assembled microtubules are concentrated in the proximal and distal regions of growing axons. J. Cell Biol. 119:867-882.

Brown A, Li Y, Slaughter T, Black MM. (1993) Composite microtubules of the axon: quantitative analysis of tyrosinated and acetylated tubulin along individual axonal microtubules. J. Cell Sci. 104:339-352.

Brown A, Lasek RJ. (1993) Neurofilaments move apart freely when released from the circumferential constraint of the axonal plasma membrane. Cell Motil. Cytoskeleton. 26:313-324.

Brown A. (1997) Visualization of single neurofilaments by immunofluorescence microscopy of splayed axonal cytoskeletons. Cell Motil. Cytoskeleton. 38:133- 145.

199 Brown A. (1998) Contiguous phosphorylated and non-phosphorylated domains along axonal neurofilaments. J. Cell Sci. 111:455-467.

Brown A. (2000) Slow axonal transport: stop and go traffic in the axon. Nat Rev Mol. Cell Biol. 1:153-156.

Brown A. (2003a) Live-cell imaging of slow axonal transport in cultured neurons. Methods Cell Biol. 71:305-323.

Brown A. (2003b) Axonal transport of membranous and nonmembranous cargoes: a unified perspective. J Cell Biol. 160:817-821.

Brown A, Wang L, Jung P. (2005) Stochastic simulation of neurofilament transport in axons: the "stop-and-go" hypothesis. Mol Biol Cell. 16:4243-4255.

Brown HG, Hoh JH. (1997) Entropic exclusion by neurofilament sidearms: a mechanism for maintaining interfilament spacing. Biochemistry. 36:15035-15040.

Brownlees J, Yates A, Bajaj NP, Davis D, Anderton BH, Leigh PN, Shaw CE, Miller CC. (2000) Phosphorylation of neurofilament heavy chain side-arms by stress activated protein kinase- 1b/Jun N-terminal kinase-3. J. Cell Sci. 113: 401–407.

Burton PR. (1985) Ultrastructure of the olfactory neuron of the bullfrog: the dendrite and its microtubules. J Comp Neurol. 242:147-160.

Burton PR, Wentz MA. (1992) Neurofilaments are prominent in bullfrog olfactory axons but are rarely seen in those of the tiger salamander, Ambystoma tigrinum. J. Comp. Neurol. 317:396-406.

Craciun G, Brown A, Friedman A. (2005) A dynamical system model of neurofilament transport in axons. J. Theor. Biol. 237:316-322.

Carden MJ, Eagles PA. (1983) Neurofillament from ox spinal nerves. Isolation, disassembly, reassembly and cross-linking properties. Biochem J. 215:227-237.

200 Carden MJ, Schlaepfer WW, Lee VM. (1985) The structure, biochemical properties, and immunogenicity of neurofilament peripheral regions are determined by phosphorylation state. J. Biol.Chem. 260:9805-9817.

Carden MJ, Eagles PA. (1986) Chemical cross-linking analyses of ox neurofilaments. Biochem J. 234:587-591.

Carden MJ, Trojanowski JQ, Schlaepfer WW, Lee VM –Y. (1987) Two-stage expression of neurofilament polypeptides during rate neurogenesis with early establishment of adult phosphorylation patters. J. Neurosci. 7: 3489-3504.

Carter J, Gragerov A, Konvicka K, Elder G, Weinstein H, Lazzarini RA. (1998) Neurofilament (NF) assembly; divergent characteristics of human and rodent NF- L subunits. J. Biol. Chem. 273:5101-5108.

Chang SH, Rodionov VI, Borisy GG, and Popov SV (1998) Transport and turnover of microtubules in frog neurons depend on the pattern of axonal growth. J. Neurosci. 18: 821–829.

Chen J, Nakata T, Zhang Z, Hirokawa N. (2000) The C-terminal tail domain of neurofilament protein-H (NF-H) forms the crossbridges and regulates neurofilament bundle formation. J Cell Sci. 113:3861-9.

Chien CL, Liem RK. (1995) The neural intermediate filament, alpha-internexin is transiently expressed in amacrine cells in the developing mouse retina. Exp. Eye Res. 61:749-756.

Chien CL, Mason CA, Liem RK. (1996) Alpha-internexin is the only neuronal intermediate filament expressed in developing cerebelar granule neurons. J. Neurobiol. 29:304-318.

Ching GY, Liem RK. (1993) Assembly of type IV neuronal intermediate filaments in nonneuronal cells in the absence of preexisting cytoplasmic intermediate filaments. J Cell Biol. 122:1323-1335.

201 Ching GY, Chien CL, Flores R, Liem RK. (1999) Overexpression of alpha- internexin causes abnormal neurofilamentous accumulations and motor coordination deficits in transgenic mice. J. Neurosci. 19:2974-2986.

Chiu FC, et al. (1989) Charaterization of a novel 66 kd subunit of mammalian neurofilaments. Neuron 21:1435-1445.

Chou CF, Smith AJ, Omary MB. (1992) Characterization and dynamics of O- linked glycosylation of human cytokeratin 8 and 18. J Biol Chem. 267:3901-3906.

Clark TG, Rosenbaum JL. (1982) Pigment particle translocation in detergent- permeabilized melanophores of Fundulus heteroclitus. Proc. Natl. Acad. Sci. USA. 79:4655-4659.

Clark EA, Lee VM-Y. (1991) Dynamics of mammalian high-molecular-weight neurofilament subunit phosphorylation in cultured rat sympathetic neurons. J. Neurosci. Res. 30: 116-123.

Cleveland DW, Monteiro MJ, Wong PC, Gill SR, Gearhart JD, Hoffman PN. (1991) Involvement of neurofilaments in the radial growth of axons. J. Cell Sci. Suppl.15:85-95.

Cleverley KE, Betts JC, Blackstock WP, Gallo JM, Anderton BH. (1998) Identification of novel in vitro PKA phosphorylation sites on the low and middle molecular mass neurofilament subunits by mass spectrometry. Biochemistry. 37:3917-3930.

Cochard P, Paulin D. (1984) Initial expression of neurofilaments and vimentin in the central and peripheral nervous system of mouse embyo in vivo. J. Neurosci. 4: 2080-2094.

Cohlberg JA, Hajarian H, Tran T, Alipourjeddi P, Noveen A. (1995) Neurofilament protein heterotetramers as assembly intermediates. J. Biol. Chem. 270:9334- 9339.

Collard JF, Cote F, Julien JP. (1995) Defective axonal transport in a transgenic mouse model of amyotrophic lateral sclerosis. Nature 375:61– 64. 202 Cote F, Collard JF, Julien JP. (1993) Progressive neuronopathy in transgenic mice expressing the human neurofilament heavy gene: a mouse model of amyotrophic lateral sclerosis. Cell 73:35– 46.

Couillard-Despres S, Zhu Q, Wong PC, Price DL, Cleveland DW, Julien JP. (1998) Protective effect of neurofilament heavy gene overexpression in motor neuron disease induced by mutant superoxide dismutase. Proc. Natl. Acad. Sci. USA 95:9626 –9630.

Cui C, Stambrook PJ, Parysek LM. (1995) Peripherin assembles into homopolymers in SW13 cells. J. Cell. Sci. 108:3279-3284.

Dabora SL, Sheetz MP. (1988) Cultured cell extracts support organelle movement on microtubules in vitro. Cell Motil. Cytoskeleton. 10:482-495.

Dahlstrand J, Zimmerman LB, McKay RD, Lendahl U. (1992) Characterization of the human nestin gene reveals a close revolutionary relationship to neurofilaments. J. Cell Sci. 103:589-597.

Dashiell SM, Tanner SL, Pant HC, Quarles RH. (2002) Myelin-associated glycoprotein modulates expression and phosphorylation of neuronal cytoskeletal elements and their associated kinases. J. Neurochem. 81:1263-1272.

De Jonghe P, Mersiyanova IV, Nelis E, Del Favero J, Martin JJ, Van Broeckhoven C, Evgrafov OV, Timmerman V. (2001) Further evidence that neurofilament light chain mutations can cause Charcot-Marie Tooth disease type 2E. Ann. Neurol. 49:245–249.

de Waegh SM, Lee VM, Brady ST. (1992) Local modulation of neurofilament phosphorylation, axonal caliber, and slow axonal transport by myelinating Schwann cells. Cell. 68:451-463.

Dent EW, Callaway JL, Szebenyi G, Baas PW, Kalil K. (1999) Reorganization and movement of microtubules in axonal growth cones and developing interstitial branches. J. Neurosci. 19:8894-8908.

203 Ding J, Liu JJ, Kowal AS, Nardine T, Bhattacharya P, Lee A, Yang Y. (2002) icrotubule-associated protein 1B: a neuronal binding partner for gigaxonin. J Cell Biol 158: 427–433.

Doetsch F, Garcia-Verdugo JM, Alvarez-Buylla A. (1997) Cellular composition and three-dimensional organization of the subventricular germinal zone in the adult mammalian brain. J. Neurosci. 17:5046-5061.

Dong DL, Xu ZS, Chevrier MR, Cotter RJ, Cleveland DW, Hart GW. (1993) Glycosylation of mammalian neurofilaments. Localization of multiple O-linked N- acetylglucosamine moieties on neurofilament polypeptides L and M. J. Biol. Chem. 268:16679-16687.

Dong DL, Xu ZS, Hart GW, Cleveland DW. (1996) Cytoplasmic O-GlcNAc modification of the head domain and the KSP repeat motif of the neurofilament protein neurofilament-H. J. Biol. Chem.271:20845-20852.

Dosemeci A, Floyd CC, Pant HC. (1990) Characterization of neurofilament- associated protein kinase activities from bovine spinal cord. Cell Mol. Neurobiol. 10: 369–382.

Drager UC. (1983) Coexistence of neurofilaments and vimentin in a neuron of adult mouse retina. Nature 303:169-172.

Dubey M, Hoda S, Chan WK, Pimenta, A, Ortiz DD, Shea TB. (2004). Reexpression of vimentin in differentiated neuroblastoma cells enhances elongation of axonal neurites. J. Neurosci. Res. 78: 245-249.

Eaker EY, Sallustio JE. (1994) The distribution of novel intermediate filament proteins defines subpopulations of myenteric neurons in rat intestine. Gastroenteroloty 107:666-674.

Elder GA, Friedrich VL. Jr, Bosco P, Kang C, Gourov A, Tu PH, et al. (1998) Absence of the mid-sized neurofilament subunit decreases axonal calibers, levels of light neurofilament (NF-L), and neurofilament content. J. Cell Biol. 141: 727–739.

204 Elder GA, Friedrich VL. Jr, Margita A, Lazzarini RA. (1999a) Age-related atrophy of motor axons in mice deficient in the mid-sized neurofilament subunit. J. Cell Biol. 146: 181–192.

Elder GA, Friedrich VL. Jr, Pereira D, Tu PH, Zhang B, Lee VM, et al. (1999b) Mice with disrupted midsized and heavy neurofilament genes lack axonal neurofilaments but have unaltered numbers of axonal microtubules. J. Neurosci. Res. 57: 23–32.

Elhanany E, Jaffe H, Link WT, Sheeley DM, Gainer H, Pant HC. (1994) Identification of endogenously phosphorylated KSP sites in the high-molecular- weight rat neurofilament protein. J. Neurochem. 63:2324-2335.

Engel A, Eichner R, Aebi U. (1985) Polymorphism of reconstituted human epidermal keratin filaments: determination of their mass-per-length ad width by scanning transmission electron microscopy (STEM). J. Ultrastruct. Res. 90:323- 335.

Errante L, Tang D, Gardon M, Sekerkova G, Mugnaini E, Shaw G. (1998) The intermediate filament protein peripherin is a specific marker for cerebellar climbing fibers. J. Neurocytol. 27:69-84.

Escurat M, Gumpel M. Lachapelle F. Gros F. Portier MM. (1988) Comparative expression of 2 intermediate filament proteins, peripherin and the 68 kDa neurofilament protein, during embryonal development of the rat. C. R. Acad. Sei. Ser. C 306: 447-456.

Eyer J, Peterson A. (1994) Neurofilament-deficient axons and perikaryal aggregates in viable transgenic mice expressing a neurofilament- betagalactosidase fusion protein. Neuron 12: 389–405.

Francis F, Roy S, Brady ST, Black MM. (2005) Transport of neurofilaments in growing axons requires microtubules but not actin filaments. J. Neurosci. Res. 79:442-450.

Frappier T, Regnouf F, Pradel LA. (1987) Binding of brain spectrin to the 70-kDa neurofilament subunit protein. Eur. J. Biochem. 169:651-657.

205 Frappier T, Stetzkowski-Marden F, Pradel LA. (1991) Interaction domains of neurofilament light chain and brain spectrin. Biochem. J. 275:521-527.

Ferri GL, et al. (1990) Neuronal intermediate filaments in rat dorsal root ganglia differential distribution of peripherin and neurofilament protein immunoreactivity and effct of capsaicin. Brain Res. 515:331-335.

Figlewicz DA, Krizus A, Martinoli MG, Meininger V, Dib M, Rouleau GA, Julien JP. (1994) Variants of the heavy neurofilament subunit are associated with the development of amyotrophic lateral sclerosis. Hum. Mol. Genet. 3:1757–1761.

Fliegner KH, Ching GY, Liem RK. (1990) The predicted amino acid sequence of alpha-internexin is that of a novel neuronal intermediate filament protein. EMBO J. 9: 749-755.

Fliegner KH, Kaplan MP, Wood TL, Pintar JE, Liem RK. (1994) Expression of the gene for the neuronal intermediate filament protein alpha-internexin coincides with the onset of neuronal differentiation in the developing rat nervous system. J.Comp. Neurol. 342:161-173.

Floyd CC, Grant P, Gallant PE, Pant HC. (1991) Principal neurofilament- associated protein kinase in squid axoplasm is related to casein kinase I. J. Biol. Chem. 266: 4987–4994.

Foster GA, Dahl D, Lee VM. (1987) Temporal and topographic relationships between the phosphorylated and nonphosphorylated epitopes of the 200 kDa neurofilament protein during development in vitro. J. Neurosci. 7:2651-2663.

Friede RL, Samorajski T. (1970) Axon caliber related to neurofilaments and microtubules in sciatic nerve fibers of rats and mice. Anat. Rec. 167:379-787.

Fuchs E, Hanukoglu I, Marchuk D, Grace MP, Kim KH. (1985) The nature and significance of differential keratin gene expression. Ann. N.Y. Acad. Sci. 455:436- 450.

206 Fuchs E, Weber K. (1994) Intermediate filaments structure, dynamics, function, and disease. Annu. Rev. Biochem. 63:345-382.

Galbraith JA, Reese TS, Schlief ML, Gallant PE. (1999) Slow transport of unpolymerized tubulin and polymerized neurofilament in the squid giant axon. Proc. Natl. Acad. Sci. USA 96:11589–11594.

Garcia ML, Lobsiger CS, Shah SB, Deerinck TJ, Crum J, Young D, Ward CM, Crawford TO, Gotow T, Uchiyama Y, Ellisman MH, Calcutt NA, Cleveland DW. (2003 ) NF-M is an essential target for the myelin-directed "outside-in" signaling cascade that mediates radial axonal growth. J Cell Biol. 163:1011-1120.

Garcia ML, Singleton AB, Hernandez D, Ward CM, Evey C, Sapp PA, Hardy J, Brown RH Jr, Cleveland DW. (2005) Mutations in neurofilament genes are not a significant primary cause of non-SOD1-mediated amyotrophic lateral sclerosis. Neurobiol Dis. Aug 2.

Geisler N, Kaufmann E, Weber N. (1985) Antiparallel orientation of the two double-stranded coiled-coils in the tetrameric protofilament unit of intermediate filaments. J. Mol. Biol. 183:173-177.

Geisler N, Schunemann J, Weber N. (1992) Chemical cross-linking indicate a staggered and antiparallel protofilament of desmin intermediate filaments and characterizes one higher-level complex between protofilaments. Eur. J. Biochem. 206:841-852.

Georgiou DM, Zidar J, Korosec M, Middleton LT, Kyriakides T, Christodoulou K. (2002) A novel NF-L mutation Pro22Ser is associated with CMT2 in a large Slovenian family. Neurogenetics 4:93–96.

Giasson BI, Cromlish JA, Athlan ES, Mushynski WE. (1996) Activation of cyclic AMP-dependent protein kinase in okadaic acid-treated neurons potentiates neurofilament fragmentation and stimulates phosphorylation of Ser2 in the low- molecular-mass neurofilament subunit. J. Neurochem. 66:1207-1213.

Giasson BI, Mushynski WE. (1996) Aberrant stress-induced phosphorylation of perikaryal neurofilaments. J. Biol. Chem. 271: 30404– 30409.

207 Giasson BI, Mushynski WE. (1997) Study of proline-directed protein kinases involved in phosphorylation of the heavy neurofilament subunit. J. Neurosci. 17: 9466–9472.

Giasson BI, Mushynski WE. (1998) Intermediate filament disassembly in cultured dorsal root ganglion neurons is associated with amino-terminal head domain phosphorylation of specific subunits. J. Neurochem. 70: 1869–1875.

Gibb BJ, Robertson J, Miller CC. (1996) Assembly properties of neurofilament light chain Ser55 mutants in transfected mammalian cells. J. Neurochem. 66:1306-1311.

Gibb BJ, Brion JP, Brownlees J, Anderton BH, Miller CC. (1998) Neuropathological abnormalities in transgenic mice harbouring a phosphorylation mutant neurofilament transgene. J. Neurochem. 70:492-500.

Gill SR, Wong PC, Monteiro MJ, Cleveland DW. (1990) Assembly properties of dominant and recessive mutations in the small mouse neurofilament (NFL) subunit. J. Cell Biol. 111:2005-2019.

Glicksman MA, Soppet D, Willard MB. (1987) Posttranslational modification of neurofilament polypeptides in rabbit retina. J. Neurobiol.18:167-196.

Goldstein ME, Sternberger NH, Sternberger LA. (1987) Phosphorylation protects neurofilaments against proteolysis. J. Neuroimmunol. 14:149-160.

Gonda Y, Nishizawa K, Ando S, Kitamura S, Minoura Y, Nishi Y, Inagaki M. (1990) Involvement of protein kinase C in the regulation of assembly- disassembly of neurofilaments in vitro. Biochem. Biophys. Res. Commun. 167:1316-1325.

Gorham JD. Baker H, Kegler D, tiff EB. (1990) The expression of the neuronal intermediate filament protein peripherin in the rat embryo. Dev. Brain Res. 57:235-248.

208 Gou JP, Gotow T, Janmey PA, Leterrier JF. (1998) Regulation of neurofilament interactions in vitro by natural and synthetic polypeptides sharing Lys-Ser-Pro sequences with the heavy neurofilament subunit NF-H: neurofilament crossbridging by antiparallel sidearm overlapping. Med. Biol. Eng. Comput. 36:371-387.

Granger BL, Lazarides E. (1983) Expression of the major neurofilament subunit in chicken erythrocytes. Science 221: 553-556.

Grant P, Diggins M, Pant HC. (1999) Topographic regulation of cytoskeletal protein phosphorylation by multimeric complexes in the squid giant fiber system. J. Neurobiol. 40: 89–102.

Grant P, Pant HC. (2000) Neurofilament protein synthesis and phosphorylation. J. Neurocytol. 29:843-872.

Greene LA. (1989) A new neuronal intermediate filament protein. Trends Neurosci. 12:228-230.

Guan RJ, Khatra BS, Cohlberg JA. (1991) Phosphorylation of bovine neurofilament proteins by protein kinase FA (glycogen synthase kinase 3). J. Biol. Chem. 266: 8262–8267.

Guan RJ, Hall FL, Cohlberg JA. (1992) Proline-directed protein kinase (p34cdc2/p58cyclin A) phosphorylates bovine neurofilaments. J. Neurochem. 58:1365-1371.

Guidato S, Tsai LH, Woodgett J, Miller CC. (1996b) Differential cellular phosphorylation of neurofilament heavy side-arms by glycogen synthase kinase- 3 and cyclin-dependent kinase-5. J. Neurochem 66: 1698–1706.

Haltiwanger RS, Kelly WG, Roquemore EP, Blomberg MA, Dong LY, Kreppel L, Chou TY, Hart GW. (1992) Glycosylation of nuclear and cytoplasmic proteins is ubiquitous and dynamic. Biochem. Soc. Trans. 20:264-269.

209 Harris J, Moreno S, Shaw G, Mugnaini E. (1993) Unusual neurofilament composition in cerebellar unipolar brush neurons. J. Neurocytol. 22:1039-1059.

Harry JG, Goodrum JF, Morell P. (1985) The postnatal development of glial fibrillary acidic protein and neurofilament triplet proteins in rat brain stem, Int. J. Dev. Neurosci. 3: 349-352.

Hart GW, Haltiwanger RS, Holt GD, Kelly WG. (1989) Glycosylation in the nucleus and cytoplasm. Annu. Rev. Biochem.58:841-874.

Hatzfeld M, Weber K. (1990) The coiled coil of in vitro assembled keratin filaments is a heterodimer of type I and II keratins use of site-specific mutagenesis and recombinant protein expression. J. Cell Biol. 110:1199-1210.

He Y, Francis F, Myers KA, Yu W, Black MM, Baas PW. (2005) Role of cytoplasmic dynein in the axonal transport of microtubules and neurofilaments. J. Cell Biol. 168:697-703.

Heins S, et al. (1993) The rod domain of NF-L determines neurofilament architecture, whereas the end domain specify filament assembly and network formation. J. Cell Biol. 123:1517-1533.

Heins S and Aebi U. (1994) Making heads and tails of intermediate filament assembly, dynamics and networks. Curr. Opin. Cell Biol. 6:25-33.

Helfand BT, Mikami A, Vallee RB, Goldman RD. (2002) A requirement for cytoplasmic dynein and dynactin in intermediate filament network assembly and organization. J. Cell Biol. 157:795-806.

Helfand BT, Loomis P, Yoon M, Goldman RD. (2003a) Rapid transport of neural intermediate filament protein. J Cell Sci. 116:2345-2359.

Helfand BT, Mendez MG, Pugh J, Delsert C, Goldman RD. (2003b) A role for intermediate filaments in determining and maintaining the shape of nerve cells. Mol. Biol. Cell. 14:5069-5081.

210 Hirokawa N. (1982) Cross-linker system between neurofilaments, microtubules, and membranous organelles in frog axons revealed by the quick-freeze, deep- etching method. J. Cell Biol. 94:129-142.

Hirokawa N, Terada S, Funakoshi T, Takeda S. (1997) Slow axonal transport: the subunit transport model. Trends Cell Biol. 7:384–388.

Hirano A, Nakano I, Kurland LT, Mulder DW, Holley PW, Saccomanno G. (1984) Fine structural study of neurofi-brillary changes in a family with amyotrophic lateral sclerosis. J Neuropathol Exp Neurol 43:471– 480.

Hisanaga S, Hirokawa N. (1988) Structure of the peripheral domains of neurofilaments revealed by low angle rotary shadowing. J. Mol. Biol. 202:297- 305.

Hisanaga S, Hirokawa N. (1989) The effects of dephosphorylation on the structure of the projections of neurofilament. J Neurosci. 9:959-966.

Hisanaga S, Hirokawa N. (1990) Molecular architecture of the neurofilament. II. Reassembly process of neurofilament L protein in vitro. J Mol Biol. 211:871-882.

Hisanaga S, Ikai A, Hirokawa N. (1990a) Molecular architecture of the neurofilament. I. Subunit arrangement of neurofilament L protein in the intermediate-sized filament. J Mol Biol. 211:857-869.

Hisanaga S, Gonda Y, Inagaki M, Ikai A, Hirokawa N. (1990b) Effects of phosphorylation of the neurofilament L protein on filamentous structures. Cell Regul. 1:237-248.

Hockfield S, Mckay RDG. (1985) Identification of major cell classes in the developing mammalian nervous system. J. Neurosci. 5:3310-3328.

Hoffman PN. (1988) Distinct roles of neurofilament and tubulin gene expression in axonal growth. Ciba. Found. Symp.138:192-204.

211 Hoh JH. (1998) Functional protein domains from the thermally driven motion of polypeptide chains: a proposal. Proteins. 32:223-228.

Hollander BA, Bennett GS, Shaw G. (1996) Localization of sites in the tail domain of the middle molecular mass neurofilament subunit phosphorylated by a neurofilament-associated kinase and by casein kinase I. J. Neurochem. 66: 412– 420.

Holt GD, Haltiwanger RS, Torres CR, Hart GW. (1987) Erythrocytes contain cytoplasmic glycoproteins. O-linked GlcNAc on Band 4.1. J. Biol. Chem. 262:14847-14850.

Ishihara T, Hong M, Zhang B, Nakagawa Y, Lee MK, Trojanowski JQ, Lee VM. (1999) Age-dependent emergence and progression of a tauopathy in transgenic mice overexpressing the shortest human tau isoform. Neuron. 24:751-762.

Ishihara T, Higuchi M, Zhang B, Yoshiyama Y, Hong M, Trojanowski JQ, Lee VM. (2001) Attenuated neurodegenerative disease phenotype in tau transgenic mouse lacking neurofilaments. J Neurosci. 21:6026-6035.

Jackson SP, Tjian R. (1988) O-glycosylation of eukaryotic transcription factors: implications for mechanisms of transcriptional regulation. Cell. 55:125-133.

Jacomy H, Zhu Q, Couillard-Despres S, Beaulieu JM, Julien JP. (1999) Disruption of type IV intermediate filament network in mice lacking the neurofilament medium and heavy subunits. J. Neurochem. 73: 972–984.

Johnstone M, Goold RG, Fischer I, Gordon-Weeks PR. (1997) The neurofilament antibody RT97 recognises a developmentally regulated phosphorylation epitope on microtubule-associated protein 1B. J. Anat. 191:229-244.

Jordanova A, De Jonghe P, Boerkoel CF, Takashima H, De Vriendt E, Ceuterick C, Martin JJ, Butler IJ, Mancias P, Papasozomenos S, et al. (2003) Mutations in the neurofilament light chain gene (NEFL) cause early onset severe Charcot- Marie-Tooth disease. Brain 126:590 –597.

212 Julien JP, Mushynski WE. (1982) Multiple phosphorylation sites in mammalian neurofilament polypeptides. J. Biol. Chem. 257:10467-10470.

Julien JP, Mushynski WE. (1983) The distribution of phosphorylation sites among identified proteolytic fragments of mammalian neurofilaments. J. Biol. Chem. 258:4019-4025.

Julien JP, Cote F, Collard JF. (1995) Mice overexpressing the human neurofilament heavy gene as a model of ALS. Neurobiol. Aging 16: 487–492

Jung C, Shea TB. (1999) Regulation of neurofilament axonal transport by phosphorylation in optic axons in situ. Cell Motil. Cytoskeleton 42:230-240.

Jung C, Yabe JT, Shea TB. (2000a) C-terminal phosphorylation of the high molecular weight neurofilament subunit correlates with decreased neurofilament axonal transport velocity. Brain Res. 856:12-19.

Jung C, Yabe JT, Lee S, Shea TB. (2000b) Hypophosphorylated neurofilament subunits undergo axonal transport more rapidly than more extensively phosphorylated subunits in situ. Cell Motil. Cytoskeleton. 47:120-129.

Kaether C, Skehel P, Dotti CG. (2000) Axonal membrane proteins are transported in distinct carriers: a two-color video microscopy study in cultured hippocampal neurons. Mol. Biol. Cell. 11:1213-1224.

Kaplan MP, Chin SS, Fliegner KH, Liem RK. (1990) Alpha-internexin, a novel neuronal intermediate filament protein, precedes the low molecular weight neurofilament protein (NF-L) in the developing rat brain. J. Neurosci. 10:2735- 2748.

Kearse KP, Hart GW. (1991) Lymphocyte activation induces rapid changes in nuclear and cytoplasmic glycoproteins. Proc. Natl. Acad. Sci. U. S. A. 88:1701- 1705.

Kelly BM, Gillespie CS, Sherman DL, Brophy PJ. (1992) Schwann cells of the myelin-forming phenotype express neurofilament protein NF-M. J. Cell Biol. 118:397-410. 213 Kelly WG, Dahmus ME, Hart GW. (1993) RNA polymerase II is a glycoprotein. Modification of the COOH-terminal domain by O-GlcNAc. J Biol Chem. 268:10416-10424.

King IA, Hounsell EF. (1989) Cytokeratin 13 contains O-glycosidically linked N- acetylglucosamine residues. J. Biol. Chem. 264:14022-140028.

Klymkowsky MW. (1995) Intermediate filaments: new proteins, some answers, more questions. Curr. Opin. Cell Biol. 7:46-54.

Koehnle TJ, Brown A. (1999) Slow axonal transport of neurofilament protein in cultured neurons. J. Cell Biol. 144:447-458.

Koonce MP, Euteneuer U, Schliwa M. (1986) Reticulomyxa: a new model system of intracellular transport. J. Cell Sci. Suppl.5:145-159.

Koonce MP, Schliwa M. (1986) Reactivation of organelle movements along the cytoskeletal framework of a giant freshwater ameba. J. Cell Biol. 103:605-612.

Krishnan N, Kaiserman-Abramof IR, Lasek RJ. (1979) Helical substructure of neurofilaments isolated from Myxicola and squid giant axons. J. Cell Biol. 82:323- 335.

Kriz J, Zhu Q, Julien JP, Padjen AL. (2000) Electrophysiological properties of axons in mice lacking neurofilament subunit genes: disparity between conduction velocity and axon diameter in absence of NF-H. Brain Res. 885: 32–44.

Kron SJ, Spudich JA. (1986) Fluorescent actin filaments move on myosin fixed to a glass surface. Proc Natl Acad Sci U S A. 83:6272-6276.

Kumar S, Yin X, Trapp BD, Hoh JH, Paulaitis ME. (2002a) Relating interactions between neurofilaments to the structure of axonal neurofilament distributions through polymer brush models. Biophys. J. 82:2360-2372.

214 Kumar S, Yin X, Trapp BD, Paulaitis ME, Hoh JH. (2002b) Role of long-range repulsive forces in organizing axonal neurofilament distributions: evidence from mice deficient in myelin-associated glycoprotein. J. Neurosci. Res. 68:681-690.

Kumar S, Hoh JH. (2004) Modulation of repulsive forces between neurofilaments by sidearm phosphorylation. Biochem. Biophys. Res. Commun. 324:489-496.

Lariviere RC, Nguyen MD, Ribeiro-da-Silva A, Julien JP. (2002) Reduced number of unmyelinated sensory axons in peripherin null mice. J. Neurochem. 81:525-532.

Lariviere RC, Julien J-P. (2003) Function of intermediate filaments in neuronal development and disease. J Neurobiol 58:131-48.

Lasek RJ, Garner JA, Brady ST. (1984) Axonal transport of the cytoplasmic matrix. J. Cell Biol. 99:212s-221s.

Lasek RJ, Brady ST. (1985) Attachment of transported vesicles to microtubules in axoplasm is facilitated by AMP-PNP. Nature. 316:645-647.

Lasek RJ. (1986) Polymer sliding in axons. J. Cell Sci. Suppl.5:161-79.

Lavedan C, Buchholtz S, Nussbaum RL, Albin RL, Polymeropoulos MH. (2002) A mutation in the human neurofilament M gene in Parkinson’s disease that suggests a role for the cytoskeleton in neuronal degeneration. Neurosci. Lett. 322:57-61.

Lawson SN, Harper AA, Harper EI, Garson JA, Anderton BH. (1984) A monoclonal antibody against neurofilament protein specifically labels a subpopulation of rat sensory neurons. J. Comp. Neurol. 228:263-272.

Lawson SN, Waddell PJ. (1991) Soma neurofilament immunoreactivity is related to cell size and fiber conduction velocity in rat primary sensory neurons. J. Physiol. Lond. 435:41-63.

215 LeBeau JM, Ellisman MH, Powell HC. (1988) Ultrastructural and morphometric analysis of long term peripheral nerve regeneration through silicone tubes. J. Neurocytol. 7:161-172.

Lee MK, Marszalek JR, Cleveland DW. (1994) A mutant neurofilament subunit causes massive, selective motor neuron death: implications for the pathogenesis of human motor neuron disease. Neuron 13:975–988.

Lee MK, Cleveland DW. (1996) Neuronal intermediate filaments. Annu. Rev. Neurosci. 19, 187-217.

Lee VM. (1985) Neurofilament protein abnormalities in PC12 cells: comparison with neurofilament proteins of normal cultured rat sympathetic neurons. J. Neurosci. 5:3039-3046.

Lee VM, Carden MJ, Schlaepfer WW, Trojanowski JQ. (1987) Monoclonal antibodies distinguish several differentially phosphorylated states of the two largest rat neurofilament subunits (NF-H and NF-M) and demonstrate their existence in the normal nervous system of adult rats. J. Neurosci. 7:3474-3488.

Lee VM, Otvos L Jr, Carden MJ, Hollosi M, Dietzschold B, Lazzarini RA. (1988) Identification of the major multiphosphorylation site in mammalian neurofilaments. Proc. Natl. Acad. Sci. U S A. 85:1998-2002.

Lendahl U, Zimmerman LB, McKay RDG. (1990) CNS stem cells express a new class of intermediate filament protein. Cell 60:585-595.

Leonard DG, Gorham JD, Cole P, Greene LA, Ziff EB. (1988) A nerve growth factor-regulated messenger RNA encodes a new intermediate filament protein. J. Cell Biol. 106:181-193.

Leterrier JF, Eyer J. (1987) Properties of highly viscous gels formed by neurofilaments in vitro. A possible consequence of a specific inter-filament crossbridging. Biochem. J. 245:93-101.

216 Leterrier JF, Kas J, Hartwig J, Vegners R, Janmey PA. (1996) Mechanical effects of neurofilament cross-bridges. Modulation by phosphorylation, lipids, and interactions with F-actin. J. Biol. Chem. 271:15687-15694.

Leung CL, Sun D, Zheng M, Knowles DR, Liem RK. (1999) Microtubule actin cross-linking factor (MACF): a hybrid of dystonin and dystrophin that can interact with the actin and microtubule cytoskeletons. J Cell Biol. 147:1275-86.

Levavasseur F, Zhu Q, Julien JP. (1999) No requirement of alpha-internexin for nervous system development and for radial growth of axons. Brain Res. Mol. Brain Res. 69:104-112.

Lewis SE, Nixon RA. (1988) Multiple phosphorylation variant of the high molecular mass subunit of neurofilaments in axons of retina cell neurons: characterization and evidence for their differential association with stationary and moving filaments. J. cell Biol. 107: 2689-2701.

Li BS, Veeranna, Grant P, Pant HC. (1999a) Calcium influx and membrane depolarization induce phosphorylation of neurofilament (NF-M) KSP repeats in PC12 cells. Brain. Res. Mol. Brain Res. 70: 84–91.

Li BS, Veeranna, Gu J, Grant P, Pant HC. (1999b) Activation of mitogen- activated protein kinases (Erk1 and Erk2) cascade results in phosphorylation of NF-M tail domains in transfected NIH 3T3 cells. Eur. J. Biochem. 262: 211–217.

Li BS, Zhang L, Gu J, Amin ND. Pant HC. (2000) Integrin alpha(1) beta(1)- mediated activation of cyclin-dependent kinase 5 activity is involved in neurite outgrowth and human neurofilament protein H Lys-Ser-Pro tail domain phosphorylation. J. Neurosci. 20: 6055–6062.

Lichtenberg-Kraag B, Mandelkow EM, Biernat J, Steiner B, Schroter C, Gustke N, Meyer HE, Mandelkow E. (1992) Phosphorylation-dependent epitopes of neurofilament antibodies on tau protein and relationship with Alzheimer tau. Proc. Natl. Acad. Sci. U S A. 89:5384-5388. Lim S-S, Sammak PJ, Borisy GG. (1989) Progressive and spatially differentiated stability of microtubules in developing neuronal cells. J. Cell Biol. 109: 253–263.

217 Lim S-S, Edson KJ, Letourneau, PC, Borisy GG. (1990) A test of microtubule translocation during neurite elongation. J. Cell Biol. 111:123–130.

Link WT, Grant P, Hidaka H, Pant HC. (1992) Casein kinase I and II from squid brain exhibit selective neurofilament phosphorylation. Mol. Cell Neurosci 3: 548– 558.

Liu Q, Xie F, Siedlak SL, Nunomura A, Honda K, Moreira PI, Zhua X, Smith MA, Perry G. (2004) Neurofilament proteins in neurodegenerative diseases. Cell Mol. Life Sci. 61:3057-3075.

Ludemann N, Clement A, Hans VH, Leschik J, Behl C, Brandt R. (2005) O- glycosylation of the tail domain of neurofilament protein M in human neurons and in spinal cord tissue of a rat model of amyotrophic lateral sclerosis (ALS). J. Biol. Chem. 280:31648-13658.

Luthi T, Haltiwanger RS, Greengard P, Bahler M. (1991) Synapsins contain O- linked N-acetylglucosamine. J. Neurochem. 56:1493-1498.

Lutter M, Fang M, Luo X, Nishijima M, Xie X, Wang X (2000) Cardiolipin provides specificity for targeting of tBid to mitochondria. Nat. Cell Biol. 2:754-761.

Macioce P, Gandolfi N, Leung CL, Chin SS, Malchiodi-Albedi F, Ceccarini M, Petrucci TC, Liem RK. (1999) Characterization of NF-L and betaIISigma1- spectrin interaction in live cells. Exp. Cell Res. 250:142-154.

Mandelkow EM, Thies E, Trinczek B, Biernat J, Mandelkow E. (2004) MARK/PAR1 kinase is a regulator of microtubule-dependent transport in axons. J. Cell Biol. 167:99-110.

Marszalek JR, Williamson TL, Lee MK, Xu Z, Hoffman PN, Becher MW. et al. (1996) Neurofilament subunit NF-H modulates axonal diameter by selectively slowing neurofilament transport. J. Cell Biol. 135: 711–724.

218 Mata M, Kupina N, Fink DJ. (1992) Phosphorylation-dependent neurofilament epitopes are reduced at the node of Ranvier. J. Neurocytol. 21: 199–210.

Matus A. (1988) Neurofilamant phosphorylation - where, when and why? TINS 11:291.

McKeon FD, Kirschner MW. Caput D. (1986) Homologies in both primary and secondary structure between nuclear envelope and intermediate filament proteins. Nature 319:463-468.

McLachlan DR, Lukiw WJ, Wong L, Bergeron C, Bech-Hansen NT. (1988) Selective messenger RNA reduction in Alzheimer’s disease. Brain Res. 427:255– 261.

Menet V, Prieto M, Privat A, Gimenez y Ribotta M. (2003). Axonal plasticity and functional recovery after spinal cord injury in mice deficient in both glial fibrillary acidic protein and vimentin genes. Proc. Natl. Acad. Sci. USA 100: 8999-9004.

Meier J, Couillard-Despres S, Jacomy H, Gravel C, Julien JP. (1999) Extra neurofilament NF-L subunits rescue motor neuron disease caused by overexpression of the human NF-H gene in mice. J. Neuropathol. Exp. Neurol. 58:1099 –1110.

Menzies FM, Grierson AJ, Cookson MR, Heath PR, Tomkins J, Figlewicz DA, Ince PG, Shaw PJ. (2002) Selective loss of neurofilament expression in Cu/Zn superoxide dismutase (SOD1) linked amyotrophic lateral sclerosis. J. Neurochem. 82:1118 –1128.

Mersiyanova IV, Perepelov AV, Polyakov AV, Sitnikov VF, Dadali EL, Oparin RB, Petrin AN, Evgrafov OV. (2000) A new variant of Charcot-Marie-Tooth disease type 2 is probably the result of a mutation in the neurofilament-light gene. Am. J. Hum. Genet. 67:37– 46.

Miller KW, Joshi HC. (1996). Tubulin transport in neurons. J. Cell Biol. 133: 1355–1366.

219 Mukhopadhyay R, Kumar S, Hoh JH. (2004) Molecular mechanisms for organizing the neuronal cytoskeleton. Bioessays. 26:1017-1025.

Myers MW, Lazzarini RA, Lee VM, Schlaepfer WW, Nelson DL. (1987) The human mid-size neurofilament subunit: a repeated protein sequence and the relationship of its gene to the intermediate filament gene family. EMBO J. 6:1617- 1626.

Nakagawa T, Chen J, Zhang Z, Kanai Y, Hirokawa N. (1995) Two distinct functions of the carboxyl-terminal tail domain of NF-M upon neurofilament assembly: cross-bridge formation and longitudinal elongation of filaments. J Cell Biol. 129:411-429.

Nakamura Y, Watanabe M, Ujihara M, Demura H. (1990) Effect of phosphorylation on 68 kDa neurofilament subunit protein assembly by the cyclic AMP dependent protein kinase in vitro. Biochem. Biophys. Res. Commun. 169:744-750.

Nakamura Y, et al. (1993) Acceleration of bovine neurofilament L assembly by deprivation of acidic tail domain. Eur. J. Biochem. 212:565-571.

Nakamura Y, Hashimoto R, Kashiwagi Y, Wada Y, Sakoda S, Miyamae Y, Kudo T, Takeda M. (1999) Casein kinase II is responsible for phosphorylation of NF-L at Ser-473. FEBS Lett. 455:83-86.

Nakamura Y, Hashimoto R, Kashiwagi Y, Aimoto S, Fukusho E, Matsumoto N, Kudo T, Takeda M. (2000) Major phosphorylation site (Ser55) of neurofilament L by cyclic AMP-dependent protein kinase in rat primary neuronal culture. J. Neurochem. 74:949-959.

Nguyen MD, Lariviere RC, Julien JP. (2000) Reduction of axonal caliber does not alleviate motor neuron disease caused by mutant superoxide dismutase 1.Proc. Natl. Acad. Sci. U S A. 97:12306-12311.

Nguyen MD, Lariviere RC, Julien JP. (2001) Deregulation of Cdk5 in a mouse model of ALS: toxicity alleviated by perikaryal neurofilament inclusions.Neuron. 30:135-147.

220 Nguyen MD, Julien JP, Rivest S. (2002) Innate immunity: the missing link in neuroprotection and neurodegeneration? Nat. Rev. Neurosci. 3:216-227.

Nguyen MD, Boudreau M, Kriz J, Couillard-Despres S, Kaplan DR, Julien JP. (2003) Cell cycle regulators in the neuronal death pathway of amyotrophic lateral sclerosis caused by mutant superoxide dismutase 1. J. Neurosci. 23:2131-2140.

Nixon RA, Lewis SE. (1986) Differential turnover of phosphate groups on neurofilament subunits in mammalian neurons in vivo. J. Biol. Chem. 261:16298- 16301.

Nixon RA. (1992) Slow axonal transport. Curr. Opin. Cell Biol. 4:8-14.

Nixon RA, Shea TB. (1992) Dynamics of neuronal intermediate filaments: a developmental perspective. Cell Motil. Cytoskeleton. 22:81-91.

Nixon RA, Paskevich PA, Sihag RK, Thayer CY. (1994) Phosphorylation on carboxyl terminus domains of neurofilament proteins in retinal ganglion cell neurons in vivo: influences on regional neurofilament accumulation, interneurofilament spacing, and axon caliber. J Cell Biol. 126:1031-1046.

Nona SN, Trowell SC, Cronley-Dillon JR. (1985) Postnatal development profiles of filamentous actin and of 200 kDa neurofilament polypeptide in the visual cortex of light- and dark-reared rats and their relationship to cortical period plasticity. FEBS Lett, 186: 111-115.

Ohara O, Gahara Y, Miyake T, Teraoka H, Kitamura T. (1993) Neurofilament deficiency in quail caused by nonsense mutation in neurofilament-L gene. J. Cell Biol. 121: 387–395.

Okabe S, Hirokawa N. (1990) Turnover of fluorescently labeled tubulin and actin in the axon. Nature 343:479–482.

Okabe S, Hirokawa N. (1992) Differential behavior of photoactivated microtubules in growing axons of mouse and frog neurons. J. Cell Biol. 117: 105– 120.

221 Okabe S, Hirokawa N. (1993) Do photobleached fluorescent microtubules move?: re-evaluation of fluorescence laser photobleaching both in vitro and in growing Xenopus axons. J. Cell Biol. 120: 1177–1186.

Okabe S, Miyasaka H, Hirokawa N. (1993) Dynamics of the neuronal intermediate filaments. J. Cell Biol. 121: 375–386.

O'Brien ET, Salmon ED, Erickson HP (1997) How calcium causes microtubule depolymerization. Cell Motil. Cytoskeleton. 36:125-135.

Pachter JS, Liem RKH. (1984) The differential appearance of neurofilament triplet polypeptides in the developing rat optic nerve. Dev. Bio. 103: 200-210.

Pachter JS, Liem RK. (1985) alpha-Internexin, a 66-kD intermediate filament- binding protein from mammalian central nervous tissues. J. Cell Biol. 101:1316- 1322

Pang L, Sawada T, Decker SJ, Saltiel AR. (1995) Inhibition of MAP kinase kinase blocks the differentiation of PC-12 cells induced by nerve growth factor. J. Biol. Chem. 270: 13585–13588.

Pant H, Veeranna. (1995) Neurofilament phosphorylation. Biochem. Cell Biol. 73: 575–592.

Pant HC. (1988) Dephosphorylation of neurofilament proteins enhances their susceptibility to degradation by calpain. Biochem J. 256:665-668.

Parry DAD, Steinert PM. (1995) Intermediate filament structure. R.G. Landes, Austin Texas.

Parysek LM, Goldman RD. (1988) Distribution of a novel 57 kDa intermediate filament (IF) protein in the nervous system. J. Neurosci. 8:555-563.

Parysek LM, Chisholm RL, Ley CA, Goldman RD. (1988) A type III intermediate filament gene is expressed in mature neurons. Neuon 1:395-401. 222 Parysek LM, McReynolds MA, Goldman RD, Ley CA. (1991) Some neural intermediate filaments contain both peripherin and the neurofilament proteins. J Neurosci Res. 30:80-91.

Popov S, Poo MM. (1992) Diffusional transport of macromolecules in developing nerve processes. J Neurosci. 12:77-85.

Portier MM, de N-B, Gros F. (1983) Peripherin, a new member of the intermediate filament protein family. Dev. Neurosci. 6:335-344.

Prahlad V, Yoon M, Moir RD, Vale RD, Goldman RD. (1998) Rapid movements of vimentin on microtubule tracks: kinesin-dependent assembly of intermediate filament networks. J. Cell Biol. 143:159-170.

Prahlad V, Helfand BT, Langford GM, Vale RD, Goldman RD. (2000) Fast transport of neurofilament protein along microtubules in squid axoplasm. J. Cell Sci. 113:3939-3946.

Qi Z, Tang D, Zhu X, Fujita DJ, Wang JH. (1998) Association of neurofilament proteins with neuronal Cdk5 activator. J. Biol. Chem. 273: 2329–2335.

Raju T, Bignami A, Dahl D. (1981) In vivo and in vitro differentiation of neurons ans astrocytes in the rat embryo. Immunofluorescence study with neurofilament and glia filament antisera. Dev. Biol. 85: 344-351.

Rao MV, Houseweart MK, Williamson TL, Crawford TO, Folmer J, Cleveland DW. (1998) Neurofilamentdependent radial growth of motor axons and axonal organization of neurofilaments does not require the neurofilament heavy subunit (NF-H) or its phosphorylation. J. Cell Biol. 143: 171–181.

Rao MV, Engle LJ, Mohan PS, Yuan A, Qiu D, Cataldo A, Hassinger L, Jacobsen S, Lee VM, Andreadis A, Julien JP, Bridgman PC, Nixon RA. (2002) Myosin Va binding to neurofilaments is essential for correct myosin Va distribution and transport and neurofilament density. J. Cell Biol. 159:279-290.

Rao MV, Campbell J, Yuan A, Kumar A, Gotow T, Uchiyama Y. et al. (2003) The neurofilament middle molecular mass subunit carboxyl-terminal tail domains is 223 essential for the radial growth and cytoskeletal architecture of axons but not for regulating neurofilament transport rate. J. Cell Biol. 163: 1021–1031.

Rao MV, Yuan A, Campbell J, Kumar A, Veeranna V, Nixon RA. (2004) The phosphorylated tain domains of neurofilament heavy and medium subunits together regulate neurofilament content but not the rate of slow axonal transport of neurofilaments in optic axons. Program No. 153.6. Society for Neuroscience annual meeting.

Reason AJ, Morris HR, Panico M, Marais R, Treisman RH, Haltiwanger RS, Hart GW, Kelly WG, Dell A. (1992) Localization of O-GlcNAc modification on the serum response transcription factor. J. Biol. Chem.267:16911-16921.

Reinsch SS, Mitchison TJ, Kirschner MW. (1991). Microtubule polymer assembly and transport during axonal elongation. J. Cell Biol. 115: 365–379.

Reles A, Friede RL. (1991) Axonal cytoskeleton at the nodes of Ranvier. J. Neurocytol. 20:450-458.

Robertson J, Beaulieu JM, Doroudchi MM, Durham HD, Julien JP, Mushynski WE. (2001) Apoptotic death of neurons exhibiting peripherin aggregates is mediated by the proinflammatory cytokine tumor necrosis factor-alpha. J. Cell Biol. 155:217–226.

Roder HM, Eden PA, Ingram VM. (1993) Brain protein kinase PK40erk converts TAU into a PHFlike form as found in Alzheimer’s disease. Biochem. Biophys. Res. Commun. 193: 639–647.

Rouleau GA, Clark AW, Rooke K, Pramatarova A, Krizus A, Suchowersky O, Julien JP, Figlewicz D. (1996) SOD1 mutation is associated with accumulation of neurofilaments in amyotrophic lateral sclerosis. Ann. Neurol. 39: 128–131. Sabry J, O’Connor TP, Kirschner MW. (1995). Axonal transport of tubulin in Ti1 pioneer neurons in situ. Neuron 14: 1247–1256.

Sacher MG, Athlan ES, Mushynski WE. (1992) Okadaic acid induces the rapid and reversible disruption of the neurofilament network in rat dorsal root ganglion neurons. Biochem Biophys Res Commun. 186:524-530.

224 Sacher MG, Athlan ES, Mushynski WE. (1994) Increased phosphorylation of the amino-terminal domain of the low molecular weight neurofilament subunit in okadaic acid-treated neurons. 269:18480-18484.

Saito T, Shima H, Osawa Y, Nagao M, Hemmings BA, Kishimoto T, Hisanaga S. (1995) Neurofilament-associated protein phosphatase 2A: its possible role in preserving neurofilaments in filamentous states. Biochemistry 34: 7376–7384.

Sanchez I, Hassinger L, Paskevich PA, Shine HD, Nixon RA. (1996) Oligodendroglia regulate the regional expansion of axon caliber and local accumulation of neurofilaments during development independently of myelin formation.J.Neurosci.16:5095-5105.

Sanchez I, Hassinger L, Sihag RK, Cleveland DW, Mohan P, Nixon RA. (2000) Local control of neurofilament accumulation during radial growth of myelinating axons in vivo. Selective role of site-specific phosphorylation. J. Cell Biol. 151:1013-1024.

Sarria AJ, Nordeen SK, Evans RM. (1990) Regulated expression of vimentin cDNA in cells in the presence and absence of a preexisting vimentin filament network. J. Cell Biol. 111:553-65.

Sasaki T, Taoka M, Ishiguro K, Uchida A, Saito T, Isobe T, Hisanaga S. (2002) In vivo and in vitro phosphorylation at Ser-493 in the glutamate (E)-segment of neurofilament-H subunit by glycogen synthase kinase 3beta. J. Biol. Chem. 277:36032-36039.

Schmidt RE, Plurad SB. (1985). Ultrastructural appearance of intentionally frustrated axonal regeneration in rat sciatic nerve. J. Neuropathol. Exp. Neurol. 44:130-146.

Schnitzer J, Franke WW, Schachner M. (1981) Immunosytochemical demonstration of vimentin in astrocytes and ependymal cells of developing and adult mouse nervous system. J. Cell Biol. 90:435-447.

Schroer TA, Steuer ER, Sheetz MP. (1989) Cytoplasmic dynein is a minus end- directed motor for membranous organelles. Cell 56:937-946. 225 Schwob JE, Farber NB, Gottlieb DI. (1986) Neurons of the olfactory epithelium iin adult rats contain vimentin. J. Neurosci. 6:208-217.

Seeman P. (1967) Transient holes in the erythrocyte membrane during hypotonic hemolysis and stable holes in the membrane after lysis by saponin and lysoslecithin. J. Cell Biol. 32:55-70.

Sengupta A, Wu Q, Grundke-Iqbal I, Iqbal K, Singh TJ. (1997) Potentiation of GSK-3- catalyzed Alzheimer-like phosphorylation of human tau by cdk5. Mol. Cell Biochem. 167: 99–105.

Seitz A, Kojima H, Oiwa K, Mandelkow EM, Song YH, Mandelkow E. (2002) Single-molecule investigation of the interference between kinesin, tau and MAP2c. EMBO J. 21:4896-4905

Shah JV, Flanagan LA, Janmey PA, Leterrier JF. (2000) Bidirectional translocation of neurofilaments along microtubules mediated in part by dynein/dynactin. Mol. Biol. Cell 11:3495-3508.

Shah JV, Cleveland DW. (2002) Slow axonal transport: fast motors in the slow lane. Curr. Opin. Cell Biol. 14:58-62.

Sharma M, Sharma P, Pant HC. (1999) CDK- 5-mediated neurofilament phosphorylation in SHSY5Y human neuroblastoma cells. J. Neurochem. 73: 79– 86.

Shaw G, Osborn M, Weber K. (1981) An immunofluorescence microscopical study of neurofilament triplet proteins, vimentin and glial fibrillary acidic protein with the adult rat brain. Eur. J. Cell Biol. 26: 68-82.

Shaw G, Weber K. (1982) Differential expression of neurofilament triplet proteins in brain development. Nature 298: 277-279.

226 Shaw G, Weber K. (1983) The structure and development of the rat retina: An immunofluorescence microscopical study using antibodies specific for intermediate filament proteins. Eur. J. Cell Biol. 30: 219-232.

Shaw G, Weber K. (1984) The intermediate filament complement of the retina a comparison between different mammalian speicies. Eur. J. Cell Biol. 33:95-104.

Sihag RK, Nixon RA. (1989) In vivo phosphorylation of distinct domains of the 70-kilodalton neurofilament subunit involves different protein kinases. J Biol Chem. 264: 457–464.

Sihag RK, Nixon RA. (1991) Identification of Ser-55 as a major protein kinase A phosphorylation site on the 70-kDa subunit of neurofilaments. Early turnover during axonal transport. J. Biol. Chem. 266:18861-18867.

Singh TJ. Grundke-Iqbal I, Iqbal K. (1995) Phosphorylation of tau protein by casein kinase-1 converts it to an abnormal Alzheimerlike state. J. Neurochem. 64: 1420–1423.

Slaughter T, Wang J, Black MM. (1997). Microtubule transport from the cell body into the axons of growing neurons. J. Neurosci.17: 5807–5819.

Slawson C, Hart GW. (2003) Dynamic interplay between O-GlcNAc and O- phosphate: the sweet side of protein regulation. Curr. Opin. Struct. Biol. 13:631- 636.

Stearns ME, Ochs RL. (1982) A functional in vitro model for studies of intracellular motility in digitonin-permeabilized erythrophores. J. Cell Biol. 94:727- 39.

Sternberger LA, Sternberger, NH. (1983) Monoclonal antibodies distinguish phosphorylated and non-phosphorylated forms of neurofilaments in situ. Proc. Natl. Acad. Sci. USA 80:6126-30.

Steinert PM, Roop DR. (1988) Molecular and cellular biology of intermediate filaments. Annu. Rev. Biochem. 57:593-625.

227 Steinert PM, Marekov LN, Parry DA. (1993a) Diversity of intermediate filament structure. Evidence that the alignment of Coiled-coil molecules in vimentin is different from that in keratin intermediate filaments. J. Biol. Chem. 268:24916- 24925.

Steinert PM, Marekov LN, Fraser RD, Parry DA. (1993b) Keratin intermediate filament structure. Crosslinking studies yield quantitative information on molecular dimensions and mechanism of assembly. J. Mol. Biol. 230:436-452.

Steiner JP, Ling E, Bennett V. (1987) Nearest neighbor analysis for brain synapsin I. Evidence from in vitro reassociation assays for association with membrane protein(s) and the Mr = 68,000 neurofilament subunit. J. Biol. Chem. 262:905-914.

Steven AC, Hainfeld JF, Trus BL, Wall JS, Steinert PM. (1983) the distribution fo mass in heteropolymer intermediate filaments assembled in vitro. STEM analysis of vimentin/desmin and bovine epidermal keratin. J. Biol. Chem. 258:8323-8329.

Stewart M. (1993) Intermediate filament structure and assembly. Curr. Opin. Cell Biol. 5:3-11.

Streifel TD, Avalos RT, Cohlberg JA. (1996) cAMP-dependent phosphorylation of neurofilament proteins NF-L and NF-M inhibits their coassembly into filaments in vitro. Biochem. Biophys. Res. Commun. 222:646-651.

Sun D, Leung CL, Liem R. (1996) Phosphorylation of the high molecular weight neurofilament protein (NF-H) by Cdk5 and p35. J. Biol. Chem. 271: 14245– 14251.

Szaro BG, Lee VM, Gainer H. (1989) Spatial and temporal expression of phosphorylated and non-phosphorylated forms of neurofilament proteins in the developing nervous system of Xenopus laevis. Brain. Res. Dev. Brain Res. 48:87-103.

Takahashi M, Amin N, Grant P, Pant HC. (1995) P13suc1 associates with a cdc2-like kinase in a multimeric cytoskeletal complex in squid axoplasm. J. Neurosci. 15: 6222–6229.

228 Takeda S, Okabe S, Funakoshi T, Hirokawa N. (1994) Differential dynamics of neurofilament-H protein and neurofilament-L protein in neurons. J. Cell Biol. 127: 173–185.

Takeda S, Funakoshi T, Hirokawa N. (1995) Tubulin dynamics in neuronal axons of living zebrafish embryos. Neuron 14: 1257–1264.

Tapscott SJ, Bennett GS, Holtzer H. (1981) Neuronal precursor cells in the chick neural tube express neurofilament protein. Nature 292:836-838.

Terada S, Nakata T, Peterson AC, Hirokawa N. (1996) Visualization of slow axonal transport in vivo. Science 273: 784–788.

Terada S, Kinjo M, Hirokawa N. (2000) Oligomeric tubulin in large transporting complex is transported via kinesin in squid giant axons. Cell 103:141–55.

Terada S, Hirokawa N (2000) Moving on to the cargo problem of microtubule- dependent motors in neurons. Curr Opin Neurobiol 10:566-573.

Terada S (2003) Where does slow axonal transport go? Neurosci Res 47:367- 372.

Terasaki M, Schmidek A, Galbraith JA, Gallant PE, Reese TS. (1995). Transport of cytoskeletal elements in the squid giant axon. Proc. Natl. Acad. Sci. USA 92:11500–11503.

Terry-Lorenzo, R. T., Inoue, M., Connor, J. H., Haystead, T. A., Armbruster, B. N., Gupta, R. P., Oliver, C. J. & Shenolikar, S. (2000) Neurofilament-L is aprotein phosphatase-1-binding protein associated with neuronal plasma membrane and post-synaptic density. J. Biol. Chem. 275, 2439–2446.

Theiss C, Napirei M, Meller K. (2005) Impairment of anterograde and retrograde neurofilament transport after anti-kinesin and anti-dynein antibody microinjection in chicken dorsal root ganglia. Eur. J. Cell Biol. 84:29-43.

229 Tomkins J, Usher P, Slade JY, Ince PG, Curtis A, Bushby K, Shaw PJ. (1998) Novel insertion in the KSP region of the neurofilament heavy gene in amyotrophic lateral sclerosis (ALS). NeuroReport 9:3967–3970.

Toyoshima I, Komiya Y. (1995) Phosphorylation and transport of neurofilament proteins in the rat spinal ganglion. Neurosci. Lett. 189:69-72.

Traverse S, Gomez N, Paterson H, Marshall C, Cohen P. (1992) Sustained activation of the mitogen-activated protein (MAP) kinase cascade may be required for differentiation of PC12 cells. Comparison of the effects of nerve growth factor and epidermal growth factor. Biochem. J. 288: 351–355.

Trojanowski JQ, Obrocka MA, Lee VM. (1985) Distribution of neurofilament subunits in neurons and neuronal processes inmmunohistochemical studies of bovine cerebellum with subunit-specific monoclonal antibodies. J. Histochem. Cytochem. 33:557-563.

Trojanowski JQ, Walkenstein N, Lee VM. (1986) Expression of neurofilament subunits in neurons of the central and peripheral nervous system an immunohistochemical study with monoclonal antibodies. J. Neurosci. 6:650-660.

Trojanowski JQ, Kelsten ML, Lee VM. (1989) Phosphate-dependent and independent neurofilament protein epitopes are expressed throughout the cell cycle in human medulloblastoma (D283 MED) cells. Am. J. Pathol. 135:747-758.

Uchida A, Yorifuji H, Lee VM, Kishimoto T, Hisanaga S. (1999) Neurofilaments of aged rats: the strengthened interneurofilament interaction and the reduced amount of NF-M. J. Neurosci. Res. 58:337-348.

Uchida A, Tashiro T, Komiya Y, Yorifuji H, Kishimoto T, Hisanaga S.(2004) Morphological and biochemical changes of neurofilaments in aged rat sciatic nerve axons. J. Neurochem. 88(3):735-745.

Uchida A, Brown A. (2004) Arrival, reversal, and departure of neurofilaments at the tips of growing axons. Mol. Biol. Cell. 15:4215-4225.

230 Vale RD, Schnapp BJ, Reese TS, Sheetz MP. (1985a) Organelle, bead, and microtubule translocations promoted by soluble factors from the squid giant axon. Cell. 40:559-569.

Vale RD, Schnapp BJ, Mitchison T, Steuer E, Reese TS, Sheetz MP. (1985b) Different axoplasmic proteins generate movement in opposite directions along microtubules in vitro. Cell. 43:623-632.

Veeranna, Shetty KT, Link WT, Jaffe H, Wang J, Pant HC. (1995) Neuronal cyclindependent kinase-5 phosphorylation sites in neurofilament protein (NF-H) are dephosphorylated by protein phosphatase 2A. J. Neurochem. 64: 2681– 2690.

Veeranna, Amin ND, Ahn NG, Jaffe H, Winters CA, Grant P, Pant HC. (1998) Mitogen-activated protein kinases (Erk1,2) phosphorylate Lys-Ser-Pro (KSP) repeats in neurofilament proteins NF-H and NF-M. J. Neurosci. 18: 4008–4021.

Veeranna GJ, Shetty KT, Takahashi M, Grant P, Pant HC. (2000) Cdk5 and MAPK are associated with complexes of cytoskeletal proteins in rat brain. Brain Res. Mol. Brain Res. 76: 229– 236.

Vickers JC, Vitadello M, Parysek LM, Costa M. (1991) Complementary immunohistochemical distribution of the neurofilament triplet and novel intermediate filament proteins in the autonomic and sensory nervous system of the guinea-pig. J. Chem. Neuroanat. 4:259-270.

Vickers JC, Costa M. (1992) The neurofilament triplet is present in distinct subpopulations of neurons in the central nervous system of the guinea-pig. Neuroscience 49:73-100.

Vickers JC, Chiu FC, Costa M. (1992) Selective distribution of the 66-kDa neuronal intermediate filament protein in the sensory and autonomic nervous system o the guinea-pig. Brain Res. 585:205-211.

Vickers JC, Morrison JH, Friedrich VL Jr, Elder GA, Perl DP, Katz RN, Lazzarini RA. (1994) Age-associated and cell-type-specific neurofibrillary pathology in

231 transgenic mice expressing the human midsized neurofilament subunit. J. Neurosci. 14:5603-5612.

Wagner OI, Ascano J, Tokito M, Leterrier JF, Janmey PA, Holzbaur EL. (2004) The interaction of neurofilaments with the microtubule motor cytoplasmic dynein. Mol. Biol. Cell. 15:5092-5100.

Wang KC, Kim JA, Sivasankaran R, Segal R, He Z. (2002) P75 interacts with the Nogo receptor as a co-receptor for Nogo, MAG and OMgp. Nature. 420:74-78.

Wang L, Ho CL, Sun D, Liem RK, Brown A. (2000) Rapid movement of axonal neurofilaments interrupted by prolonged pauses. Nat. Cell Biol. 2:137-141.

Wang L, Brown A. (2001) Rapid intermittent movement of axonal neurofilaments observed by fluorescence photobleaching. Mol. Biol. Cell. 12:3257-3267.

Wang L, Brown A. (2002) Rapid movement of microtubules in axons.Curr. Biol. 12:1496-1501.

Watson DF, Griffin KP, Fittro KP, Hoffman PN. (1989) Phosphorylation- dependent immunoreactivity of neurofilaments increases during axonal maturation and beta, beta’-iminodipropionitrile intoxication. J. Neurochem. 53: 1818-1829.

Willard M, Simon C. (1983) Modulations of neurofilament axonal transport during the development of rabbit retinal ganglion cells. Cell. 35:551-559.

Wong NK, He BP, Strong MJ. (2000). Characterization of neuronal intermediate filament protein expression in cervical spinal motor neurons in sporadic amyotrophic lateral sclerosis (ALS). J. Neuropathol. Exp. Neurol. 59:972–982.

Wong PC, Cleveland DW. (1990) Characterization of dominant and recessive assembly-defective mutations in mouse neurofilament NF-M. J. Cell Biol. 111:1987-2003.

232 Wong PC, Marszalek J, Crawford TO, Xu Z, Hsieh ST, Griffin JW, Cleveland DW. (1995) Increasing neurofilament subunit NF-M expression reduces axonal NF-H, inhibits radial growth, and results in neurofilamentous accumulation in motor neurons. J. Cell Biol. 130:1413-1422.

Wong ST, Henley JR, Kanning KC, Huang KH, Bothwell M, Poo MM. (2002) A p75(NTR) and Nogo receptor complex mediates repulsive signaling by myelin- associated glycoprotein. Nat. Neurosci. 5:1302-1308.

Willard MB, Simon C. (1983) Modulations of neurofilament axonal transport during the development of rabbit retinal ganglion cells, Cell 35: 551-559.

Xia CH, Roberts EA, Her LS, Liu X, Williams DS, Cleveland DW, Goldstein LS. (2003) Abnormal neurofilament transport caused by targeted disruption of neuronal kinesin heavy chain KIF5A. J. Cell Biol. 161:55-66.

Xu Z, Cork LC, Griffin JW, Cleveland DW. (1993) Increased expression of neurofilament subunit NF-L produces morphological alterations that resemble the pathology of human motor neuron disease. Cell 73:23–33.

Xu ZS, Liu WS, Willard MB. (1992) Identification of six phosphorylation sites in the COOH-terminal tail region of the rat neurofilament protein M. J. Biol. Chem. 267:4467-4471.

Yabe JT, Pimenta A, Shea TB. (1999) Kinesin-mediated transport of neurofilament protein oligomers in growing axons. J Cell Sci. 112:3799-3814.

Yabe JT, Jung C, Chan WK, Shea TB. (2000) Phospho-dependent association of neurofilament proteins with kinesin in situ. Cell Motil. Cytoskeleton. 45:249- 262.

Yabe JT, Chan WK-H, Chylinski TM, Lee S, Pimenta A, Shea TB (2001) The predominant form in which neurofilament subunits undergo axonal transport varies during axonal initiation, elongation and maturation. Cell Motil. Cytoskeleton 48:61-83.

233 Yamasaki H, Itakura C, Mizutani M. (1991) Hereditary hypotrophic axonopathy with neurofilament deficiency in a mutant strain of the Japanese quail. Acta Neuropathol. 82:427–434.

Yamashita T, Higuchi H, Tohyama M. (2002) The p75 receptor transduces the signal from myelin-associated glycoprotein to Rho. J Cell Biol. 157:565-570.

Yang Y, Dowling J, Yu QC, Kouklis P, Cleveland DW, Fuchs E. (1996) An essential cytoskeletal linker protein connecting actin microfilaments to intermediate filaments. Cell. 86:655-665.

Yen SH, Fields KL. (1981) Antibodies to neurofilament, glial filament, and fibroblast intermediate filament proteins bind to different cell types of the nervous system. J. Cell Biol. 88: 115-126.

Yin X, Crawford TO, Griffin JW, Tu P, Lee VM, Li C, Roder J, Trapp BD. (1998) Myelin-associated glycoprotein is a myelin signal that modulates the caliber of myelinated axons. J. Neurosci. 18:1953-1962.

Yoshihara T, Yamamoto M, Hattori N, Misu K, Mori K, Koike H, Sobue G. (2002) Identification of novel sequence variants in the neurofilament-light gene in a Japanese population: analysis of Charcot-Marie-Tooth disease patients and normal individuals. J. Peripher. Nerv. Syst. 7:221–224.

Yu W, Baas PW. (1995) The growth of the axon is not dependent upon net microtubule assembly at its distal tip.J Neurosci. 15:6827-6833.

Yu W, Schwei MJ, Baas PW. (1996) Microtubule transport and assembly during axon growth. J. Cell Biol. 133; 151–157.

Yuan A, Rao MV, Kumar A, Julien JP, Nixon RA. (2003) Neurofilament transport in vivo minimally requires hetero-oligomer formation. J. Neurosci. 23:9452-9458. Erratum in: J. Neurosci. 23:10463-10465.

Zhu Q, Couillard-Despres S, Julien JP. (1997) Delayed maturation of regenerating myelinated axons in mice lacking neurofilaments. Exp. Neurol. 148: 299–316. 234 Zhu Q, Lindenbaum M, Levavasseur F, Jacomy H, Julien JP. (1998) Disruption of the NF-H gene increases axonal microtubule content and velocity of neurofilament transport: relief of axonopathy resulting from the toxin beta, beta’- iminodipropionitrile. J. Cell Biol. 143: 183–193.

235