ELECTRON MCROSCOPICAL INVESTIGATIONS OF MYELIN BASIC PROTEIN:

LIPID INTERACTIONS AND THEIR IMPLICATIONS IN MULTIPLE SCLEROSIS

A Thesis

Presented to

The Faculty of Graduate Studies

of

The University of Guelph

by

DANIEL ROBERT BENIAC

In partial fulfilment of requirements

for the degree of

Doctor of Philosophy

April, 1998

O Daniei Robert Beniac, 1 998 National Library Bibliothèque nationale 1*1 ofCanada du Canada Acquisitions and Acquisitions et Bibliogtaphic Services services bibliographiques 395 Wellingtori Street 395, rue Wellington OîîawaON KIAM OtfawaON K1AW canada canada

The author has granted a non- L'auteur a accordé une licence non exclusive licence allowing the exclusive permettant à la National Libmy of Canada to Bibliothèque nationale du Canada de reproduce, loan, distriiute or seli reproduire, prêter, distn'buer ou copies of this thesis in microform, vendre des copies de cette thèse sous paper or electronic formats. la forme de microfichelnlm, de reproduction sur papier ou sur format électronique.

The author retains ownership of the L'auteur conserve la propriété du copyright in this thesis. Neither the droit d'auteur qui protège cette thèse. thesis nor substantial extracts fiom it Ni la thèse ni des extraits substantiels may be printed or otherwise de celle-ci ne doivent être imprimés reproduced without the author's ou autrement reproduits sans son permission. autorisation. ABSTRACT

ELECTRON MICROSCOPICAL INVESTIGATIONS OF MYELIN BASIC PROTEIN:

LIPID INTERACTIONS AND THEIR IMPLICATIONS IN MULTIPLE SCLEROSIS

Daniel Robert Beniac Supervisor:

University of Guelph. 1997 Professor George Harauz

In the development of multiple sclerosis (MS), it is suspected that post translational modifications of myelin basic protein (MBP)cause a loss in cationicity in MBP. which in tum leads to myelin destabilization. In normal individuals. the CI charge isomer predominates. In MS individuals the less cationic C8 isomer predominates: in a rare form of MS (Marburg type) an even less cationic isomer MC8 is found. Due to the complexity of the myelin sheath. it is not possible to study in vivo. to high resolution. the ultrastructure of the myelin sheath and its components. In vitro studies of the isolated components are consequently the best alternative by which to study myelin. Electron microscopie examination of different isomers of MBP interacting with a lipid monolayrr of phosphatidylserine and monosialoganglioside GMl revealed the isomers to behave distinctly. It was found that C8, Cl, and MC8 had increasing arnounts of interaction with the lipid monolayer. Single particle computer image analysis was applied to each charge isomer population of MBP particles to produce 2D and 3D density maps. The 3D reconstructions were solved to 2.5 nm resolution. The reconstructions have three basic components, the cap, annular ring, and base. If one considers the Cl reconstruction to be "normal" (predominant species of MBP in a hedthy individuaij then the C8 reconstruction appears to be a compacted version of it and the MC8 reconstructions are the most extended, with potential C3 symmetry. The exact composition of the reconstmctions are unknown, but two models are proposed. Either (i) the complex is composed of a single MBP particle with a major lipid component. or (ii) the cornplex is a hexamer of MBP. with a minor lipid component. Mode1 (ii) is favoured based on the potential C3 syrnmetry. Based on the results of this in vitro analysis, it is possible to hypothesize that the different MBP charge isomers interact uniquely with lipids at the molecular level. At present, the development of MS is unknown; this study provides information on the putative nature of the disturbance of myelin at the molecular level. 1

ACKNOWLEDGEMENTS

1 wish to thank Dr. Terry Beveridge. Dr. Rickey Yada Dr. John Phillips. and especially Dr. George Haraw for their assistance and counselling. 1 would also like to extend my gratitude to Dr. Thomas Tompkins and Maria Luckevich who helped initiate this project. 1 also wish to thank Dr. Mario Moscarello. Dr. Denise Wood. and Teresa

Miani for assistance in the purification of myelin basic protein, providing me with human myelin basic protein, especially the recently characterised Marburg type MBP. and providing me with antibodies to myelin basic protein. 1 would like to extend my special gratitude to Bob Harris whose technical knowledge about electron microscopy over the years has been invaluable to me. 1 aiso wish to thank Dr. Peter Ottensmeyer and Allan

Fernandes for use of, and assistance in, the operation of the scanning transmission electron microscope, and Dr. Greg Czarnota for assistance with IQAD. I would also like to thank my labmates; Ross Ridsdale for help with Insight, Dr. Nades Palaniyar for the TEM imaging of hC1, and Wen Li for help with IQAD. TABLE OF CONTENTS

Page # Subject vi LIST OF TABLES v LIST OF FIGURES viii LIST OF ABBREVIATIONS

CHAPTER 1: Introduction

1.1: Introductory précis 1.2: Myelin sheath 1.3: Multiple sclerosis 1.4: Myelin basic protein 1.5: Protein structure determination by electron microscopy 1.6: Statement of thesis

CHAPTER 2: Transmission electron rnicroscopic imaging, and image analysis. of bovine Cl charge isomer of myelin basic protein

summary Lipid monolayers and bovine myelin basic protein. Materials and methods Purification and characterisation of bovine myelin basic protein Analysis of bovine myelin basic protein by polyacrylarnide gel electrophoresis Silver staining polyacrylamide gels Western blot analysis of bovine myelin basic protein Preparation of plastic carbon coated grids Lipid monolayer technique for preparation of myelin basic protein for EM imaging Electron microscopy of bovine myelin basic protein Immuno-gold electron rnicroscopy of bovine myelin basic protein Single particle countdunit area for bovine myelin basic protein Two-dimensional image analysis of bovine myelin basic protein Three-dimensional reconstruction of bovine myelin basic protein Results Isolation and electron microscopy of bovine Cl 65 2.4.2: Two-dimensional analysis and three-dimensional reconstruction of bovine C 1 75 2.5: Discussion 77 2.5.1: Computer image analysis 82 2.6: Conclusions

83 CHAPTER 3: Electron microscopic imaging, and image analysis. of human myelin basic protein charge isomers

Smary Electron microscopy and rnyelin basic protein Materials and methods Purification and characterisation of human myelin basic protein Analysis of human myelin basic protein by polyacrylamide gel electrophoresis and silver staining polyacrylamide gels Western blot analysis of human myelin basic protein Antibody assay for hurnan myelin basic protein Preparation of plastic carbon coated grids, and lipid monolayer technique for preparation of myelin basic protein for EM imaging Electron microscopy of human myelin basic protein Immuno-gold electron microscopy of hurnan rnyelin basic protein Single particle countdunit area for human myelin basic protein Two-dimensional image analysis of human myelin basic protein Three-dimensionai reconstruction of human myelin basic protein Results Isolation, immuno-gold electron microscopy, and single particle counts of human myelin basic protein Two-dimensional analysis of human myelin basic protein TEM images Three-dimensional analysis of human myelin basic protein TEM images Two-dimensionai analysis of human myelin basic protein STEM images Three-dimensional analysis of human myelin basic protein STEM images Discussion Interaction of myelin basic protein with PS/GM, lipid monolayers 2D computer image analysis of TEM and STEM images of human myelin basic protein 3D computer image analysis of TEM and STEM images of human myelin basic protein Conclusions CHAPTER 4: Thesis conclusion Future research

REFERENCES

APPENDIX: The ribosome: a test of angular reconstitution

A-1: A bstract A-2: Introduction A-3 : Materials and methods A-4: Results and discussion A-5 : Concluding remarks LIST OF TABLES

Page # Subject

CHAPTER 2

64 2- 1: t-Test for irnmuno-gold counts of bovine C 1.

CHAPTER 3

103 3-1 : t-Test for antibody assay of human MBP. 11 1 3-2: t-Test for immuno-gold counts of hurnan MBP. 1 17 3-3: t-Test for particle counts of human MBP. LIST OF FIGURES

Page # Subject

CHAPTER 1

1- 1: Electron micrograph of bovine myelin sheath. 1-2: Chernical structure of arginine and citnilline. 1-3: Arnino acid sequence of rnyelin basic protein.

CHAPTER 2

Cornputer image analysis: an ovemiew. Schematic representation of the lipid monolayer technique. Holey plastic support film. Lipids used to produce the lipid rnonolayer for myelin basic protein. Photographs of bovine brain. Protein incubation container. Purification and biochemical characterization of bovine rnyelin basic pro tein. Electron micrographs of bovine C 1: low salt. Electron micrographs of bovine C 1: high salt. Particle counts for bovine C 1. Irnmuno-gold electron rnicroscopy of bovine Cl on PSIG,, lipid mono layers. Immuno-gold counts for bovine C 1. Class averages of bovine C 1 produced by MSA. Selected "Mn classes of bovine C 1 produced by MSA. Selected "C" classes of bovine Cl produced by MSA. Resolution calculation of 2D averages of bovine C 1. 3D angular orientations of bovine C 1, plotted cartographically as points on a hemispherical surface. 3D Reconstniction of bovine C 1: low salt. 3D Reconstruction of bovine C 1: high salt. Resolution calculation of 3D reconstmctions of bovine C 1. Cornparisons of TEM reconstructions of bovine Cl.

3-1 : Biochernical characterization of human myelin basic protein. 3-2: Antibody assay for human MBP. vii

3-3: Immuno-gold electron microscopy of PSIG,, lipid monolayers.

Immuno-gold electron microscopy of hurnan Cl on PS/GMI lipid monolayers. Imrnuno-gold electron microscopy of hurnan C8 on PS/GMl lipid monolayers. Irnrnuno-gold electron microscopy of human MC8 on PS/GMl lipid monolayers. Immuno-gold MBP complexes. Immuno-gold counts for human MBP. Electron rnicrographs (TEM) of lipid rnonolayers and human MBP. Particle counts for human MBP. Class averages of particles produced on lipid monolayers of PS/GMland high salt buffer. TEM class averages of human Cl produced by MSA. TEM class averages of hurnan C8 produced by MSA. TEM class averages of human MC8 produced by MSA. TEM class members and averages of Cl produced by MSA. TEM class members and averages of C8 produced by MSA. TEM class members and averages of MC8 produced by MSA. TEM resolution calculation of 2D averages of human myelin basic protein. 3D Reconstruction of human Cl fiom TEM images. 3D Reconstruction of hurnan C8 from TEM images. 3D Reconstruction of human MC8 from TEM images. Angular orientations of myelin basic protein TEM 3D reconstructions. plotted cartographically as points on a hemispherical surface. TEM C 1 3D reconstruction; class averages, fonvard-projections, and shaded surface representations. TEM C8 3 D reconstruction; class averages, forward-projections, and shaded surface representations. TEM MC8 3D reconstruction; class averages, forward-projections. and shaded surface representations. TEM resolution caiculations of 3D reconstructions of hurnan myelin basic protein. Cornparisons of 3D TEM reconstructions of human myelin basic protein Electron micrographs (STEM) of human myelin basic protein. STEM class averages of human Cl produced by MSA. STEM class averages of human C8 produced by MSA. STEM class averages of human MC8 produced by MSA. STEM class members and averages of Cl produced by MSA. STEM class members and averages of C8 produced by MSA. STEM class members and averages of MC8 produced by MSA. STEM resolution calculation of 2D averages of hurnan myelin basic protein. 3D Reconstruction of human Cl fiom STEM images. 3D Reconstruction of human CS from STEM images. 3D Reconstruction of human MC8 from STEM images. Angular orientations of myelin basic protein STEM 3D reconstructions. ploaed cartographically as points on a hemispherical surface. STEM Cl 3D reconstniction: class averages, fonvard-projections. and shaded surface representations. STEM C8 3 D reconstruction; class averages, fonvard-proj ections. and shaded surface representations. STEM MC8 3D reconstniction; class averages, fonvard-projections. and shaded surface representations. STEM resolution calculations of 3D reconstructions of hwnan myelin basic protein. Cornparisons of 3D STEM reconstnictions of human myelin basic protein. Overlaying TEM and STEM reconstructions. Probing the internal structure of the hurnan Cl TEM reconstruction. Probing the intemal structure of the human Cl STEM reconstruction. Probing the intemal structure of the human C8 TEM reconstruction. Probing the internal structure of the hurnan C8 STEM reconstniction. Probing the internal structure of the human MC8 TEM reconstruction. Probing the intemal structure of the human MC8 STEM reconstniction. Mode1 of the human myelin basic protein complex. Symmetrised human C 1 STEM reconstruction. Symmetrised human C8 STEM reconstruction. Symmetrised human MC8 STEM reconstruction. Neuraminidase and MBP-possible structural correlations? Fitting MBP reconshuctions into the myelin sheath.

APPENDIX

Schematic of ribosomal subunits. 3D reconstruction of E. coli 30s ribosomal subunit. 3D reconstniction of E. coli 30s ribosomal subunit and 3D location of ribosomal proteins. 3D reconstniction of E. coli 50s ribosomal subunit. 3D reconstruction of E. coli 30s and 50s ribosomal subunits. 3D reconstruction of E. coli 305 and 50s ribosomal subunits, shown as a functional entity . 3D reconstniction of T. Zanuginosus 40s ribosomal subunit. Conparison of the structures of the E. coli 30s (prokaryotic) and ir lanuginosus 40s (eukaryotic) small ribosomai subunits. LIST OF ABBREVUTIONS 2D two-dirnensional 3D three-dimensiond bMBP bovine myelin basic protein dW,O deionized, distilled water Cl-C6. C8 charge isomers of myelin basic protein CAB cellulose acetate butyrate CAPS 3-(cyclohexylamino)- 1-propanesulfonic acid CCD charge coupled device CM52 carboxyrnethyl cellulose CNS central nervous system CSF cerebrospinal fluid DHAA dehydroabietylamine EAE experimental autoimmune encephalomyelitis EM electron microscopy FTIR Fourier transform infared

GM 1 monosialoganglioside GM, KAC hierarchical ascendant classification HLA human leukocyte antigen hMBP human myelin basic protein IEM immunoelectron microscopy [QAD iterative quaternion-assisted angular determination MAG myelin associated glycoprotein MBP myelin basic protein MC8 Marburg type C8 isomer of myelin basic protein MMBP Marburg type myelin basic protein (human) MOG myelin oligodendrocyte glycoprotein MRI magnetic resonance imaging MS multiple sclerosis MSA multivariate statistical analysis MMS Marburg type multiple sclerosis NMR nuclear magnetic resonance PAGE pol yacry lamide gel electrophoresis PE perivenous encephalomyelitis PLP proteolipid protein PNS penpheral nervous system PS phosphatidy lserine SDS sodium doedecyl sulfate STEM scanning transmission electron microscope TEM transmission electron microscope w/v weight per volume CHAPTER 1

INTRODUCTION

1.1: Introductory précis

Multiple sclerosis (MS) is one of the major autoimmune diseases of the central nervous system (CNS). At the root of this severe and debilitating condition lies a disturbance in myelin. It is essential to determine, to high resolution, the ultrastructure of the myelin sheath in normal and MS-afflicted individuals in order to be able to understand the molecular etiology of these pathologies. The lipid and protein composition of the myelin membrane is well known, but a detailed comprehension of the various lipid- protein and protein-protein interactions requires high resolution structures of these complexes, which are presently lacking. The primary reason for this lack of information is that myelin oligodendrocytes, the myelin forming cells of the CNS. grow in intimate contact with many other cells, and thus these cells cannot be isolated with their native structure intact. The only way to study to high resolution the structure of myelin is by the in vitro investigation of its isolated components, alone and in concert with other membrane constituents. Through this, one can determine likely interactions in vivo within the intact myelin sheath and changes in these interactions which lead to destabilization of myelin.

The two major proteins of the myelin sheath are myelin basic proteins (MBPs) and proteolipid proteins (PLPs). These proteins in particular, as membrane proteins in generai, have not and cannot easily be crystallised for X-ray diffiactometry. Electron 2 microscopy is the best direct method of structure determination that cm be used to study

the structures of these proteins both in isolated form and in defined complexes with one

another and with lipids. The three-dimensional structures can be determined to high

resolution by computerised image analysis of the randomly oriented proteins on lipid

monolayers created at an air-water interface. This study will yield data on potentiallp different tertiary and quaternary structures of specific microheteromers of MBP. when it

interacts with lipids. MBP appears to be the central player in the pathogenesis of MS.

Knowledge of the tertiary structures of MBP charge isomers. and their organisation on

lipid layers and within the compacted myelin multilayers, will be attained by this study and will enhance the understanding of the mechanisms of myelin destabilisation in MS.

The MBPs are a family of basic proteins encoded by a single gene. Through a

mechanism of aitemate gene splicing, a number of isoforms are formed. In the bovine

and hurnan brains the 18.5 kDa isoform is the dominant one. Minor 17 kDa and 2 1.5

kDa isoforms are also present. The 1 8.5 kDa isoform consists of 170 amino acids of which 12 are lysines and 19 are arginines which give the molecule its basic character.

The absence of cysteine suggests that disulphide bonds do not stabilize the structure.

In addition to various isoforms, the 18.5 kDa isofonn can be resolved into several species called cornponents or charge isomers. These differ from one another by loss of positive charge so that C 1 , the most cationic, has a net positive charge of + 1 9. C2 has a net positive charge of +18, C3 has a net positive charge of + 17, etc. C8 is the least cationic, and has a net positive charge of +12. These various charge isomers can be separated on carboxymethyl cellulose columns at pH 10.5. In this way they are available 3

in pure form to study.

Knowledge of the structure of MBP, and its association with lipids. is essential to

understanding the mechanism by which MBP stabilises the myelin membranes. MBP (as

many other membrane proteins) has not proved crystallisable for structural snidies by X-

ray diffiactometry, and nuclear magnetic resonance WMR) is not an appropriate

alternative in this instance. High resolution cryo transmission electron microscopy (TEM)

and cryo scanning transmission electron microscopy (STEM) are novel and feasible

approaches to determine the structures of specific microheteromers of MBP complexed

with lipid. This technique will yield results directly relevant to that of the natural proteins

in the myelin sheath, and forms a basic expenmental system for investigation of

conformational transitions of MBP and its interaction with lipids. MBP isolated frorn

victims of MS is different from that of normal, age-matched controls. Thus. a knowledge

of the structure and interactions of the various charge isomers with lipids will help us to

understand the nature of the instability of the myelin sheath in MS.

This thesis is divided into three chapters, and one appendix. The first chapter

provides background information on the topics exarnined. The second chapter drals specifically with the cryoTEM investigation of the CI charge isomer of bovine myelin

basic protein (bMBP). This chapter focuses on the technicai development of the lipid monolayer technique, and its application to MBP. The third chapter deals specifically with three different charge isomers of human MBP (hMBP) associated with heaithy and

MS afflicted people. This chapter addresses directly the dynarnic and structural differences between these different charge isomers of MBP when they interact with a lipid 4 monolayer. This chapter also describes two electron rnicroscopic techniques. cryoTEM. and low dose cryoSTEM. The appendix at the back of this thesis demonstrates the use of the IQAD angdar reconstitution process on electron images of ribosomal subunits.

This appendix provides an example for the angular reconstitution of a "known" 3D structure, and serves as a control for the determination of the "unknown"3D structure of the MBP-lipid cornplex.

1.2: Myelin sheath

This investigation focuses on the structure and dynarnics of MBP-lipid interactions. and their possible implications to myelin stabilisatioddestabilisation in the development of MS. This section provides a brief overview of the structure and function of the myelin sheath.

Myelin is found in both the central nervous system (CNS) and peripherd nervous system (PNS). It is involved in the conduction of electrical impulses in both sensory and motor nerves. In the CNS the myelin is formed fiom an oligodendrocyte. and in the PNS the myelin is formed fiom a Schwann cell. Myelin is composed of tightly packed repeating layers of lipids and proteins. The myelin sheath can be composed of several to over one hundred layers of plasma membrane wrapped around an axon (Fig. 1-1). The lipid component has a high concentration of cholesterol and phospholipids. The major function of myelin is to act as an insulator to increase the speed of conduction of impulses dong the axon. A demyelinated axon will have a markedly slower speed of conduction than a myelinated mon. An example of conduction can be seen in the unmyelinated giant 5 squid axon which is 500 Fm in diameter. This mon has a conduction velocity of 25 m/second. If conduction velocity were proportional to fibre radius, a human nerve fibre

(myelinated) with a 10 pm diameter would conduct at 0.5 mkecond. This however is not the case; due to myelination the human nerve conducts at 50 m/second. This value is twice the speed of the squid giant mon, thus demonstrating that the increase in conduction velocity is attributable to rnyelination. The myelin sheath does not encase the entire axon.

There are a series of regular interruptions in the myelin called nodes of Ranvier ( 1 Fm wide). The myelin sheath increases the capacitance of the mon, by allowing action potentials only to be generated at the nodes of Ranvier. The process by which a nerve impulse is conducted in "jumps" between the nodes of Ranvier is called saltatory conduction. In MS the myelin sheath is degraded, and thus the ability of the axon to propagate nerve impulses is decreased.

Upon doser examination of the myelin sheath, it becomes apparent that there is a high level of order within this structure. in Fig. 1-1 the intraperiod line formed by the apposition of the extracellular surfaces, and major dense line formed by the apposition of the cytoplasmic surfaces have been identified. Both histological and polarized light studies at the beginning of this century demonstrated that myelin has both lipid and protein components, with the lipid being the major component (Morell, 1977). Myelin is a relatively dehydrated structure; it has been estimated that the myelin in white matter was 40% water, whereas non-myelinated white matter was 80% water. Chemically. human myelin is 30% protein and 70% lipid (dry weight). The lipid fraction is composed oE 27.7% cholesterol; 27.5% galactolipid; 22.7% cerebroside; 3.8% sulfatide; and 43.1% mintraperiod line

major dense line

Figure 1-1: Electron micrograph of bovine myelin sheath. This figure shows a cross section of a CNS myelin sheath and axon (a), and a close-up view of the myelin sheath

(b). In (b) the intraperiod line (extracellular), and major dense line (cytoplasmic) have been identified. The myelin sheath shown in this figure was fixed with glutaraldehyde and osmium tetroxide, and was imbedded in LR White plastic, and subsequently sectioned for EM visualization. Scale bar 6 Pm. 7 phospholipid (al1 values are expressed as % total lipid weight). The major lipid type of myelin is the phospholipids (43.1%), which are composed of 15.6% ethanolamine phosphatides, 11 -2% choline phosphatides, 4.8% serine phosphatides, 0.6% inositol phosphatides, 7.9% sphingomyelin, and 12.3% plasmalogens (Morell. 1 977). The separation of the protein fraction of myelin yields two major proteins. the proteolipid protein farnily (with PLP being the major species) 50%, and myelin basic protein familp

(MBP) 2025% (Boggs and Moscarello, 1978b; Deber and Reynolds, 1991: Lees and

Brostoff, 1984; Rumsby, 1978; Smith 1992; Staugatis et al.. 1996) of the dry weight.

There are also severai minor proteins present in myelin. There are two proteins collectively referred to as DM20 (Agrawal et al., 1972), at least one glycoprotein (Quarles ef ai., 1972. 1973a), and a family of high molecular weight proteins. This snidy focuses on the 18.5 kDa isoform of MBP. Its role in the development of MS. and biochemical properties are further addressed in Sections 1.3 and 1.4 of this chapter.

t .3: Multiple sclerosis

Multiple sclerosis (MS) was first detected over a centuty and a half ago (Carswell.

1838; Compston, 1988). MS is an inflarnrnatory demyelinating disease of the CNS. In

MS, cells of the immune system destroy the myelin. In regions of severe demyelination multiple hardened sclerotic lesions (plaques) are formed (hence the narne multiple sclerosis). MS is a relapsing-remitting disease, with both acute and chronic types. At present the etiology of MS is still unknown. What is known about MS is that both environmental, and genetic factors are involved in the development of the disease. When 8 examining the global distribution of the occurrences of MS it is apparent that the disease is less prevalent in equatorial regions of the globe. Several studies have helped to establish the north-south distribution of MS; in northem Scotland 175 pet 100.000 have

MS (Robertson and Compston. 1995), in southern England 99 per 100.000 have MS

(Robertson and Compston. 1999, in Enna, Sicily 53 per 100.000 have MS (Dean et al..

1979), and in Malta 4.2 per 100,000 individuals develop MS (Vassal10 er aL. 1979).

Migration also displays a north-south trend in the development of MS. The data indicate that if an individual (European Caucasian) moves to a southem environment post adolescence, from a northem environment they carry the sarne risk of developing MS as that in the country of ongin. An individual moving south during childhood develops MS with the same risk as one bom in the country (Dean. 1967). The same principle also applies in reverse to Asian and Afro-Caribbean individuals who migrate north. In the case of those individuals who migrate. there is a low prevalence of MS compared to the rest of the population, whereas their offspring have the sarne incidence of MS as the rest of the population (Elian et al., 1990). These data suggests to a potential environmenrd factor in the development of MS; however, to date no such factor(s) hasihave been identified.

The other major component involved in the development of MS is the genetic factor. It has been obsemed that the incidence of MS is low in oriental populations irrespective of where they live (Detels et al., 1977; Kuroiwa et al.. 1983; Miller et al..

1986). These data indicate that perhaps genetic susceptibility exists to the development of MS. Additional studies by Ebers et al. (1986) and Mumford et al. (1994)

10 remyelinated and non-remyelinated lesions exist. The lesions can be found in both the

PNS and CNS. In the CNS, the lesions can be in the grey or white matter (dthough they are more fiequent in the white matter), with the boundaries of the lesions having no bearing on the distribution of the grey or white matter. Thése lesions are commonly found in the optic nerve, spinai chord, brain stem, and the cerebellum. These lesions can be detected by magnetic resonance imaging (MN);however they are not unique to MS.

Similar lesions are found in disserninated encephalomyelitis (Kesselring et al.. 1990). sarcoidosis (Miller et al., 1988), and cerebral vasculitis (Miller et al., 1987). Analysis of the cerebrospinal fluid (CSF) of an MS individuai will show an elevation in the ce11 count

(more than 60 per mm3), and an increase in IgG concentration. As was the case with the

MN information, the CSF results, although indicative of MS, are not unique to MS. In combination, the MRI and CSF data provide strong evidence for the potential of an individual having MS. The only way to unequivocally confi~rrnMS is by autopsy.

In MS afflicted individuals, a number of sharply contoured plaques are present ranging in size fkom 2-10 mm in diameter in sections of the brain. In chronic active plaques macrophages are concentrated at the edges of these plaques. In the rnajority of the patients with long term MS, the plaques are hypocellular, with a glial population of small fibrous astrocytes. Either astroglial fibres occupy the space created by myelin loss. or the tissue has a bare appeamnce with axons lying in a expanded extracellular space

(Barnes ef al., 1991). These plaques are referred to as inactive chronic plaques. Active chronic plaques are lesions with active myelin breakdown. These plaques are identified by the presence of myelinic macrophages, which contain recently ingested myelin 11

(Petrescu. 1969), and are immunoreactive to MBP. myeiin associated glycoprotein

(MAG), and other myelin specific constituents (Itoyarna et al.. 1980: Prineas ef al.. 1984).

The pathogen responsible for the deveiopment of MS is unknown. At present either a viral or bactenai pathogen is suspected. It is known that the immune system has a major role in the progression of the disease. Indirect evidence suggests that MS is an autoimmune disease, which is T cell-mediated. It has been hypothesized that through the process of molecular mimicry the T ce11 is responding to an epitope on the pathogen that bears structural similarity to a myelin component. MBP. PLP. and myelin oligodendrocyte glycoprotein (MOG) are the major candidates in the theory for molecular mimicry which would lead to myelin destruction.

It must be emphasised that at present there is no direct evidence implicating autoreactive T cells, or myelin specific antibodies to the pathogenesis of MS (Martin et al., 1992; Utz and McFarland. 1994). Animal mode1 systems have helped to provide information on the development of MS . Experimental auto immune encephalomyelitis

(EAE) is an animal system that exhibits plaque formation, demyelination, remyelination. and histological similkties to MS. EAE can be induced in genetically susceptible animals by sensitising them to the cornponents of myelin. and it is believed that there are potentiai similarities in the pathogenesis of MS and EAE (Shaw et al., 1988). EAE can be induced specifically by the irnrnunization of animds with whole myelin. or MBP. or

PLP, or encephalitogenic peptides (of MBP or PLP), or lymphocytes from affected animais, or fdly by using clones of T cells specific for MBP or PLP. At present the exact way in which myelin is destroyed in EAE is unknown. It is believed that anti- myelin antibodies recognize the myelin, and activate complement which promotes Fc or complement receptor mediated phagocytosis (Trotter et al., 1986: L imington et al.. 1988:

Goldenberg et al., 1989). Activated macrophages then release myelinotoxic cytokines and proteases (Cammer et al., 1978; Brosnan et al., l988), which lead to myelin degradation.

Oligodendrocytes and myelin appear not to express MHC class 1 or II molecules: therefore, cytotoxic T cells may not be involved in the demyelination process (Wekerle et al., 1994). Morphologically MS and EAE lesions exhibit similar configurations of myelin destruction via infiltration of macrophages.

As mentioned previously, the major argument against autoimmunity in MS is the failure to identify autoreactive T cells and anti-myelin antibodies (Martin et al.. 1992: Utz and McFarland, 1994). Although EAE resembles MS in many ways, there are also differences between the two. EAE is probably closer to the human disease perivenous encephalomyelitis (PE) which is a post vaccination demyelination disease. In cases of PE innumerable rnicroscopic lesions occur within days of onset with no increase in size (Han and Earle, 1975). In MS, the lesions are large fiom the beginning, fewer in nurnber. and continue to increase in size. The exact relationship of EAEPE to MS is unknomn: however, EAE still serves as an excellent mode1 system for the understanding of autoimmunity in the process of myelin degradation, even if the pathogenic nature of the two diseases may differ.

MS is a complex disease, as has been outlined above. To study MS one must also address the fact that there are varied types of MS. In this thesis, MBP isolated from chronic MS and acute Marburg type MS were investigated. The following two paragraphs 13 provide brief information on these two types of MS.

In the case of chronic MS the mean age of onset is 30 years, with a life expectancy of an additional 25 years. In approximately 113 of the cases the patients have little disability for the first 15 of these years (benign MS). In a study by Allen er ul.

(1978), 74% of the MS individuals died of complications caused by the disease. and 26% died fiom causes apparently unrelated to MS. MS has also been reported in association with numerous secondary diseases including: typhus, syphilis, diphtheria Parkinson's disease, tuberculosis, encephalitis lethargica, cervical spond!osis, syringomyelia malabsorption syndrome, ulcerative colitis, myasthenia gravis, systemic lupus erythematosus, systemic arnylodosis, cerebrovascular amyloid angiopathy, eosinophilic vasculitis, ankylosing spondylitis, spongiform encephalopathy, aqueductal stenosis.

Guillain-Barré syndrome, and human irnmunodeficiency virus type 1 (Werner. 1939: Brain and Wilkinson, 1957; Payan and Levine, 1963; Matthews, 1968; Petit et al.. 1969: Aira et ai.. 1974; April and Vansonnenberg, 1976; Heffner et al., 1976: Fantelli. 1978:

Forrester and Lascelles, 1979; Kinney et al., 1979; Tanphaichitr. 1980: Lûssmann rf uZ..

198 1 ; Peiffer, 1982; Rang et al., 1982).

Acute multiple sclerosis (Marburg Type) differs from the chronic MS due to the extremely rapid progression of the disease, with death occurring within 1-6 months frorn clinicd onset. Acute MS was first identified by Marburg (1906, cited in Graham and

Lantos, 1997), and was diagnosed as a type of MS with a rapid development. Other studies have provided support for the similarities of the two forms of MS (Williamson.

1894; Symonds, 1924; Barré et al., 1926; Reuter and Gaupp, 1932, cited in Graham and 14

Lantos, 1997). The lesions of both forms of MS have been identified as histologically identical (Juba, 1939, cited in Graham and Lantos, 1997), including intense hypercellularity, oedema, giant astrocytes, and an absence of glial ce11 fibre formation

(Peters, 1958). The CSF protein and lymphocyte count are normal or moderately elevated in Marburg type MS.

In a recent study of Marburg type MS (MMS) by Wood et al. (1996). the biochemical propertïes of MBP were studied. This study Mersupported the theory that the enzymatic modification of MBP is involved in myelin membrane destabilization in MS

(Moscarello et al, 1994). In this study, MBP was found to be less cationic in MMS individuals than in normal individuals or chronic MS individuals. due the deimination of arginine to citnilline. Citrulline is an arnino acid for which there is no known codon. It arises fiom the post-translational modification of arginine by the enzyme peptidylarginine deiminase (Fig. 1-2). Peptidylarginine deiminase specifically uses arginine in peptide linkage. and thus fiee arginine is not a substrate for this enzyme. Within the myelin sheath MBP exists as a number of charge isomers (Cl-Cg), with C 1 being the most cationic and C8 being the least cationic (there is no C7). In normal individu& the Cl charge isomer of MBP (18.5 kDa isofom) predominates (C8IC1 = 0.82); this isorner has

19 arginine residues, and no citnilline residues (Fig. 1-3). MS afflicted individuals have a CWC1 ratio of 2.5, where the C8 isomer has 13 arginines and 6 citrulline residues. This represents a loss of 6 positive charges. In the Marburg case the C8Kl ratio was drarnatically increased to 6.7, and 18 of the 19 arginines were converted to citdline (a loss of 18 positive charges). The major role of MBP in the myelin sheath is believed to H-c-NH~ H-C-NH~l coo- coo-

Figure 1-2: Chernical structure of arginine and citrulline. The chernical structures of the amino acids arginine (a) and citrulline (b) are illustrated here. There is no known codon for citrulline. Citrulline is produced by the deimination of arginine. and results in the loss of one positive charge. The C8 charge isomer of MBP differs from Cl by the deimination of 6 of the arginine residues to citmlline (Wood and Moscarello. 1989) 16 be its ability to form stable compact multilayers, which is directly attributed to its ner positive charge (Moscarello et al., 1986). The data of this midy indicate that for MBP there rnay be a fingerprint biochemical modification that cm be attributed to myelin destabilization. The differential modifications to MBP in chronic and Marburg Type MS appear to support this hypothesis, and rnay account in part for the rapid progression of

Marburg Type MS. No information was available on the effect of these post-translational modifications on the folded structure of MBP until the work on this thesis was complete.

1.4: Myelin basic protein

The protein fiaction of central nervous system myelin accounts for approximately

30% of the dry weight, and includes MBP (Fig. 1-3) and PLP: 20-25% and 50°/i. respectively (Boggs and Moscarello, 1978b; Deber and Reynolds. 1991 : Lees and

Brostoff, 1984; Rurnsby, 1978; Smith, 1992; Staugaitis et al., 1996). MBP is generally considered, although this is unproven, to be the major candidate for the antigen responsible for autoimmunity in MS. In human myelin, the 18.5 kDa protein (PI - 10.6) is the major isoform due to differential splicing of a single mRNA transcript (Campagoni.

1988). The mRNA transcnpt of MBP is then selectively translocated through the ce11 to a region proximal to the myelin membrane. Once at the membrane. the mRNA is translated by fiee ribosomes and can be incorporated directly into the membrane. This differs from PLP, the other major myelin protein. The PLP mRNA is first translated on the rough endoplasmic reticulum, processed in the Golgi apparatus. and finally transported in vesicles to the myelin membrane (Kalwy and Smith, 1994). a 17 1 ASQKRPSQRH GSKY LATAST MDHARHGFLP

91 WTAHYGSLPQ PVVHFFKN IV 121 TPRTPPPSQG ~GAEGQRPG 151 FGYGGWASDY DAQGTLSKIF 181 KLGGWDSWSG

b 1 ASQKRPSQWH GSKY LATAST MDHARHGFLP 31 RHWDTGILDS [GRFFGGDRG APKRGSGKDS 61 HHPARTAHYG SLPQKSHGRT QDENPVVHFF 91 KNIVTPRTP -PSQGKGRGLS LSRFSWGAEG 121 QWPGFGYGGR ASDYKSAHKG FKGVDAQGTL 151 SKIFKLGGWD SWSGSPMARW

C 1 ASQKW PSQRH GSKYLATAST MDHARHGFLP 31 RHWDTGILDS IGRFFGGDRG APKRGSGKDS 61 HHPARTAHYG SLPQKSHGRT QDENPVVHFF 91 KNIVTPWTP -PSQGKGAEGQ 2PGFGYGGWA 121 SDYKSAHKGF KGVDAQGTLS KIFKLGGRDS 151 WSGSPMAWR

Figure 1-3: Amino acid sequence of myelin basic protein. The arnino acid sequences of die 3

isoforms of human MBP are tisted here. Shown here are (a) the entire 196 residue MBP isofom

(21.4 ma), (b) the 170 residue MBP (18.5 kDa) isoform. (c) and the 159 residue ( 17.3 kDa)

isoform. The 196 residue sequence has ken shaded to show the regions of the sequence that are

rnissing in the smaller isofom. The 18.5 kDa isofom is the most cornmon in humans (and is the focus of this study) and is rnissing residues 59-89 fiom the 196 residue isofom. Whereas the 17.3 kDa isoform is missing residues 59-89, and 132- 142. The unique ûiproline repeat has ben underlined in al1 three figures, and dl the arginine residues have been outiined (Carnegie el af..

1971). 18

The 18.5 kDa isoform of MBP exists as a number of charge isomers which are the result of myriad post-translational modifications (Moscarello er al.. 1992: Ulmer. 1988:

Wood and Moscarello, 1996). The charge isomers which mise from the microheterogeneity of the protein can be separated on a carboxymethyl cellulose (CM52) column, and are narned Cl to C6, and C8, of which Cl is the most cationic and considered to be the least modified (Chou et al., 1976). The differences between the charge isomers are as follows: C2 differs from C 1 essentially by dearnidation. whereas

C3. C4, and CS differ by combinations of dearnidation and phosphorylation. C8 has the most drarnatic modification, where 6 of the 19 arginines have been converted to citrulline

(Wood and Moscarello. 1989).

MBP has long been known to be one of the agents in brain or spinal cord hornogenates responsible for EAE, the animal mode1 for MS (Chou et al.. 1983: Kies.

1965; Nag et al., 1993; Shapira et al., 1971). MBP isolated from victims of MS has been show to be considerably less cationic than that from normal, age-matched convols because of a decrease in the most cationic foms (Cl) and a relative increase in the less cationic isomen (narnely C8) (Moscarello et al., 1994). The distribution of MBP charge isomers also varies in myelin: Cl predominating in the major dense line (cytoplasmic side of plasma membrane), and C8 in the intraperiod line (extracellular side of plasma membrane) (McLaurin et al., 1992, 1993).

Myelin basic protein-lipid interaction is generally believed to be critical for the formation and stability of the multilarnellar myelin sheath although the mechanism by which this occurs is not clearly defined (Boggs and Moscarello, 1978a; Boggs et al.. 198 1. 19

1982; Brady et al., 1979, 198I ; Kirschner and Ganser, 1980: Mateau et al.. 1996: Menon et al., 1990; Moscarello et al., 1986; Sqdzik et al., 1984; Sedzik and Blaurock, 1995; Sixl et al., 1984; Smith, 1977% 1982b. 1992; Smith et al., 1983; ter Beest and Hoekstra

1993). The above rnentioned studies include the in vitro demonstration that MBP can aggregate acidic lipid vesicles and that the lipid reforms into stable multilarnellar layers which resemble in vivo myelin as deterrnined by X-ray diffraction and NMR studies

(Boggs el al.. 1977, 1981, 1982; Fraser et al., 1989; Lampe and Nelsestuen. 1982:

Maggio and Yu, 1989, 1992; Pêli et al., 1987; Rand et al.. 1979). Additional physical- chernical, NMR and Fourier transform infiared spectroscopic (FTIR) studies of MBP interaction with phospholipids have confirmed that the attraction between MBP and lipid is largely electrostatic but also has a hydrophobic component whose magnitude is still debated (Banik and Davison, 1974; Boggs et al., 1981; Demel et al.. 1973: Gould and

London, 1972; Gow et al.. 1990; Hayer-Hartl et al., 1993: Hughes et al.. 1982: Jones and

Epand, 1980; Maggio and Yu, 1989, 1992; Mendz et al.. 1 988, 1 99 1 : Menon er al.. 1990:

Monferran et al.. 1986: Moore. 1982: Nezil el al.. 1992; Pérez-Gil and Keough. 1994:

Persaud et al., 1989; Reid and Bagerl, 1993; Roux et al.. 1994; Sankaram et al.. 1989a.

1989b, 1991; Sixl et al., 1984; Smith, 1982b; Smith et ul., 1983; Suewicz et al.. 1987; ter Beest 1993; Young et al., 1987). The strongest interaction of MBP is with negatively- charged lipids such as cerebroside sulphate, or with gangliosides (Bach and Sela. 1985:

Chan et al., 1990; Demel et al., 1973; Fidelio et al., 1982; MacNaughtan et al.. 1985:

Sela and Bach, 1984). Many spectroscopic studies using circular dichroism (CD) or FTIR have revealed that MBP undergoes a significant conformationai change in the presence 30 of lipid vesicles or detergent micelles, as well as after phoçphorylation (Anthony and

Moscarello, 197 1; Deibler et al.. 1990; Keniry and Smith. 1979. 198 1: Lees and Brostoff.

1984; Men& et al.. 1988, 1991; Moore. 1982; Rarnwani et al.. 1989: Smith. 1977b:

Stuart, 1996; Surewicz et al., 1987), adopting some a-helical and P-sheet structure. The microheterogeneity of MBP gives nse to structural variance and changes in the protein's ability to interact with lipids (Cheifetz and Moscarello, 1983; Jones and Epand. 1980:

Wood and Moscarello, 1989). The conformation of MBP might be predominantly a random coi1 in aqueous solution (Chapman and Moore. 1976; Epand et al.. 1974; Gow and Smith. 1989; Krigbaum and Hsu, 1975), so its structure when associated with lipid is probably more representative of its structure in situ. Several qualitative or semi- quantitative models of the structure of MBP or smdl portions of it. based on biochemical or NMR data. have been published (Fraser and Deber, 1985; Golubovich. 1989: Inouye and Kirschner, 1991; Martenson, 198 1, 1986; Mendz et al.. 1995: Mendz and Moore.

1985; Pnce et al., 1988; Stoner 1984, 1990). However, protein structure prediction algorithms are still in an early stage of development and the models of the whole protein must be verified against direct empirical knowledge.

MBP also interacts with itself (Afshar-Rad et al., 1987; Bellini et al.. 1987:

Moskaitis et al., 1987; Smith, 1980, 1982a) and with other myelin proteins such as PLP

(Edwards et al., 1989; Golds and Braun, 1978). MBP has been found in association with the proteins of the cytoskeleton (Dyer and Benjamins, 1989; Pereyra et al., 1988). where it has been identified as a Ca2+-calmodulinregulated agent for actin polymensation

(Barylko and Dobrowolski, 1984; Chan et al., 1990; Dobrowloski et al., 1986; Roth et 31 al., 1993) and tubulin stabilisation (Pirollet et al.. 1992). MBP has been shown to stimulate phosphoinositide-specific phospholipase C (PI-PLC)activity (Tompkins and

MoscarelIo, 1991, 1993, 1994) and to bind GTP (Chan et al., 1987). It has thus been suggested that MBP may function to transduce cellular signals from the membrane to the

PI-PLC which produces second- messages, and so might be necessary for the coupling of cytoskeletal reorganisation and cellular signalling during myelination and remyelination events. It appears that MBP plays an integral role in a myriad of tasks within the myelin heath. Any information on MBP will thus play a key role in deciphenng the function of this unique protein.

1.5: Protein structure determination by electron microscopy

The two most common current methods of determining the structures of proteins to near atomic resolution are X-ray crystdlography and NMR. The former method requires 3D crystals of suitable size (-0.5 mm or larger) and high regularity (Glasel and

Deutscher, 1995; McRee, 1993). The inability of many proteins, especialIy mernbrane- associated ones, to form such crystals is a significant bottleneck in their structural and functional analyses. NMR is increasingly being used to determine the structures of srnaIl

(< 20 kDa) water-soluble proteins in aqueous solution (Glasel and Deutscher. 1993). It cannot be readily applied to membrane-associated proteins, especially when these are complexed with detergents or lipids to solubilise and enrich them, or when the proteins form aggregates at the high concentrations required (like MBP). The technique certainly has yielded valuable information on MBP, and especially on small segments of MBP --33

(Fraser and Deber, 1985; Koshy et al., 1996; Mendz et al., 1995: Mendz and Moore.

1982, 1985; Price et al, 1988).

High resolution TEM is becoming a viable alternative to the determination of structures of biological macromolecules and their complexes. In the absence of crystals.

TEM can define the size. shape, and quatemary organisation of the macromoIecule in question. TEM can be equally applied to the study of glycoproteins. lipoproteins. proteolipids, nucleoproteins, and, of course, membrane-associated proteins. Advances in

TEM technology and technique over the past two decades have enabled structures to be determined to increasingly higher resolution in both two and three dimensions. CryoTEM allows one to image fieeze-dried or fiozen-hydrated specimens, presumably in their native state. or even simply at low temperatures where radiation induced structural alterations are minimised (Chiu, 1986, 1993). When the macromolecule is imaged in isolated form. computational techniques ("single particle electron crystallography") can be used to determine its 3D structure (Frank, 1996; van Heel et al.. 1996; Yada et al.. 1995). Recent resolution achievements are routinely 3 nrn and often approach 2 nm in 3D reconstruction

(Frank et al., 1995a, b; Schatz et al., 1995; Stark et al.. 1993). There is a report of resolution approaching 1 nm (Czarnota et al., 1994), and it has been stated that there is no fundamental limit to attaining higher resolution by analysing extremely large sets

(>1O*) of macrornolecular images (van Heel et al., 1996). If the specimen is arranged in the form of a 2D crystal, then even lower radiation doses can be maintained. and 2D crystallographic computational approaches can be used to form a projection map of the unit ce11 to very high resolution, permitting fitting of the amino acid primary structure. 23

The same process can be applied to the specimen tilted at different angles in the TEM. to obtain high resolution projection maps of the structure at different orientations which can be applied computationally to give the 3D structure. The successful determination of the structure of bacteriorhodopsin to near-atomic resolution (Grigorieff et al.. 1996:

Henderson et al., 1990), as well as subsequent high resolution analyses of other important complexes (e-g., Jap et al., 1991 ; Wang et al., 1994), sparked a renaissance of interest in the 2D crystailisation of soluble and integral membrane proteins on lipid monolayers and bilayers (Aoyama et al., 1995; Engel et al., 1992; Jap et al.. 1992; Kubalek rf al.. 1994:

Kühlbrandt, 1992; Scheybani et al., 1994, Schmitt et al., 1994). Published resolutions of

1.5- 1.7 nm in 2D projection are almost routine (Perkins et al., 1994; Voges rf al.. 1994).

Another type of regular array, the helical filament, yields 3D (reconstmcted) structures to the resolution range 1.1- 1-6 nm (Akiba et al., 1996; McGough et al.. 1994: Morgan et al., 1995).

The present study combines single particle electron crystallography. with the lipid monolayer technique to study MBP. In doing this the MBP will be in an environment

(lipid) which mimics its native environment in the myelin sheath. This method is superior for imaging MBP compared to the standard method of EM preparation in which proteins are imaged in aqueous solutions. Under these conditions MBP would be in a random coi1 configuration (Chapman and Moore, 1976; Epand et al., 1974: Gow and Smith. 1989:

Krigbaum and Hsy 1975), and not representative of its structure in myelin where it is in intimate contact with the plasma membrane.

The previous paragraphs of this section have dealt with the favourable aspects of 24 protein imaging by electron microscopy (EM), and specifically the advantages of imaging

MBP in association with lipids. However, there is one aspect of electron imaging that must also be considered in this discussion of protein structural determination by EM. namely resolution and its impact on structural determination. In X-ray crystallography a protein crystai is used to produce high resolution diffiaction data. which are subsequently used to calculate an electron density map. From a 5 A map (low resolution for X-ray) one can determine the shape of a molecule, and potentially identifi an a helix as a rod of electron density. With a 3 A rnap (medium resolution for X-ray) it becomes possible to trace the path of a polypeptide chah, and fit the known primary arnino acid sequence into the map. Finally at high resolutions precise structural features can be identified. At 2 A resolution one can discriminate leucine and isoleucine side chains from each other, and at 1 A resolution individual atoms can be seen as discrete balls of density.

In X-ray crystailography the combination of low, medium, and high resolution maps are used dong with cornputer rnodelling to produce a final structure for the protein under investigation (Branden and Tooze, 199 1 ).

In the present investigation the structure of MBP complexed with lipid were solved to a maximum estimated resolution of 9 A in 2D, and 21 A in 3D. Although this is considerably high resolution for EM, it is low resolution compared to X-ray. This fact leads to the question: What structural information can be gathered from the present electron images? Obviously, the present study does not contain information about the atomic structure of MBP on lipid monolayers. Furthemore, at this resolution it is not possible to discriminate lipid from protein in the mass of the final 3D reconstruction. 25

However, the present iimited resolution of this data does not provide grounds for the dismissai of these data. These data provide key structural information on the 2D and 3 D shape of MBP associated with lipid. Based on the shape and volume derived from these reconstructions, it is possible to propose putative models for MBP and lipid within the structural volumes which have been produced. In addition. at this resolution it is possible to discriminate between the different MBP charge isomer-lipid complexes. Although one cannot resolve the molecular structures within the reconstructions. it is possible to differentiate the structures between the various charge isomers of MBP complexed with lipids in the lipid monolayer. Since MBP is believed to have a structurai role in the myelin sheath the ability to differentiate between the isomers when they interact with lipid provides key structural information on MBP where previously there was none. A final point for consideration in this discussion is that many of the techniques used in this study have only recently been developed, especially the angular reconstitution schemes (Farrow and Ottensmeyer, 1992, 1993; van Heel, 1987b). As such the techniques are still under development, and in time continued development in electron imaging, sarnple preparation. and computational analysis techniques can only lead to improved resolution. and subsequent molecular modeiling based on the data produced by electron imaging. The data presented in this study provide impetus for the further development of these electron irnaging techniques. 1.6: Statement of thesis

This thesis describes an investigation of the interaction of the different charge isomers of the 18.5 kDa isoform of MBP with lipid monolayers of phosphatidylsenne and rnonosialogangiioside GM,(PS/GM,) (4: 1) (w/w). Electron microscopy was employed to study the interaction, and high resolution structures of the MBP-lipid associations.

MBP piays an integrai role in the structure and function of the myelin sheath. It is believed that MBP is involved in the development of MS, although the exact cause of

MS, and the role of MBP in the development of MS, is unknown at present. What is known is that in healthy individuals the C 1 charge isomer of MBP predominates. whereas in MS individuals the C8 charge isomer is the dominant species of iMBP. This investigation provides both dynarnic and structurai information on the different charge isomers of MBP that are present in healthy and MS afflicted individuals.

In this study four species of MBP were snidied: the bovine Cl charge isomer

(Beniac et al., 1997a); human Cl and C8 charge isomers; and unique human MBP from an individual afflicted with Marburg type multiple sclerosis (MMS) similar to C8 (referred to as Marburg C8 (MC8)). The Cl charge isomer has a net positive charge of + 19. the

C8 isomer has a net positive charge of +12, and the MC8 isomer has a net positive charge of +I. Prior to this investigation, the principal structural information available on MBP was a secondary structure prediction by Stoner (1984) based on the arnino acid sequence of MBP. The quatemary structures of the three isomers determined in this study were compared, in order to ascertain the effect of these post-translational modifications on folded MBP in association with lipid. MBP is a lipid associated protein, and it is believed that it requires lipid to maintain its native structural configuration. In this study MBP was imaged by electron microscopy (cryoTEM, and cryoSTEM) on simple lipid monolayers of PS/GM, (4: l ) to mimic its adherence to the lipid in the myelin sheath. When MBP was added to the monolayer a statistically larger number of single particles were observed than in its absence. Furthemore, the different charge isomers of MBP exhibited significantly different amounts of single particle production. The MC8 produced the highest number of particles, followed by Cl and finally C8. This indicated that the three charge isomers of MBP with different net positive charge and folded structure have different hctional abilities to associate with the lipid monolayer of PSIG,,, (4:l).

Electron images were analyzed by the computer image analysis techniques of single particle analysis (van Heel et al., 1996), and angular reconstitution (Fanow and

Ottensmeyer, 1992, 1993; Czamota et al., 1994; Beniac et al., 1997a). The 3D analysis of the different charge isomers on the PSIG,, (4:l) lipid monolayers produced unique structures for each isomer. The Cl cornpiexes were somewhat circular. the C8 complexes were smaller and prolate, and the MC8 complexes were the most extended. and were triangularo with apparent three-fold symmetry. The 3D reconstructions of the three MBP charge isomer-lipid complexes exhibited major strucniral differences and similarities. .411 three reconstructions shared three basic components: the cap, annular ring. and base. The

Cl reconstruction contained al1 three components, whereas the C8 reconstruction had a defuiite base, but the presence of the annuiar ring, and cap were stain dependent. The

Marburg CS (MC8) reconstructions had al1 three features, with a well pronounced pore 28 passing through the base region of the reconstruction. If one assumes C 1 is the "normal" configuration of this structure the complex contracts or relaxes, depending on which charge isomer makes up the complex. C8 was compacted. and MC8 was more extended.

Based on this study the composition of the complex is unknown; however. two models are proposed: (i) the complex is composed of a single MBP molecule with a major lipid component, and (ii) the complex is a hexamer of MBP, with a rninor lipid component.

Based on the potential C3 syrnrnetry of the MC8 data model (ii) is favored.

Although an exact implication on myelin stability is not yet warranted based pureiy on these results, meaningfil conclusions can be drawn from this investigation. Previous research has indicated that C8 has a loss of six positive charges, and MC8 has a loss of

18 positive charges (Wood et al., 1996). compared to Cl. The results of the present study indicate that the charge isomen of MBP Cl. Cg, and MC8 display both structural and dynamic association differences with the lipid monolayer of PS/GM, (4: 1). In MS. the loss in cationicity by MBP is believed to be attributable in the destabilization process in rnyelin (Moscarello et al., 1986). One cannot directly imply that the structurakharge differences in MBP are the cause of MS; perhaps it is a secondary response to the actual disease. Since the exact nature of the development of MS is unknown at present. the curent study provides insightful information on the putative nature of the disturbance in myelin at the molecular level. CHAPTER 2

TRANSMISSION ELECTRON MlCROSCOPIC IMAGENG, AND IMAGE

ANALYSIS, OF BOVINE Cl CHARGE ISOMER OF MYELIN BASIC

PROTEIN

2.1: Summary

MBP plays an integral role in the structure and function of the myelin sheath. In humans and cattle, an 18.5 kDa isoform of MBP predominates. and exists as a multitude of charge isomers resulting from extensive and varied post-translational modifications.

The Ieast modified isomer (named Cl) of the 18.5 kDa isoform of bovine MBP (bMBP) fkom fresh brain has been purified. Transmission electron microscopy was used to image negatively-stained single particles of Cl-lipid complexes on a lipid monolayer. Under these conditions, the Cl-lipid complex presented diverse projections whose relative orientations were determined using an iterative quaternion-assisted angular reconstitution scheme. In different büffers, one with a low salt and the other with a high salt concentration, the conformation of the complex was slightly different. In low salt buffer. the three-dimensional reconstruction, solved to a resolution of 4 nm, had an overall "Cu shape of outer radius 5.5 nm. The three-dimensional reconstruction of the complex in high sait buffet, solved to a resolution of 3.4 nm, was essentially the same in terms of overall dimensions but had a somewhat more compact spherical architecture. 30

2.2: Lipid moaolayers and bovine myelin basic protein

Myelin basic protein was the first agent in brain or spinal cord homogenates discovered to be responsible for experimental allergic encephalomyelitis (EAE). which is considered to be an animal mode1 for the human disease multiple sclerosis (Carnegie.

1971; Eylar et al., 1971; Kies, 1965). The 18.5 kDa isoform of MBP is the most common in mamals, including humans and cattle. The amino acid sequences of the

18.5 kDa isoforms of human and bovine MBP were reported independently by Carnegie et al. (1 97 1) and by Eylar et al. (197 l), respectively. Extensive post-translational modifications of MBP (e.g., GTP-binding, ADP-ribosylation, deimination of arginyl residues to citnitlinyl residues) create a high degree of microheterogeneity (Chou et al..

1976; Chan et al., 1987; Wood and Moscarello, 1989, 1996: MoscarelIo et al., 1992:

Moscarello, 1996). The resdtant charge isomers of MBP can be separated on a cation exchange column at high pH (Chou et al., 1976). Component Cl is the least modified and most basic component, and C8 is the most modified. MBP isolated fi-om victims of multiple sclerosis has been shown to be considerably less cationic than that from normal. age-matched controls due to a decrease in the most cationic forrn (Cl) and a consequent relative increase in the less cationic isomers, particularly C8 (Moscarello et al.. 1994;

Wood and Moscarello, 1996; Wood et al., 1996). This chapter focuses on the C 1 charge isomer of the 18.5 kDa form of bovine MBP when it interacts with a lipid monolayer.

Myelin has been generally considered to be an inert structure, facilitating saltatory conduction of nerve impulses. Thus, the role of MBP has long been assigned to be simply the compaction at the cytoplasmic surface of oligendrocytes or Schwann cells. 3 1 which eventually becorne the major dense line (Napolitano et al., 1967: Brunner et al..

1989; Kirschner et al., 1989; Deber and Reynolds, 1991; Mateu et al., 1996; Staugaitis.

1996). However, with the identification of cimilline-containing MBP (Wood and

Moscarello, 1989), its implication in demyelination in multiple sclerosis (Moscarello et al., 1994; Wood et al., 1996; Wood and Moscarello, 1W6), and its localisation to the intraperiod line of myelin (McLaurin et al., 1992, 1993), it appeared that a division of labour was operative for the plethora of post-translationally modified MBP molecules

(Wood and Moscarello, 1996; Moscarello, 1996). Presently, MBP is generally believed to be the agent responsible for the autoimmune response which precipitates the active degradation of the rnyelin sheath in multiple sclerosis (e.g., Vanguri et al.. 1993:

Mastronardi et al.. 1993).

Knowledge of the tertiary structure of MBP, and its association with lipids and other proteins within the compacted rnyelin multilayers, would facilitate a greater understanding of how it carries out its diverse functions. It is known by spectroscopie studies that MBP probably has a disordered conformation in its isolated form in aqueous solution (Epand et al., 1974; Krigbaum et al., 1975; Chapman and Moore, 1976: Gow and

Smith, 1989, 1992; Stuart, 1996) and that its secondary structure changes der phosphorylation or in the presence of lipids or detergents (Anthony and Moscarello, 197 1;

Smith, 1977b, 1992; Keniry and Smith, 1979, 198 1; Mendz and Moore, 1982; Ramwani et al., 1989; Deibler et al., 1990). The protein interacts mongly with itself and with other proteins (Golds and Braun, 1978; Smith, 1980, 1982% 1982b, 1992; Ashfar-Rad et al..

1987, Bellini et al., 1987, Moskaitis et al. 1987; Edwards et al., 1989), and sequesters 32 zinc (Cavatorta el al., 1994), a divalent cation essential for the stability of myelh (Inouye and Kirschner, 1984). The structures of srnall subsegments of MBP have been probed by nuclear magnetic resonance (Keniery and Smith, 1979; Fraser and Deber, 1985: Mendz and Moore, 1985; Mendz et al., 1995; Koshy et al., 1996; Price et al., 1988) or predicted by computational sequence analyses (Martenson, 198 1, 1986; Stoner. 1984. 1990:

Golubovich et al., 1989; Inouye and Kirshner, 1991; Perez-Gil and Keough. 1994).

However. ail attempts to obtain three-dimensional crystals of MBP suitable for X-ray difiactometry, and very many have been pursued over the past two decades. have failed

(Sedzik and Kirschner, 1992).

The preceding paragraphs of this section indicated the ability of MBP to associate with itself, and with lipid monolayers. Under these conditions purified MBP (random coi1 in aqueous solution) is believed to adopt a structure sirnilar to that in vivo in the rnyelin sheath. This thesis is an investigation of MBP-lipid interactions. Several lipids were considered for this study, but two lipids produced the best results. The first lipid. phosphatidylserine (PS), was chosen phmily due to its negative charge on the polar head group. Since MBP is a basic protein, it carries a net positive charge, and thus it should interact eiectrostatically with PS. The second lipid that was used, monosialoganglioside

GM, was chosen for two reasons. Firstly, several studies have been conducted on

MBP/GM, interactions (Sela and Bach, 1984; Ong and Yu, 1984; Bach and Sela, 1985;

Chan et al., 1990; Smith, 1992). These studies have shown that MBP and GMlinteract both hydrophobicaily and electrostatically with each other, thus making CiMl an excellent candidate for this study. The second reason for choosing GM,as a lipid cm be found in 33 the successful crystallization of cholera toxin on lipid films (Reed et al.. 1987: Ribi et al..

1988; Mosser and Brisson, 1991). In these studies 2D crystals of cholera toxin were successfully produced on a lipid monolayer that contained GM,. In a study by Caamaiio and Zand (1989) it was demonstrated the cholera toxin A and B subunits had homologous sequences to MBP which were involved in GM,binding. Based on this information both

PS and GM,were prime candidates for the lipids that were tested for MBP interaction. and subsequently these two lipids showed the greatest degree of single particle production upon interaction with MBP.

The primary focus of this study is the determination of the structural differences by electron microscopy and cornputer image analysis (Fig. 2-1) between the different charge isomers of human MBP in association with lipid monolayers. The present chapter serves as a feasibility study on the potential application of the lipid monolayer technique to MBP. A bovine source was chosen for this introductory study to perform the numerous trials necessary to optimise the lipid-MBP interactions.

The primary function of a lipid monolayer in EM imaging is to provide a surface to which the macromolecule being studied can adsorb. In the case of MBP the interaction with lipid is believed to maintain its native structure (Anthony and Moscarello. 1971:

Keniry and Smith, 1979, 1981; Lees and Brostoff, 1984; Mendz et al.. 1988. 1991:

Moore, 1982; Ramwani et al., 1989, Smith, 1977b; Surewicz et al., 1987). Once adsorbed to a lipid monolayer the molecule can have diffenng degrees of structural order on the lipid surface, the lowest order being randomly conf'ined to a plane, with increasing levels of order such as, preferentially ordered along a specific mis, Figure 2-1: Cornputer image analysis: an overview. This figure provides an overview of the computer image anaiysis process. First, digitized single particles of bovine Cl were interactively selected from a large single image (a). Once the particles are selected they are placed in a single computer file with thousands of individual particles (b). The individual images then undergo a pretreatment process, which includes: contrast reversal. variance normalization, and finally band pass filtering (c). Once al1 the particles are pretreated, they then undergo several rounds of reference alignment, multivariate statistical analysis (MSA), and hierarchical ascendant classification (HAC), to produce final image averages (d). At this point the 2D image analysis is completed, and the 3D image analysis begins for the angular reconstitution process (e-0. The first step in the processing involves the removal of dlbackground. leaving only the single particle (e(i)). and then for each particle a sinogram is cdculated (e (ii)). A sinogram is a stack of one dimensional projections (x axis) rotated 360° (y axis). The sinograms are then utilized in the angular reconstitution process to determine the relative angular orientation of two images to produce the final 3D reconstruction (0. Scale bar 10 nm.

36 dimers, a paracrystal, or a perfect 2D crystal (Jap et al., 1992).

When using the lipid monolayer technique to irnmobilize proteins a specalized well in a Teflon block is used (Fig. 2-2). Essentially one has a tiny well (10 pl volume) with the diameter of an EM grid. A protein suspension (10 p1) is first pipetted into the well. and then a dilute lipid suspension (1 pl) is spread on top of the protein solution. The well is then kept humid and the sarnple is ailowed to incubate for a set penod of time. after which the lipid and protein are withdrawn by an EM grid coated widi a holey plastic support film (Fig. 2-3).

In engineering the proper lipidprotein system several factors mut be considered.

Interactions at the atomic level are well understood (Israelachvili. 1985). whereas protein interactions at a macroscopic level, are less understood. Proteins interact strongly with their natural environment, which may be water or lipid, with the latter being the case with

MBP. Therefore, solvation forces must be considered (Jap er al., 1992). Detergents can also be utilized to assist in lipid solubilization to affect membrane fluidity. and assist in the formation of vesicles and micelles if they are desired. Detergent can also interact with protein molecules via association with hydrophobie regions of the molecule. Caution must be observed when using detergents since a harsh detergent may denanire a protein. or form a thick coat around the protein which may be deleterious when one wants to image the molecule.

In the present study detergents were not used, thus they were not an experimental parameter. Factors such as the influence of pH, salts, and lipid components were the primary conditions which were tested. Since MBP is a basic protein (positive charge). Figure 2-2: Schematic representation of lipid monolayer technique. This figure outlines the steps involved in the lipid monolayer technique. First the protein sample ispipetted into the well (a), then the lipid solution is added on top of the protein suspension (b). After a penod of incubation the protein molecules accumulate at the protein-lipid interface (c). Finally the protein sample is removed with an EM grid for observation the TEM (d). Figure 2-3: HoIey plastic support film. This figure shows a holey plastic (celIulose acetate butyrate) support film as viewed in a cornpound light microscope (a). and a TEM

(b). The holey plastic film has been placed on a TEM grid (100 X 400 Mesh), and one cm see the grid bars (black) in (a). The holes in the plastic (white) are where the MBP and lipid support film are placed. After the holey plastic has been placed on the TEM grid, it has carbon evaporated ont0 it to provide additional support. Scale Bars (a) 10

Pm,(b) 6 pm. 39 a lipid monolayer was produced with a net negative charge. so that there would be an electrostaîic attraction between the lipid and MBP. The lipids that produced the highest degree of MBP interaction (Fig. 2-4) were phosphatidylserine (negative charge). and monosialoganglioside GM,, which interacts with MBP (Chan er al.. 1990). Figure 2-4: Lipids used to produce the lipid monolayer for myeiin basic protein. The structure of phosphatidylserine (PS) (a), and monosialoganglioside GM 1 * (b) are shown here. PS and GM~were chosen since they both have a negative charge. which would attract the cationic MBP protein by electrostatic interactions. GMI was ais0 chosen based on a previous study by Chan et ai. (1990), which demonstrated that MBP and monosialoganglioside GM 1 interact.

* Monosinloganglioside GM~is an abbreviation for GnlP 1 +3GnlNAcB 1 +4Gd

(3+ 2aNerrAc)Bl + 4GlcpI+ Icernmide. 2.3: Materials and methods

2.3.1: Purification and characterisation of bovine myelin basic protein

Bovine brain was obtained from 2-4 year old cattle immediately der death

(Ontario Agricuitural College). Ail subsequent manipulations were performed at -1°C unless otherwise specified. The entire brain was immediately placed in approximately 1

1 of isotonic saiine (145 rnM NaCl). The meninges and cortex were carefully removed and blood washed away from the remaining tissue with isotonic saline. The gray matter was removed from the white matter by gently using a spatula (Fig. 2-5). This white

matter was cut into small pieces approximately 3-5 cm in dimension and either frozen rapidly and stored at -70°C until use, or used irnmediately as a source for bovine C 1.

The Cl component of myelin basic protein was isolated essentially as describcd

previously (Cheifetz and Moscarello, 1985; Wood and Moscarello, 1989). Briefly. a total of approximately 36 g of white matter was homogenised in 4-5 small portions in 284 ml chloroform:methanoI (2: 1 v:v) for 2 minutes using a probe sonicator (ProZjO. DiaMed

Corp., Mississauga, Ontario) at a setting of 4-5. This homogenate was stirred ovemight at 4"C, and then filtered through Whatman filter paper under gravity at room temperature

in the fume hood. The residue remaining on the paper was washed with 130 ml ice cold chioroform:methanol (2:l v:v) followed by 172 ml of ice cold acetone. The residue was then air dried and either frozen for up to 2 days at -20°C before use. or irnmediately

resuspended in 130 ml of 100 mM H2S0,, 0.1 mM PMSF (phenylmethylsulfonylfluoride).

The homogenate was gently stirred overnight and centrifùged in a Beckman JA20 rotor at 6,000 x g for 1 h. The pellet was resuspended in 52 ml of the same solution and Figure 2-5: Photographs of bovine brain. A bovine brain is shown here. with three figures which follow the processes involved in dissecting the brain for purification of myelin basic protein. In (a) the brain is shown with the meninges still intact. the arrow shows an area of the brain where the meninges are evident. Figure (b) shows a brain with the meninges removed, and (c) is a crossection of the brain. The white amow in (c) identifies the white matter, and the grey mow the grey matter. In the myelin basicprotein isolation process the grey matter is stripped off, and the white matter is kept for homogenization and further purification. 43

recentrifuged at 6,000 x g for 30 min. The acid-soluble proteins in the pooled

supematants were precipitated by adding an equal volume of ice-cold absolute ethanol and

leaving the mixture overnight at -20°C.

The precipitated acid-soluble proteins were recovered by centrifugation at 6.000

x g for 1 h at -10°C. The precipitate was washed twice in 150 ml of 90% ethanol at a

temperature of -20°C, and recovered by centrifugation at 6.000 x g for 30 min at -10°C.

The final protein pellet was resuspended in 10 ml of buf5er- 1 (80 mM glycine. 6 M urea.

pH 9.5). and dialysed ovemight against 500 ml of buffer-1 at room temperature.

Insoluble material was removed by centrifugation at 6000 x g for 15 min at 4°C. The pH

of the supernatant was adjusted to 9.5 with NaOH.

An amount of 80-90 (Azg0)units of sarnple was loaded at a rate of 0.1 mVrnin onto

a Whatman CM52 column (35 cm x 1 cm) which had been pre-swollen and pre-

equilibrated with buffer-1 at pH 10.5. Chromatography was performed at room

temperature (-18-20°C). Afier sample loading, the colurnn was washed with 60 ml of

buffer-1 at 12 ml/h for 5 h. Bound protein was eluted by 200 ml of buffer-2 (80 mM

glycine, 2M urea, pH 10.5) in a linear gradient of 0-200 rnM NaCl at 12 mlh. The eluate

was monitored at 280 nm and collected in 3 ml fractions. Under these conditions.

component Cl of MBP eluted last. The Cl-containing fractions were dialysed against

either 1 1 of buffer-G (80 rnM glycine, pH 7.5 with NaOH, 0.75 mM NaN,) or buffer-S

(10 mM HEPES, pH 7.5 with NaOH, 150 rnM NaCl, 10 mM CaClz, 0.73 rnM Na,) at

4°C for 24 h, and then dialysed further against 1 1 of the sarne buffer for another 24 h.

The protein (MBP) concentration was estimated via a commercial colorimetric assay using 44 bicinchoninic acid (Pierce Corp., Rockford, Illinois), and bovine senim albumin as a standard, and diluted to a final concentration of 2.5 or 0.25 mg/ml. Aliquots of bovine

Cl were either used immediately or frozen at -20°C until use.

2.3.2: Analysis of bovine myelin basic protein by polyacrylamide gel electrophoresis

The identity of purified bMPB/CI was confirmed by discontinuous SDS-PAGE

(12% polyacrylamide gel electrophoresis) (Laemrnli, 1970). In brief, protein samples

fiom various steps of the purification were mixed with an equal volume of SDS loading

buffer (0.0625 M Tris-HCl, pH 6.8, 2% SDS, 5% P-mercaptoethanol. 87% glycerol) and

boiled for 5 minutes. The gel was then run at 88 V for 30 minutes. and the voltage was

increased to 110 V, and the gel was run for a Mer 30 minutes. Following electrophoresis the gels were either silver stained, or prepared for Western blot analysis.

as described below.

2.3.3: Silver staining polyacrylamide gels

Following SDS-PAGE electrophoresis, gels were prepared for silver staining with

a 2 hour fix in 50% (v/v) methanol. For this and al1 subsequent steps the gels were

gently shaken on an orbital shaker, and the solution was drained prior to the addition of the next solution. The gels were subsequently rinsed four times with ddH,O. for a period

of 10 minutes each. The rinses were followed by a final 2 hour fix in 50% methanol.

The gels were then stained for 15 minutes in silver stain [(21 ml of 0.36 % NaOH), 2 ml

NH,OH, (0.8 g silver nitrate in 4 ml dW,O). and 73 ml ddH,O], and the staining was 45 followed by 2 washes with ddH,O for a period of 2 minutes each. The silver stained gels were then submerged in developer [(2.5 ml of 1.0 % citric acid), (0.25 ml of 37% formaidehyde), and 500 ml ddH,O] . The gels were allowed to develop until the bands in the gel becarne visible, and the reaction was stopped by draining off the developer. and adding 50% (v/v) methanol.

2.3.4: Western blot anaiysis of bovine myelin basic protein

Western blotting was performed on the gels using a commercial kit (amplified aikaline phosphatase immun-blot assay kit, goat anti-rabbit IgG (H+L)- biotin. Bio-Rad

Laboratories (Canada) Ltd., Mississauga, Ontario, # 170-64 12). In al1 subsequent steps the gels were gently shaken on an orbital shaker unless otheMrise stated, and the solution was drained prior to the addition of the next solution. The gel was initially equilibrated in transfer baer (10 rnM CAPS, pH 11.0 with NaOH, 20% methanol) for 10 min. and electroblotted at 170 mA for 3 h in a BioRad Mini-Protean II blotting apparatus ont0 a nitrocellulose membrane. The gel was then washed for 10 minutes in TBS (20 rnM Tris-

HCl, pH 7.5,500 mM NaCl), and the unbound sites in the nitrocellulose membrane were biocked using 3% gelatin in TBS. The membrane was subsequently washed 3 times with

TBS, for a period of 10 minutes each. The membrane was then probed with rabbit anti- bovine Cl polyclonal IgG antibody (0.03 mg/ml),(a generous gift fiom Dr. Mario

Moscarello, Hospital for Sick Children, Toronto) in TBS for 3 hours. The membrane was then washed 3 times with TBS, for a penod of 10 minutes each. The membrane was then probed with biotinylated goat anti-rabbit antibody (1:3,000) for 1 hou. While the blot 46 was being probed the streptavidin-biotinylated aüsaline phosphatase complex was formed by adding equal amounts of streptavidin to alkaline phosphatase (13.000 dilution for each) in TBS. mer the second antibody incubation, the membrane was washed twicr. for a penod of 10 minutes each. The membrane was then probed with the streptavidin- biotinylated alkaline phosphatase complex for 1 hour. This step was then followed by 4 washes with TBS, for a penod of 10 minutes each. Fresh developer was made by adding

0.2 ml of reagent A, and 0.2 ml of reagent B, to 20 ml of colour development buffer.

The membrane was then submerged in the developer (for approximately 5 minutes) and allowed to sit in the dark without agitation until the bands developed on the membrane.

The reaction was then stopped by removing the developer, and submerging the membrane in ddH,O.

2.3.5: Preparation of plastic carbon coated grids

Copper and nickel 400 mesh grids were coated with plastic support films to provide mechanical support for observation of MBP in the TEM (Fig. 2-3). Copper grids were used for standard EM observation and subsequent cornputer analysis. These grids were coated with holey plastic, so that the micrographs could be taken on the specimens over the holes, thus minimising the background signal. Nickel grids were used for imrnuno-gold electron microscopy. Due to the number of washes and the fragile nature of lipid monolayers, solid plastic films were used to provide additional mechanical support. Preparation of solid and holey plastic support films was essentially the same. the only difference was the omission of one step. 47

To prepare the plastic support film, a cleaned glass microscope slide was coated on one side with hurnan nose grease and immersed in 0.2% celIulose acetate butyrate

(CAB) in ethyl acetate for 5 seconds. To prepare holey plastic, one breathes on the slide several times, once it is removed fiom the 0.2% CAB. (If one is producing solid plastic this step is omitted.) The slide was allowed to dry in a humid chamber for 10 minutes.

Once dry, the slide was placed in wash solution (arnylacetate:methanoI:ddHO, 1 :150: 10) for 1 minute and allowed to air dry. The edges of the slide were scored with a razor blade, and the plastic film was floated off ont0 a surface of ddHzO. TEM gnds were then placed on the plastic film, and removed from the ddH20 with a piece of hard filecard paper. The grids and film were allowed to air dry. The gnds were stabilized with a layer of carbon film by direct evaporation in an Edwards coating system (E306A) until they developed a light grey colour, as rnonitored simply by eye. Individual grids were examined in a phase contrast light microscope, in order to appraise the quality of the support films. The best grids were then used for TEM.

2.3.6: Lipid monolayer technique for preparation of myelin basic protein for EM imaging

A 13 pl drop of Cl was incubated in a Teflon weii (Fig. 2-6) and a 4 pl droplet of dehydroabietylarnine (DHAA),(ICN), (1 mg/mI in hexane) (Aoyarna er al.. 1995) or iipid mixture (Avanti Polar Lipids, Birmingham Al., USA),(O. 1-0.4 mglml in chiorofom- hexane (1 :1, v:v)),(Jap et al., 1992; Scheybani et al., 1994) was touched to the surface where it formed a monolayer. Afier some time in a humid environment, the lipid-protein Figure 2-6: Protein incubation container The protein incubation container is shown here from the top (a), from an angle (b), and as a schematic from the side (c). The protein samples were incubated at 32OC, and the wells were maintained at constant humidity due to the exterior water trough. Each well has a diameter of 3 mm, and a height of 1 mm. this allows one to add 10 pl of protein sample and 1 pl of lipid on top as is shown in the closeup of a single well (d). 49 layer on the sudace was removed with a copper grid coated with holey plastic (cellulose acetate butyrate), negatively stained with unbuffered (pH -4) uranyl formate (Bremer ef al., 1992), and air-dried. A number of factors were varied in this process: protein concentration (2.5 - 0.25 mg/ml), substrate at air-water interface (DHAA. or mixtures of the following lipids: (Avanti Polar Lipids, Birmingham AL.. USA). phosphatidylethanolarnine (>99% purity), phosphatidylcholine (>99% purity, from egg), phosphatidylserine (>99% purity, from brain), and (Sigma, St. Louis. USA) monosialoganglioside G,, (95% purity, fiom bovine brain)! incubation temperature (20°C or 32OC), incubation time (0.5, 1, 2, 6, 8, and 12 h), pH of the protein solution (10.5 or

73, buffer conditions (concentrations of HEPES-NaOH, glycine, NaCI. CaCl?), and subphase density (with or without 20% glycerol). Cl at 0.25 mg/ml at 32°C interacted best (based on the Frequency of single particles) with a monolayer formed by a 4: 1 (w:w) mixture of phosphatidylserine (0.4 mg/ml): monosialoganglioside GMl(0.1 mglrnl) (Fig.

2-4), at 3Z0C, suspended in buffer-S (1 50 mM NaCI, 10 mM CaCl,), and is the basis of the results presented here.

2.3.7: Electron microscopy of bovine myelin basic protein

Electron imaging was perfonned in a Philips EM4OOT at a nominal magnification of 60,000 X using a low-dose unit, and with the sample kept at liquid nitrogen temperature (-185OC) in a Gatan cryo-holder. Al1 images were recorded on Kodak electron microscope film (ESTAR thick Base 4489), with an exposure time of 1 second.

In order to maximise the signal-to-noise ratio, images were only taken in previously 50

unirradiated areas of the grid over a hole in the plastic where the MBP molecules were

being supported by a lipid monolayer.

2.3.8: Immuno-gold electron microscopy of bovine myelin basic protein

Imrnuno-gold electron microscopy of MBP was performed on MBP-lipid monolayers using nickel grids coated with carbon on a solid CAB film. MBP suspended

in buffer-S was fira incubated in a Teflon trough with a phosphatidylserine : monosialoganglioside GM, lipid monolayer for 0.5 hou at 32OC. as was outlined above.

A coated nickel gnd was then used to lifi off the lipid monolayer, and the grid was ailowed to dry for 2 minutes. The grid was then floated on a drop of 3X PBS-BSA ( 10 mM NaHPO, pH 7.2, 0.45 M NaCl. 0.02 % Na&, 2% bovine serum alburnin) for 30 minutes. In this step and ail subsequent steps, the drops (50 pl) on which the grids were placed are sitting on top of Parafilm and incubated at room temperature. The grid was then placed on a drop of 1" antibody (0.03 mg/ml of anti-bovine MBP polyclonal IgG in

3X PBS-BSA; kindly provided by Dr. Moscarello. Hospital for sick children. Toronto) for 1.5 hours. The grid was then washed with 7 separate drops of 3X PBS-BSA: for one minute each. The grid was then placed on a drop of protein A gold (ASz0= 1. in 3X

PBS-BSA; Sigma Chernical Company, St. Louis) for 20 minutes, and then washed by placing it on 4 separate drops of ddH20, for one minute each. and finally stained with 2% uranyl acetate for 30 seconds.

In order to test for specific binding of the antibody to MBP, four different variants of the above experiment were conducted. These were: probing with primary anti-MBP 5 1

IgG plus protein A gold, probing with only protein A gold, probing with non-immune IgG plus protein A gold, and probing wiîh a monoclonal antibody (IgM) that was not specific for MBP plus protein A gold. The above four variants were also applied to a lipid monolayer that was incubated with buffer-S, but no MBP.

Since MBP spreads randomly on lipid monolayers. the number of gold particles in a 0.5 pm X 0.5 pm area were counted on several micrographs in order to obtain a measureable quantity of gold binding. The mean number of gold particles per unit area were then calculated, and the Students t-Test was used to determine if the gold particle counts per unit area were significantly different. The Students t-Test was calculated for independent samples with unequal variances. The variance, the 't' value. and the effective degrees of freedom (df) were calculated using equations 3.1, 3 2, and 3.3, respectively.

Equation 3.1

Equation 3.2

Equation 3.3 (s:l", +&q2 Effectivedf = Ks:lnJ21(n, - 111 +[(s21n~1~l(n~-111

The various calcdated Students 't' values were then tested for statistical significance (Steel and Tome, 1980). In order to expedite the calculation of the 't' values the Quatro-Pro for Windows (Version 5) spread sheet program (Borland International, Inc.) which is equipped with a statistical package was used.

2.3.9: Single particle countdunit area for bovine myelin basic protein.

In order to quanti@ the effect of bMBP on lipid monolayers of phosphatidylserine

: monosialoganglioside GM,,the nurnber of single particles were counted in a 0.3 pm X

0.3 pm unit area for samples containing lipid and bufTer-S only vs. samples containing lipid, buffer-S, and MBP. In both experiments the samples were incubated in Teflon troughs at 32°C as outlined above.

The mean nurnber of particles per unit area were then calculated. and the Students t-Test was used to determine if the particle count per unit area was significantly different in the two samples. The Students t-Test was calculated for independent samples with unequai variances, as previously outlined (Eqs. 3.1, 3.2, 3.3).

2.3.10: Two-dimensional image analysis of bovine myelîn basic protein

Image analysis by "single particle electron crystallography" (Yada et al.. 1995:

Frank, 1996) was performed using the IMAGIC-V program (van Heel et al.. 1996), with additional angular reconstitution (van Heel, 1987b) programs using a quaternion-assisted approach (Fmow and Otîensmeyer, 1992, 1993). Images were digitised at a resolution of 0.31 nm per object level using an 1s-1000 Gel Scanner (Canberra-Packard.

Mississauga, Ontario) on a M3B dissecting microscope (Wild Leitz), and an EMPIX

(Rockwood, Ontario) uniform intensity light box. Totals of 228 and 3 1 17 (preparations in buffer-G and buffer-S, respectively) smailer images of size 64x64 pixels ("picture 53 elernents") and containing a single particle within them were extracted from the larger images. Two-dimensional single particle electron image analysis was performed using standard approaches (Fr* 1989; van Heel et al.. 1996). using the IMAGIC V cornputer program (van Heel et ai.. 1996). Successive cycles of alignment with respect to a varies of references (Harauz et al., 1988), correspondence anal ysis and hierarchical ascendant classification (van Heel, 1989) were used to define homogeneous subsets for two- dimensional averaging. Both populations were analysed independently. In the case of the smaller population from buffer-G, the class averages comprised only small numbers of members and it was decided to use the individual single particle images as projections for three-dimensional reconstniction (Harauz and van Heel, 1986). The larger population from buffer-S gave statistically significant classification results. and the class averages were deemed suitable to serve as projections.

2.3.11: Three-dimensional reconstniction of bovine myelin basic protein

In preparation for three-dimensional reconstniction, projection image pretreatment comprised mainly background removal (Beniac et al., 1997e). Determination of relative

Euler angles of projections within each set was performed using quaternion-based angular reconstitution algorithm (Farrow and Ottensmeyer, 1992, 1993). Three-dimensional reconstructions using weighted back-projection with exact filters were then performed

(Harauz and van Heel, 1986), and were manipulated (thresholded and rotated) using

INSIGHT II (BioSym Inc., Parsippany, NJ) molecular graphics software. To get two independent class averages or reconstructions, the projection sets were randomly halved. 54

The Fourier ring and shell correlation functions were used to estimate the reproduciblr spatial resolution between two independent two-dimensional averages or three-dimensionai reconstructions, respectively (van Heel, 1987b). 2.4: Resuits

2.4.1: Isolation and electron microscopy of bovine Cl

To obtain material for these experiments, charge component C 1 of the 1 8.5 kDa isoform of MBP from fresh bovine brain was purified under conditions that minimised proteolytic degradation (Fig. 2-7). A bovine source of the protein yields large and fresher quantities of material than human sources, i.e.. autopsies. The differences between MBP sequences from the two species are small (Carnegie, 1971; Eylar et al.. 197 1).

Preparations of bovine MBP for electron microscopy were undertaken primarily using a lipid monolayer approach (Jap et al., 1992; Scheybani et al., 1994). The best interaction of MBP with the interface substrate was achieved using a mixture of negatively-charged phosphatidylserine and monosialoganglioside GM,. B y "best". it is rneant that the MBP adhered to the lipid monolayer and large numbers of particles could be seen in the TEM (Figs. 2-8, 2-9). Under other incubation conditions. few single particles could be seen adsorbing to the lipid film. Occasionally. two-dimensional microcrystals could be found but these diffiacted to only one order and were not analysed

Mer(results not shown). There is no underlying carbon support film in the samples.

In these micrographs the protein complex is suspended in negative stain on lipid monolayers supported over holes in a holey plastic film. It was fond that cooling the sarnple to liquid nitrogen temperatures in the electron microscope helped preserve these fiagile lipid films during image recording.

In electron micrographs, bovine Cl single particles in low salt buffer-G appears to have a toroidal or "C" shape and a diameter of approximately 11 nrn (Fig. 2-8). In 1 Fractions collected 60 Figure 2-7: Protein purification and biochemical characterization of bovine myelin basic protein. (a), Silver stained polyacrylamide gel of protein from bovine brain. Lane

1. bovine brain homogenate; Lane 2 . mixture of al1 charge isomers of MBP (this sample was loaded ont0 the CM52 anion exchange column); Lane 3, Cl charge isomer: Lane 4.

C2 charge isomer; Lane 5, C3 charge isomer; Lane 6. C4 charge isomer. (b) Western immunoblot of the polyacrylamide gel panel in (a) using rabbit anti-bovine MBP IgG.

(c) CM52 anion exchange column elution profile, charge isomers CI, C2, C3. and CL1 have been identified. For al1 studies in this chapter, the fractions enriched in Cl were pooled and used. Figure 2-8: Electron rnicrographs of bovine Cl A field of view of bovine Cl is shown

here. The Cl was irnaged in a TEM, and negatively stained with 2% uranyl formate. A

single particle appears here as a bright structure surrounded by a dark halo. The Cl is shown

here adsorbed to a lipid monolayer of GM~and PS: the Cl was incubated in buffer G. In this preparation the single particles tend to be clustered together. Figure (b) is a close-up of the region outlined by a rectangle in figure (a); a single particle has been circled in (b). Scale bar 50 nrn. Figure 2-9: Electron micrographs of bovine Cl A field of view of bovine Cl is shown here. The C 1 was imaged in a TEM, and negatively stained with 2% uranyl formate. the single particles appear here as a bright structure surrounded by a dark halo. The CI is shown here adsorbed to a Iipid monolayer of GM~PS; the C 1 was incubated in buffer S.

Figure (b) is a close-up of the region outlined by a rectangle in figure (a): a single particle has been circled in (b). Scale bar 50 nm. 59 higher salt buffer-S, the protein complex has a more compact spheroidal appearance but is approximately the same size (Fig. 2-9). The homogeneity of appearance (size and degree of staining) of the preparations makes acceptable the application of both established and emerging techniques of single particle electron crystallography (Yada ri al., 1995; Frank, 1996). The first step in this process is two-dimensional alignment. multivariate statistical analysis, and hierarchical ascendant classification to define characteristic (recurring) projection views (Harauz et al., 1988; van Heel. 1989: Frank.

1989). The second step is determination of the relative orientations of different projections (van Heel, 1987b; Farrow and Ottensmeyer, 1992, 1993) and three- dimensional reconstruction (Harauz and van Heel, 1986) (Fig. 2- 1).

In order to Merquanti@ the interaction of MBP with the lipid monolayer. an expenment was performed in which two suspensions were incubated with the lipid monolayer of PSIG,, (4: l), the first suspension contained only buffer-S. and the second contained Cl suspended in buffer-S. The results of this experiment are shown in Fig. 7-

10. The number of single particles were counted in 0.3 pm X 0.3 pm areas for both preparations; the buffer-S population had 4.2 f 5.1 particles per unit area. and the Cl buffer-S preparation had 123.5 + 26 particles per unit area. The Students t-Test was used to compare the two sample means, and based on the t-Test the C 1 populations were found to have significantly more single particles. The next point to be addressed was the composition of the particles, specifically did they contain MBP? Immuno-gold electron rnicroscopy was used to determine this (Figs. 2- 1 1, and 2- 12). MBP-PS/G,, (4:1 ) and

PS/GM,(4: 1) lipid monolayers were probed with combinations of anti-bovine IgG, non- -- Buffer S bovine C 1 Sample Type

Figure 2-10: Particle counts for bovine Cl. MBP - lipid monolayrr interactions were quantified by counting the number of single particles in a unit area (0.3 pm X 0.3 Pm).

Two sarnple preparations were compared, one with buffer S and lipid, and one with bovine Cl suspended in buffer S and lipid. In both preparations the number of single particles were counted after a 0.5 hour incubation at 32' C. The buffer S preparation had

4.2 +/- 5.1 single particles per unit area. and the bovine C 1 preparation had 123.5 +/- 36 single particles per unit area. This data were then analyzed by the Students t-Test for two samples with unequal variances (Eqns. 2-1, 2-2, and 2-3). Based on these formulas the pooled standard deviation was caiculated to be 16.1, with a t value of 16.1 with 84 degrees of freedom. Since the critical t vaIue for 84 degrees of freedom is 1.66, one can say that

MBP has a significant effect on the lipid monolayen, with regards to single particle nucleation. Figure 2-1 1: Immuno-gold electron microscopy of bovine Cl on PS/GM~ lipid monolayen. Eight fields of view of PS/% 1 lipid monolayers that were probed with protein A gold (prA) are shown here. This figure can be broken up into two prirnary groups: the first set (a-d) consisted of buffer S incubated with the lipid monolayer. the second set (e-h) consisted of bovine Cl in buffer S incubated with the lipid monolayer.

Both groups were probed with four different antibody-gold combinations. These were as follows; (a,@ only prA, (b.0 non-immune IgG and prA. (c,g) non-specific monoclonal antibody and prA, and (d,h) antibovine IgG and pr A. In figs. (a-g) there was little if any gold label. whereas in fig. (h) there was considerable gold labelling. The black arrow points to a single gold particle in (h). Scale bar 100 nm.

1 2 3 3 1 231 Buffer S Bovine Cl Sample Type

Figure 2-12: immuno-gold counts for bovine Cl. A similar analysis to that camed out in Fig. 2-10. was applied to the immuno-gold data for bovine C 1 (Fig. 2- l 1 ). In this figure the mean number of gold particles per unit area (0.5 ym X 0.5 Pm)was plotted for eight different sample preparations. These data sets cmbe divided up into two primary groups: the first set consisted of buffer S incubated with a lipid monolayer. and the second set consisted of bovine C 1 suspended in buffer S with a lipid monolayer. Both sarnples were incubated at 32 OC. Both groups were probed with 4 different antibody-gold combinations

(labeled 1-4). These combinations were; ( 1 ) only protein A gold (prA), (2) non-immune

IgG and prA, (3) non-specific monoclonal antibody and prA, and (4) anti-bovine IgG and prA. The mean +/- standard deviation are plotred in this figure- and these values were then analyzed by the Students t-Test. The results of this analysis are presented in Table 2- 1. t Variable I Variabte 2 Y, + s, Y2 +- ~1 S 1-2 df ResuI t B, 1, A 8.1 + 3.1 O. 1 + 0.3 2.6 149 30.6 VI > V2

B, A 8.1 I3.1 0.0 2 0.0 2.6 143 31.1 VI > VZ

B, N, A 8.1 f 3.1 O. 1 + 0.3 2.6 150 30.4 VI > VZ

B, M, A 8.1 2 3.1 O. 1 + 0.2 2.6 145 30.9 VI > V2 Cl, 1, A Cl, A 8.1 + 3.1 1.0 I0.9 2.6 186 25.2 VI > V2 Cl, N, A 8.1 +- 3.1 0.8 f 1.1 2.6 197 25.1 VI > V2 Cl, M, A 8.1 + 3.1 0.4 I 0.6 2.6 166 28.3 VI > V2

Table 2-1: t-Test for irnmuno-gold counts of bovine Cl. The Students t-Test results for the immuno-gold data for bovine C 1 (Figs. 2- 1 1. and 2- 12) are presented here. In this analysis the t' values were calculated for two sarnples, assuming unequal variances. Based on the degrees of fieedom, a critical t value (0.05) of 1.7 was used. In al1 of the tests based on this criteria, the first variable (VI : bovine C 1 probed with primary antibody, and protein A gold) was significantly larger then the second variable (V2: seven controls). The terms that were used to detïne the variables are as follows: CL - Cl charge isomer of MBP, B - buffer S; 1 - prirnary anti-bovine

MBP IgG; A -protein A gold; N - non-immune IgG; M - monoclonal antibody that is not specific for MBP. 65 immune IgG? monoclonal antibody not specific for MBP, and protein A gold (prA). Fig.

2-1 1 shows sample fields of view for this experirnent, and Fig. 2- 12 shows a graph of the mean prA counts per unit area (0.5 pn X 0.5 pm). In Table 2-1 these data were analyzed by the Students t-Test where it was determined that there was significantly more prA labelling in the positive control (MBP-PSIG,, (4: 1) probed with anti-bovine IgG and prA) then in the seven negative controls. This demonstrated that MBP was integrated in the single particles that had been imaged, and there was an epitope(s) at the surface of the particles that was accessible to the polyclonal anti-bovine MBP IgG.

2.4.2: Two-dimensional analysis and three-dimensional reconstruction of bovine Cl

The two-dimensional single particle analysis of the srna11 population (228 particles) of bovine C 1 in buffer-G yielded class averages (of approximately 10 mernbers each) that mainly appeared to be subtle variations of the basic "CMshape (not shown). The 228 original images of individual single particles were considered to be projections of the sarne three-dimensional siructure at different orientations. Two-dimensional single particle analysis of the larger data set (3117 particles) of bovine Cl in buffer-S yielded convergence to a set of statistically significant class averages shown in Fig. 2- 13. This preparation on the whole was much better in tems of yield of particles and definition of their shape than the one in buffer-G. The class averages of Fig. 2-13 were also interpreted to represent projections of the same three-dimensional stmcture at different orientations. The class averages correspond well with their individuai class members.

The selected "Mnclasses (Fig. 2-14), and "C" classes (Fig. 2- 15) illustrate the high degree Figure 2-13: Class averages of bovine Cl produced by MSA. Cl was image(d by

TE1M, at liquid nitrogen temperature, and stained with 2% uranylformate, 3 1 17 part icles weI -e used in the cornputer analysis. Indicated here are two charactenstic vi ews.

inte :rpreted to be an "end-on C" (black circles) and a "side-on M" (white circles) orie ntation. Scale bar 10 nm. Figure 2-14: Selected "MWclasses of bovine Cl produced by MSA. Classes 37.42. and 5 1 (a, b, and c respectively) are shown here. These classes correspond to the class averages highlighted in Fig. 2-13. The above figure shows al1 of the individual class members, plus the total class average (circled). These three classes are referred to as the "side-on M" orientation. Scale bar 10 nm. Figure 2-15: Selected "CMclasses of bovine Cl produced by MSA. Classes 12. 29. and 89 (a, b, and c respectively) are shown here. These classes correspond to the class averages highlighted in Figure 2-13. The above figure shows al1 of the individual class members, plus the total class average (circled). These three classes are referred to as the

"end-on C" orientation. Scale bar 10 nm. 69 of similarity between the images. The resolution of the 2D images was calculated by the

Fourier ring correlation method (the canonical significance threshold of three standard deviations as suggested by van Heel (1987b) was used), and found to be 1 -8 nrn (Fig. 2-

16).

Both projection sets of 228 images and 100 class averages, were analysed by the iterative quaternion-assisted angular determination process (Farrow and Ottensmeyer.

1992, 1993). The distribution of Euler angles on the unit sphere for each set is shown in Fig. 2-17. The three-dimensional reconstructions of both projection sets. computed using filtered back-projection, are shown in stereo in Figs. 2- 18 and 2-19 from various perspectives. The dimensions are an outer radius of 5.5 nm, an inner radius of 3 nm. an overall circurnference of 15 nm, and a height of 4.7 m. For the first data set. almost al1 orientations depict the "Cmshape, indicating a faithful correspondence with the original projection data. The second three-dimensional reconstruction is not as open as the first one, but has roughly the sarne dimensions. Resolution estimates calculated by the Fourier shell correlation method for the 3D structure produced an estimated resolution of 3.4 nm and 3.2 nm for two independent three-dimensional reconstnictions of bMBP in buffer-G and S respectively (Fig. 2-20). -.. -.. sigma

Pixel

Figure 2-16: Resolution calculation of 2D averages of bovine Cl. An estimate of the spatial resoluticn of the aligned 2D image data from the high salt buffer was calculated by

Fourier ring correlation (FRC).The images were scanned at 0.3 1 nrnlpixel. Using a o value of 3, the resolution was estimated to be 1.8 nm. l . . -* ,... - - . -:,. -:. --.. . .

I s - -. .,:.; '.. - -

l .. . . '2.'. 5.:' -;' p increasing . . ., .-+ ' -. 90 1 * .* i -. m. i . * -- . 1

Figure 2-17: Angular orientations of bovine Cl 3D reconstructions, plotted cartographically as points on a hemispherical surface. The angular orientations of

228 single particles from (a) the low salt buffer (Figs. 2-8. 2-18). and (b) 100 class averages of 3 1 17 particles in the high salt buffer (Figs. 2-9. 2- 19) are shown here. The square points represent angular orientations in the foreground hemisphere. and the crosses show angular orientations in the background hemisphere. Under low sa1 t conditions (a) kt particles tend to cluster in one central region; this is indicative of their tendency to adsorb to the lipid in the "C" orientation (Fig. 2-8). The high salt preparation

(b) has less clustering, and the points seem to lie in a band in the upper region of the plot. Figure 2-18: Reconstruction of bovine Cl: low salt. Bovine Cl single particles that adsorbed to the lipid monolayer in the unsalted buffer system (Fig. 2-8) were used to produce this reconstruction. The reconstruction is illustnted here rotated at 90 O angles to the vertical axis of this page (a-d), the top and bottom views are shown in (e) and (f) respectively. The reconstruction is shown in stereo here as a shaded surface representation. Figure 2-19: Reconstruction of bovine Cl: high salt. Bovine CI class averages of the particles that adsorbed to the lipid monolayer in the calcium chloride and sodium chloride buffer system (Fig. 2-13) were used to produce this reconstruction. The reconstruction is illustrated here rotated at 90 O angles to the vertical axis of this page

(a-d), the top and bottom views are shown in (e) and (0 respectively. The reconstruction is shown here in stereo as a shaded surface representation. -.. -.. sigma

O 123 4 5 6 7 8 9101112131415 Pixel

FS C -.. -.. sigma -

O 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 Pixel Figure 2-20: Resolution calculation of 3D reconstructions of bovine Cl. An estirnate of the spatial resolution of the 3D reconstructions from the low sait (a) and high salt (b) buffer reconstructions were calulated by Fourier shell correlation (FSC). Using a o value of 3, the resolution was estimated to be 3.2 nm for the low salt, and 3.4 nm for the high salt reconstructions. 75

2.5: Discussion

High resolution TEM is a viable technique to determine structures of biological macromolecules and their complexes (Yada et al., 1995, Frank, 1996). especially when like MBP (Sqdzik and Kirschner, 1992) these do not form three-dimensional clstals of suitable size and order for X-ray diffractometry. To date there is only one published

TEM study of MBP, which suggested that MBP spread fiom solution and negatively stained appeared to be fibrous and of dimensions 15 nrn by 1.5 nm (Epand et al.. 1974).

The analysis in this work was qualitative and performed under greatly different conditions

(see below). A decade ago, it was reported that MBP polymerised on the surfaces of erythrocytes and that this effect could be visualised by scanning electron microscopy, but no micrographs were presented (Bellini et al., 1987). Since MBP is a peripheral membrane protein within the myelin sheath. it was reasonable to hypothesise that it would interact with and then be visualised on lipid monolayers formed at the air-water interface. and potentially even form ordered arrays (Jap et al., 1992; Scheybani er a/.. 1994).

Although large, two-dimensional crystals, were not produced, one could analyse using other approaches the images of large numben of bovine Cl particles adhering to the lipid monolayer. The best interaction of MBP with the interface substrate was achieved using a mixture of negatively-charged phosphatidylserine and monosialoganglioside GM,. This interaction was Mercharacterized by the statistical analysis of single particle counts

(Fig. 2- 1O), and irnmuno-gold electron microscopy (Figs. 2- 1 1 and 2- 12, Table 2- 1). both of which confïied a distinct positive correlation between the presence of single particles and MBP. 76 These results are consistent with previous biochemicai data in many respects. It

has long been known that MBP cm interact strongly with and aggregate acidic lipid

vesicles in vitro and that the lipid reforms into stable multilarnellar layers which resemble

in vivo myelin as determined by X-ray difiaction and NMR studies (Boggs et ai.. 1977:

Lampe and Nelsestuen, 1982; Chiefetz et al., 1985; MacNaughtan et al.. 1985; Monferran et al., 1986; Fraser et al., 1989; Maggio and Yu, 1989, 1992). The strongest interaction

is with lipids such as cerobroside sulphate or gangliosides (Demel et al. 1973; Fidelio et al., 1982; Bach and Sela, 1985; MacNaughtan et al., 1985; Monferran et ai.. 1986:

Maggio and Yu, 1989, 1992; Chan et al., 1990). nius, it is not surprising that positively- charged MBP interacted with negatively-charged lipid monolayers (Fig. 2-4) and could

be visualised by transmission electron microscopy.

MBP also has a marked tendency to self-associate, an effect suggested to be due

to hydrophobic interactions (Smith, 1980, 1982% 1982b, 1992; Ashfar-Rad et ai.. 1987:

Bellini et al., 1987; Moskaitis et al., 1987). Lt also interacts with other proteins such as

PLP (Golds and Braun, 1978; Edwards et al., l989), calmodulin (Chan et al.. 1990). actin

(Dobrowolski, et al., 1986), and tubulin (Pirollet et al., 1992), and sequesters zinc

(Cavatorta et al., 1994). Thus, it is not unexpected that MBP foms extensive clusters on

negatively-charged lipid monolayers, especially in the low salt buffer-G (Fig. 2-8).

Finally, aithough MBP in solution is probably fully disordered (Epand et al.. 1974:

Krigbaurn and Hsu, 1975; Chapman and Moore, 1976; Gow and Smith. 1989; Smith.

1992; Stuart, 1996), CD and other analyses show that the degree of secondary structure. primarily the amount of a-helix but also P-sheet, increases substantially afier 77 phosphorylation and in the presence of organic solvents. detergents. and lipids (Anthony and Moscarello, 1971 ; Smith. 1977% 1977b; Keniry and Smith, 1979, 198 1; Mendz and

Moore, 1982; Stuart. 1996). tt can be argued that the conformation of bovine Cl cornplex which is formed on lipid monolayers, that has been visualised. is a biologically more relevant form than isolated protein spread on a hydrophobic carbon film. for example (Epand et al., 1974).

2.5.1: Cornputer image analysis

The two-dimensional single particle analysis of the srna11 population (228 particles) of bovine C 1 in buffer-G yielded class averages (of approximately 10 members each) that mainly appeared to be subtle variations of the basic "C" shape (not shown). This result was interpreted to mean that the particles "wobbled" around this basic projection direction.

Because these class averages were not statistically significant, they were not considered

Mer. These individual images were then used as projections for three-dimensional reconstniction (Fig. 2-1 8). Two-dimensional single particle analysis of the larger data set

(31 17 particles) of bovine Cl in buffer-S yielded convergence to a set of statistically significant classes and class averages shown in Figs. 2- 13, 2-14, 2- 1 5 with an estimated resolution of 1.8 rn (Fig. 2-16). The 2D averages were used as two-dimensional projections for three-dimensional reconstruction (Fig. 2- 19), with an estimated resolution of 3.4 nm (Fig. 2-20). This preparation on the whole was much better in terms of yield of particles and definition of their shape than the one in buffer-G.

The conditions of EM preparation are favourable for generating rnany orientations 78 of MBP complexed with lipid: the protein has a large nurnber of positively-charged residues distributed throughout its structure, each of which can interact with the negatively-charged monosialoganglioside GM, or phosphatidylserine headgroups; and the lipid monolayer is mechanically more flexible than even a thin carbon film and could potentially wrap around and cradle each molecule. As well. the results of the two- dimensional electron image analyses indicate that the sets of individual images of MBP- lipid complexes should be interpreted as representing randomly oriented projections of the sarne three-dimensional s-tnicture. Other situations have been described where such isotropic orientations were achieved, especially using methylamine tungstate as a negative stain (Bremer et al., 1992; Stoops et al., 1992).

Both pretreated projection sets (228 images and 100 class averages. respectively). were subjected to an iterative quaternion-assisted angular determination process (Figs. 7-

18, 2-19) (van Heel, 1987b; Farrow and Ottensmeyer, 1992, 1993). This approach has recently been applied to the signal sequence binding protein (Czarnota et a[.. 1994). nucleosomes (Bazett-Jones et al.. 1996; Czarnota and Ottensmeyer. 1996), ribosomal subunits (Beniac et al., 1997b, 1997c, 1997e), and other protein complexes. These successfùl previous endeavours provide confidence in the further application of this algorithm to the randomly-oriented projections of the present data. The distribution of

Euler angles on the unit sphere for each set is shown in Fig. 2-17. As suggested from considerations of the class averages, the preparation in buffer-S is seen visually to have a more isotropic orientational distribution than the one from buffer-G.

The three-dimensional reconstructions of both preparations are shown separately Figure 2-21: Cornparisons of TEM reconstructions of bovine Cl. The bovine Cl low sait reconstruction (left), is compared to the high salt reconstruction (right). In this figure side (a,b), and top (c.d) view orientations are shown. as both full (a.c). and cut away (b,d) representations. When the reconstnictions are compared this way it is apparent that the low sait reconstruction is less compact then the high salt reconstruction.

The reconstructions are also similar in that they both are essentially spherical and hollow. 80 in Figs. 2- 18 and 2- 19, and together in Fig. 2-2 1. For the fira data set of C 1 in buffer-

G, almost al1 orientations depict the "Cmshape, indicating a faithful correspondence with the original projection data. The dimensions are an outer radius of 5.5 nm, an inner radius of 3 nm, an overall circderence of 15 nm, and a height of 4.7 nrn. Epand et al.

(1974) proposed that MBP was a fibrous protein of dimensions 15 nm by 1.5 nrn. but these workers had imaged a mixture of al1 charge isomers (C 1 to CS) in aqueous solution with no Iipids present. Thus, the conditions of preparation for EM here are significantly different from those of this earlier work. The second three-dimensional reconstruction of proteins in calcium-containing buffer-S (Fig. 2-19) is not as open as the first one, but has roughly the same exterior diameter (Fig. 2-21). The effect of monovalent and divalent cations possibly may have forced the structure to close up somewhat.

These reconstructions of bovine C 1 complexed with lipid have several implications for the packing of myelin. The thickness of the major dense line in the myelin sheath is roughly 1.7 nm, and that of the intraperiod line is 2.5 nrn. A lipid bilayer is at least about

4.5 rn thick. The charge isomer Cl is located primarily in the major dense line. while

MBPIC8 is located primarily in the intraperiod line (McLaurin et al., 1992. 1993).

Biologically relevant questions that arise are (i) how does the Cl isorner sit in the major dense line?; and (ii) how does a conversion to C8 allow a migration through the bilayer into the intraperiod line? The first question is perhaps the easier one, but the answer is still not clear. Clearly, any scenario represents a forced and tight fit with the volumes that were reconstmcted. However, myeiin may not necessarily form on pre-existing lipid bilayers being laid down initially. Myelin basic protein could be the starting foundation 8 1 around which lipid is laid, an idea given credence by the observation already made by

Napolitano et al. (1967) that myelin retains a lamellar structure in the absence of lipids.

It is essential to pursue higher resolution structural analyses of MBP under varied conditions to clariQ such issues. 82

2.6: Conclusions

Myelin basic protein is an unusuai protein in rnany respects. It is microheterogeneous in its multitude of isoforms and isomers. and is the only myelin- associated protein that is absolutely essential for the maintenance of the myelin sheath.

This chaptrr provided an examination of bovine MBP (1 8.5 kDa isoform) interacting with a lipid monolayer of PS/GM, (4:l) by transmission electron microscopy and three- dimensional reconstruction of single particles on lipid monolayers. The results of this analysis demonstrated that Cl had a strong interaction with the PSIG,, (4: 1) lipid monolayer, and single particles of MBP-lipid were reconstmcted to produce a hollow sphencal structure with a 10 nm diameter. Under low salt conditions the particle has a open C shape, and the structure closes up into a more spherical shape in the presence of salt. Based on the encouraging results of this study, the next logical step. and topic of tte next chapter, was the application of this system to different charge isomers human

MBP, found in normal and MS afflicted individuals. This will provide insightful information on the potential structural and dynamic behavior of MBP in MS. CHAPTER 3

ELECTRON MICROSCOPIC IMAGING, AND IMAGE ANALYSIS, OF

HUMAN MYELIN BASIC PROTEN CHARGE ISOMERS

3.1: Summary

Three distinct charge isomers (Cl, C8, MC8) of human ( 18.5 kDa) interacting with a lipid monolayer of phosphatidyiserine and monosialoganglioside G,,

(PS/GMl)(4: 1) were investigated in this chapter. Component C 1 is the least modified and most basic component. C8 is the most modified citrullinated component found in chronic

MS affiicted individuals. MC8 is an extremely modified citrullinated form of MBP found in an acute type of MS called Marburg type MS (Moscarello et al., 1994; Wood and

Moscarello, 1996; Wood et al., 1996). The biochemicai differences between these proteins, and their implications in MS were discussed in sections 1.3 and 1.4 of Chapter

1 of this tea. MBP was imaged at liquid nitrogen temperature, negatively stained with uranyl formate and methylarnine vanadate for visuaiization by TEM and STEM. respectively. As was the case in Chapter 2 of this thesis, the negativeiy stained particles were imaged on a lipid monolayer of PS/GMl(4: 1), under the high sait conditions needed to maximise particle adsorption.

When equirnolar amounts of the three charge isomers were incubated with the lipid monolayer they demonstrated significantly different single particle production. This single particle production was interpreted to be a measure of MBP lipid interaction. It 84 was found that the hMC8 isomer had the greatest ability to produce single particles followed by hC 1, and finally hC8. Irnrnuno-gold electron microscopy positively identified that MBP was a component of the single particle complexes. The differences in the degree of immuno-gold labelling for the three different charge isomers on the PS/G,,

(4: 1) lipid monoiayers paralleled the results of the single particle data (MC8 > C 1 > C8).

When the TEM and STEM 3D reconstructions of the same charge isomer complex are compared they have a similar shape; however, the reconstructions are extremely different with regards to intemal mass distribution. and surfaces of the reconstructions differ. The TEM reconstructions had smooth surfaces, whereas the STEM reconstructions had convoluted surfaces filled with outcroppings and indentations. Since in both cases the MBP-lipid complexes were imaged by negative stain, the majority of the image contrast was being generated fiom the metal stain and not from the particle itself. As such, these reconstructions provide strong structural evidence for the surface topology of particle where the stain is in intimate contact with it. Due to incomplete stain penetration. it is difficult to interpret the interior densities of the reconstructions. The methylamine vanadate (STEM) reconstructions had a solid interior, whereas the uranyl formate stained

(TEM) reconstructions were hollow inside; the results (TEM and STEM) were consistent for al1 three charge isomer reconstructions.

When comparing the hC 1, hC8 and hMC8 reconstructions to one another the three reconstructions are quite similar in that they share similar cap, annular ring. and base structural elements. If one considers the hC1 reconstruction to be the "nonnd" structure. the hC8 appears to be a compact contracted version of the structure, and the hMC8 is a 85 relaxed and extended version of the structure. The structural data hint at the apparent three-fold symmetry of the MBP reconstructions. Due to the size of the reconstructions

(8-10 nrn diameter) the complexes are too large to be a single monomer of MBP. There are two possible interpretations of the reconstructions. either the complex is cornposed of one MBP monomer with a large lipid component, or the complex is a oligomer of MBP with a small lipid component. The data favour the second interpretation, and the three- fold symmetry and size of the complex suggests that the complex is a hexarner.

The current data clearly illustrate that there are interactive and structural differences between the different charge isomers of MBP when they interact with a lipid monolayer of PS/G, , (4: 1). Biochemical data have implicated the different charge isomers of MBP to be involved in the development of MS. This study provides corroborative information that the MS and non-MS forms of MBP behave differently at the molecular level when they interact with lipids in a monolayer.

3.2: Electron microscopy and myelin basic protein

Section 1.5 of this thesis introduced electron microscopy as a method of protein structure determination. In diis chapter two distinct methods of electron microscopic examination were employed in the study of MBP. The first mode of electron microscopy that was employed was transmission electron microscopy (TEM), and the second mode of electron rnicroscopy was scanning transmission electron microscopy (STEM). Bright field TEM is the standard method of determination of macromolecular structure by EM

(Valpuesta et al., 1992; Dube et al., 1993; Boisset et al., 1993; Radermacher et ai.. 1994; 86

Frank et al., 1995a, b; Schatz et al.. 1995; Stark et al.. 1995; Serysheva et al.. 1995:

Snvastava et al.. 1995; San Martin et al., 1995: Harauz et al., 1996; Verschoor et al..

1996; Orlova et al., 1996; Kolodziej et al., 1996; San Martin et al., 1997; Stoops et al..

1997; Stark et al., 1997), and this was the method of TEM observation employed here.

In both types of microscopes (TEM and STEM) the sarnple is visualised in a high vacuum environment, since air would deflect the electrons, interfering with illumination and the image forming process; therefore the biological sample must be visualised without any aqueous water present.

In the TEM the electrons typically are emitted by a tungsten wire filament. or a lanthanum hexaboride (LaB,) cathode, or a field emission gun. The emitted electrons are collimated by a condenser lens and aperture before they interact with the specimen. Afier the electrons of the imaging bearn interact with the specimen, the objective lens focuses and magnifies the image of the specimen. The objective aperture removes high angle scatter. Final magnification is provided by intermediate lenses, and the projection lens system which projects the image on a phosphor-coated screen, film. or on a video carnera.

The STEM, altematively, has only two basic components. the illumination system. and the detection system. The illumination system consists of a field emission gun (which produces a sub-nanometer probe of electrons) and an objective lens fitted with an aperture. The exiting electrons (once they have interacted with the sampie) are then collected by a senes of detectors, including large and small angle annuiar dark field detectors, and an optional energy loss spectrometer. It should be noted that TEMs cm also be equipped with in-column and post-column spectrometers as well (Ottensmeyer and 87

Andrew, 1980; Ottensmeyer. 1984; Bauer, 1988). What is unique about the STEM is that it is capable of multi-signal imaging, at lower doses than in the TEM. The primary reason for the lower dose in the STEM is that the eficiency of detection is quite high. since nearly al1 of the electrons that have gone through the sample are collected by the detectors.

In the present study the small angle annular dark field collectors on the STEM were used to provide data with regards to the structure of MBP-lipid complexes. The negative stain methylarnine vanadate was used to produce high signal-to-noise ratio images of MBP showing fine detail (Hainfeld et al., 1994). Negative staining or freeze drying of proteins, nucleoproteins or nucleic acids, and subsequent imaging in the STEM can give high quality results with fine structurai detail (Mory et al.. 198 1 ; Trent et al..

1991; Czarnota et al., 1994).

The present EM study is the first of its kind for MBP associated with lipid. As such, little structurai information was available on the tertiary and quatemary structure of

MBP (Fraser and Deber, 1985; Golubovich, 1989; Inouye and Kirschner. 1991 :

Martenson, 198 1, 1986; Mendz et al., 1995; Mendz and Moore, 1985; Price et al.. 1983:

Stoner 1984, 1990), especiaily on the differences between the different charge isomers of

MBP. Due to the lack of any information it was determined that negative staining was the sample preparative method of choice for the structural investigation of the MBP charge isomers.

The technique of negative staining (Brenner and Home, 1959) has been widely employed in the structural investigation of macromolecules. Historically. the EM 88 structurai study of a macromolecule by negative staining has always been the essential

fvst step in the determination of the structure under investigation. The second phase of the investigation then follows in which the macromolecule is imaged unstained. and the structure is subsequently solved, and compared to the stained structure. Although the

unstained structure will be of higher resolution, the overall structure of the two 3D reconstructions tend to be similar (Radermacher et al., 198% 1992, 1994: Boisset et al..

1990, 1992; Hinshaw et al., 1992; Akey and Radermacher, 1993; Wagenknecht et al..

1989). The primary use for negative stain in the initial analysis of a macromolecule is because biological materials have a low atomic nurnber, and thus they tend to produce low

contrast images. Negative stains in general are composed of a 2% aqueous solution of a heavy metal salt, which produces high contrast images with a high signal-to-noise ratio.

The major drawback of a negatively stained sample is that the image produced represents the stain distribution around a macrornolecule, and not the mass distribution of the

macromolecule itself. Thus when the macromolecule is reconstmcted the exterior surface of the 3D volume is representative of the structure. However, the mas distribution within

the reconstruction, and its interpretation cm be problematic. This problem is mer exacerbated by incomplete staining, positive staining, and preferential staining of components, ie., nucleic acid is preferentially stained in a nucleoprotein complex (Zobel and Beer, 1961, 1965). Thus, specific components in the 3D volume can be over represented (i-e..preferential staining will lead to an increased mass in the 3D volume in the region of the higher stain distribution).

In this investigation hiro negative stains were used. For the TEM analysis the 89 common negative stain uranyl formate was used, and for the STEM analysis the recently developed and less cornmon stain meîhylamine vanadate was used (Hainfeld er al.. 1994).

Uranyl formate (U02(CH02)3(Leberman, 1965) was chosen as a negative stain since it possesses the highest density of the negative stains (Bremer et al.. 1992). Uranyl formate also generates fme granules on drying, and thus is exnemely well suited for the investigation of the structures of small protein molecules. Uranyl formate is preferable to uranyl acetate (the most cornmonly used negative stain), since it has a smaller grain size, and thus superior stain penetration. The major drawbacks of uranyl formate. are that it is only soluble in water at a pH range of 3.5-5.0, and der it is prepared it must be used immediately, since it begins to decompose within 2-3 minutes of preparation. Once it has decomposed it loses its superior ability to penetrate the biological sarnple being investigated. Furthemore, al1 uranyl stains are ionic, and thus exist as a multitude of charged species (positive and negative), which exhibit preferential staining. For example. in the investigation of protein structure uranyl ions preferentially stain the phosphate. carboxyl, and hydroxyl groups of the amino acid residues. As such. one must consider that the preferential staining will affect al1 the results of the structural analysis of the protein being studied. With regards to mass distribution dùs will be deletenous due to the non-unifonn staining which will provide an overestimation of mass in areas of preferential staining. The benefit of this type of preferential staining is that it will provide insight into the location of the preferentially stained charged groups when compared to a mildly ionic stain; methylamine vanadate is such a stain.

Methylamine vanadate (CH,NH2V0,), the second stain used in this study was 90 specifically developed for use in the STEM (Hainfeld et al.. 1994). Methylamine vanadate has several advantages as a stain; it can be used at a near physiological pH at

8, it is stable in the electron beam with little granularity at a dose of 104 e/nrn2. and it does not produce any positive staining like the uranyl stains tend to do. As such rnethylarnine vanadate is basically a non-ionic stain, and any computer analysis will produce results free of positive stain artifact. Methylamine vanadate has a density that approximates that of nucleic acid, and thus it is an ideal contrasting agent for the investigation of the structures of proteins which have a lower density.

When comparing reconstructions of biological sarnples fiom methylamine vanadate vs. uranyl formate, density is a crucial factor to consider. Most biological sarnples are composed of primarily carban, hydrogen, oxygen, and nitrogen which have atomic masses of 12.0 Da, 1.0 Da, 16.0 Da, and 14 Da respectively. Vanadinium (atomic number 23) has an atomic mass of 50.94 Da, whereas uranium (atomic number 92) has an atomic mass of 238.0 Da. The heavy metals in the stains have higher atomic numbers than the biological sarnple. and thus will produce a faithfid representation of the particle surface morphology when they stain a biological sample. However. based on atomic mass the vanadiniun images will tend to obscure the interior mass density of the biological sample less than uranium does. The hi& atomic mass of uranium. and its superior staining abilities are why it has served as an excellent TEM stain over many years. In the case of 3D reconstructions one wants the highest amount of biological structural information with the lowest amount of stain contribution in the signal necessary to produce a useful image for analysis. With the advent of the low dose STEM, with its 91 highly efficient electron detectors, it is now possible to use lower atomic number stains

(ie. vanadinium) to produce 3D reconstructions from negatively stained particles. The vanadate stained reconstructions will have a lower çtain contribution in the negatively stained particle being investigated.

In using two different stains, and electron imaging techniques. this investigation was able to address the issues of differentiai imaging and staining. At the same tirne the

3D structures of the three charge isomer-lipid complexes were determined under different conditions, thus providing more substantial stmctural data on MBP. This is the first time structurai EM has been conducted on the different post-translationally modified forms of

MBP when they interact with a lipid monolayer. 92

3.3: Materials and methods

Preface: The majority of the materiais and methods that were used in this chapter were described in Chapter 2 of this thesis. which focused on the development of the lipid rnonolayer system using a bovine source MBP. This chapter then applied these techniques to different charge isomers fiom a human source. To avoid repetition the subsequent sections refer to the corresponding section from Chapter 2 where the technique has already been descnbed. In any situations where new protocols were used. or the protocols differ, the according information has been provided.

3.3.1: Purification and characterisation of human myelin basic protein

Human brain preparation and purification was perfomed exactly as described in section 2.2 of Chapter 2 of this thesis. The source of the MBP was fiom human autopsy material. Al1 of the 18.5 kDa isoform of human MBP was provided purified fiom the lab of Dr. Mario Moscarello, Hospital for Sick Children, Toronto, Canada. Components C8 and MC8 were collected fiom the protein fraction that does not bind the CM52 cellulose. and were isolated by reverse phase HPLC. The unbound protein (2 mg/ml) was dissolved in 0.05% (v/v) trifluoroacetic acid, vortexed, and centrifuged at 12,000 rpm on a table top centrifuge for 5 minutes at room temperature. The supematant was removed and the pellet was discarded. At this point some of the supematant was saved for SDS-PAGE to test for the presence of the 18.5 kDa MBP. The supematant (4 mg) was then injected into a Phamacia (LKB) SuperPAC Pep-S 5 pm C2/C,, column, with a flow rate of 0.5 dmin. Once loaded a linear gradient starting at 0.05% hifluoroacetic acid and ending 93 at 40% (0.05% trifluoroacetic acid) and 60% acetonitrile was applied over a 60 minute period. During this period the fractions were collected with the being measured, and plotted on a chart recorder. The fractions were gathered, and those containing C8 or MC8 were collected.

3.3.2: Analysis of human myelin basic protein by polyacrylamide gel electrophoresis and silver staining polyacrylamide gels

Al1 electrophoretic preparative procedures were conducted as outlined in section

2.2, Chapter 2. The only difference was that KI,hC8, and hMC8 were analyzed.

3.3.3: Western blot analysis of human myelin basic protein

Western blotting was performed on the gels using a commercial kit (Amplified alkaline phosphatase immun-blot assay kit, goat anti-rabbit IgG (H+L)-biotin. Bio-Rad

Laboratories, Mississauga, Ontario, #170-6412), as descnbed in section 2.2, Chapter 2.

The only difference was that hC1, hC8, and hMC8 were analyzed.

3.3.4: Antibody assay for human myelin basic protein

The antibody assay was performed on a nitrocellulose membrane using a commercial kit (Amplified alkaline phosphatase immun-blot assay kit, goat anti-rabbit IgG

(H+L)-biotin, Bio-Rad Laboratones , Mississauga, Ontario, # 170-64 12). This section bnefly surnrnarises the steps involved in this process. In ail subsequent steps the blot \vas gently shaken on an orbital shaker unless otherwise stated, and each soiution was drained 94 pnor tu the addition of the next solution. Equimolar amounts (0.12 mg/ml) of hC1. hC8. and hMC8 were diluted by a factor of IOX, 50X. and lOOX in TBS (20 rnM Tris-HCl, pH 7.5, 500 mM NaCl). Two pl of each sample were then pipetted ont0 a nitrocellulose membrane that had been pre-wetted with TBS, and air dried for 2 minutes. The membrane had been divided up into a grid pattern so that 10 replicates of each dilution could be pipetted ont0 the membrane. Once al1 the samples had been pipetted ont0 the same membrane it was allowed to dry for 5 minutes. After drying, the unbound sites in the nitrocellulose membrane were blocked using 3% gelatin in TBS. The membrane was subsequently washed 3 times with TBS, for a period of 10 minutes each. The membrane was then pro bed with rabbit anti-bovine C 1 polyclonal IgG antibody (a generous gifi from

Dr. Mario Moscarello, Hospital for sick children, Toronto) in TBS for 3 hours. The membrane was subsequently washed 3 times with TBS, for a period of 10 minutes each.

The membrane was then probed with biotinylated goat anti-rabbit antibody (1 :3.000) for

1 hour. While the blot was being probed. the streptavidin-biotinylated alkaline phosphatase complex was formed by adding equal amounts of streptavidin to alkaline phosphatase (1:3,000 dilution for each) in TBS. Mer the second antibody incubation. the membrane was washed twice. for a period of 10 minutes each. The membrane was then probed with the streptavidin-biotinilated alkaline phosphatase complex for 1 hour.

This step was then followed by 4 washes with TBS, for a penod of 10 minutes each.

Fresh developer was made by adding 0.2 ml of Bio-Rad kit reagent A, and 0.2 ml of Bio-

Rad kit reagent B, to 20 ml of colour development buffer. The membrane was then submerged in the developer and allowed to sit in the dark without agitation until the dots 95 developed on the membrane (approximately 5 minutes). The reaction was then nopped by removing the developer, and submerging the membrane in ddH20.

Mer development the blot was digitaily scanned by an 1s-1000 densitometer

(Canberra-Packard, Mississauga, Canada). The 1s-1000 software was then used to measure the relative optical density of each spot. which corresponded to the intensity of antibody binding in that region. The data were then surnmed and the mean optical densities of each treatment were compared to each other by the Students t-Test for independent sarnples with independent variances. The Students t-Test was conducted exactly as outlined in section 2.2 of Chapter 2 of this text.

3.3.5: Preparation of plastic carbon coated crids, and Lipid monolayer technique for preparation of myelin basic protein for EM imaging

Al1 the TEM sample preparative techniques for the imaging of the three charge isorners of human MBP (Cl, C8, and MC8) were conducted as outlined previously in section 2.2 of Chapter 2 of this text. The only difference was that oniy the high sait buffer (buffer-S) method of sample adsorption to a lipid monolayer of phosphatidylsenne and monosialoganglioside GM,was used. In the STEM analysis, MBP was negatively stained with rnethylamine vanadate (Nanoprobes Inc. Stony Brook, N.Y.). In both cases

(TEM and STEM) the sample was air dried, just as was done in the TEM analysis of

MBP in Chapter 2. 96

3.3.6: Electron microscopy of human myelin basic protein

TEM imaging was performed in a Philips EM400T at a nominal magnification of

60,000 X using the low-dose unit, or a JEOL EM-100CX at a nominal magnification of

50,000 X. In both cases the sample was kept at liquid nitrogen temperature (- 185°C) in a Gatan cryo-holder in the Philips TEM, or an Oxford cryo-holder, in the EOL TEM.

Al1 images were taken at an accelerating voltage of 80 kV. and recorded on Kodak electron microscope film (ESTAR thick Base 4489), with an exposure time of 1 second.

In order to minimise radiation damage images were only taken in previously unirradiated areas of the grid over a hole in the plastic where the MBP particles were being supported by a lipid monolayer. The micrographs were converted to digital format using a

MicroMAX CCD scanner (Princeton Instruments Inc., Stittsville, Ontario) mounted on an

M3B dissecting microscope (Wild Leitz), which scanned the microscope film lying on an

EMPIX (Rockwood, Ontario) uniform intensity light box. The digital images for hC1 were scanned at a pixel size of 2.23 A per pixel at the object level, and the hC8 and hMC8 images were scanned at 2.76 A per pixel at the object level.

STEM imaging was conducted on a Vacuum Generators HB601UX STEM at a nominal magnification of 500,000 X. Al1 images were digitally acquired directly in the microscope, at a pixel size of 3.3 A per pixel. The sample was imaged at liquid nitrogen temperature, and al1 images were taken at an accelerating voltage of 100 kV, at a dose of

17 e/A2. For the structural analysis of MBP, the low angle annula dark field detector was used for image acquisition. As was the case in the TEM imaging, the signal-to-noise ratio in the images was maximised by taking images in previously unirradiated areas of 97 the grid over a hole in the plastic where the MBP particles were being supported by a lipid monolayer.

3.3.7: Immuno-gold electron microscopy of human myelin basic protein

Immuno-gold electron rnicroscopy, and the subsequent statistical analysis were performed on MBP, as was descnbed in section 2.2. Chapter 2. The only difference was

that hC 1, hC8, and hMC 8 were used.

3.3.8: Single particle countslunit area for human myelin basic protein

Single particle counts, and the subsequent statistical analysis were performed on

MBP, as was described in section 2.2, Chapter 2. The only differences were that hC 1.

hC8, and hMC8 were used. Secondly, equimolar amounts (0.1 2 mg/ml) of hC 1 . hC8. and

hMC8 in buffer-S were loaded into the Teflon troughs and incubated at 32°C for 0.5 h.

1.0 h, 2.0 h, and 6.0 h with the lipid monolayer of PS/GM, (4:l). The single particle

counts for each treatrnent in the time course were then counted and averaged. and plotted

graphically. The entire data sets for each treatment were also pooled, and andyzed by

the Students t-Test.

3.3.9: Two-dimensional image analysis of human myelin basic proteia

Image andysis by "single particle electron crystallography" (Yada et al.. 1995.

Frank, 1996) was perforrned using the IMAGIC V program (van Heel et al.. 1996)

exactly as described in section 2.2, Chapter 2. The only major differences were that the 98 numbers of particles that were analyzed were considerably larger, 33,660 single particles in totaI. There were six distinct data sets that were analyzed independently of each other.

The numbers of single particles in each data set were as follows. For the TEM data there were three populations: hC1 (6172 particles), hC8 (5017 particles), and hMC8 (5266 particles). For the STEM data there were also three populations: hC1 (5420 particles). hC8 (6346 particles), and hMC8 (5439 particles). The other major differences in these analyses were that after several successive rounds of MSA and HAC were conducted. three sets of 100 class averages each pooled and used as input for the subsequent 3D reconstruction process. The reason for this change in analysis was to minimise the potential reference bias affect from the 2D analysis on the 3D analysis.

3.3.10: Three-dimensional reconstruction of human myelin basic protein

Al1 3D angular reconstitution, back-projection, and visualization by INSIGHT II

(Biosym Inc., Parsippany, NJ) procedures for both the TEM and STEM were conducted as outlined in section 2.3, Chapter 2. The only difference in the analyses was that 300 image averages were used as the input for the 3D analysis of each individual population.

In each analysis of the six populations, dl calculations were done independently for each data set. 99

3.4: ResuIts

3.4.1: Isoiation, immuno-gold electron microscopy, and single particle counts of human myelin basic protein

Pnor to TEM observation of MBP, the protein samples for the three charge isomen of human MBP were tested by SDS-PAGE and Western blot analysis (Fig. 3- 1 ).

The SDS-PAGE gel (Fig. 3-1 (a)) showed that the three sarnples contained one species of protein each, with a molecular weight of 18.5 ma. This gel showed that the 18.5 kDa bands were fiee of proteolytic digestion products. Western immunoblotting was then used to identifi the three bands as MBP (Fig. 3- l(b)). In this blot a polyclonal anti-bovine CI

IgG was used, and it successfùlly identified al1 three bands. In this investigation a human source of MBP was used. Al1 the MBP was obtained via autopsy. and was purified at the

Hospital for Sick Children, Toronto, Canada, in the laboratory of Dr. Mario Moscarello.

In Chapter 2 of this thesis the anti-bovine IgG was used to show that bovine MBP bound to the lipid monolayer of PS/GM,(4:l). Since this polyclonal antibody was raised against bovine C 1, its binding to the three different human charge isomers of MBP which was demonstrated in Fig. 3-1 was further investigated. The three protein samples were diluted by 10X, 50X, and lOOX and blotted on the membrane, and probed with anti- bovine Cl/MBP (Fig. 3-2). They were then probed with a secondary antibody. and developed via the alkaline phosphatase method to provide a detectable signal. The blot was digitally scanned, and the mean relative opticai densities of the three different MBP charge isomers were calculated for the three different dilutions ( 1OX, 50X. 100X). When the Students t-Test was applied to the nine data sets (Table 3-1) it was found that the Figure 3- 1 : Biochemical characterization of human myelin basic protein. ( a,. Si lver stained polyacrylamide gel of myelin basic protein from human brain. Lanes I -3 wereloaded with charge isomers hC 1, hC8, and hMC8 respectively. (b) Western immunoblot of the polyacrylamide gel panel in (a) using rabbit anti-bovine MBP IgG. Figure 3-2: Antibody assay for human MBP. The polyclonal anti-bovine MBP IgG was tested for its ability to bind different human charge isomers of MBP (Cl. C8. and

MC8). (a) hC1. hC8, and hMC8 were diluted by 10X. 50X. and lOOX . blotted on nitrocellulose in replicate. and probed with the anti-bovine MBP IgG. The nitrocellulose blot was scanned by a digital carnera, and the opticai densities were measured for each spot. The mean relative optical density +/- standard deviation are plotted in this figure for the 10X (b), SOX (c), and lOOX (d) dilutions. These optical density values were then analyzed by the Students t-Test. The results of this analysis are presented in Table 3- 1. Re plic ates

23456 789101

Charge borner Charge S orner Charge Borner Variable 1 Variable 2 Y, + s, Y2 + % s,-~ df t Result hCl (10X) hC8 (IOX) 20.3 t 4.1 14.8 + 2.5 3.4 15 4.0 VI > VZ

hCl(10X) hM8(lOX) 20.5k4.1 13.6 I2.4 3.4 15 3.6 VI > V2

hC8 (10X) hM8 (10X) 14.8 k 2.5 13.6 f 2.4 2.4 18 0.7 V1 = V3 hC1 (SOX) hC8 (50X) 8.7 + 1.9 5.1 + 1.3 1.6 16 5.0 VI > VZ hC1 (50X) hM8 (50X) 8.7 + 1.9 5.2 k 0.8 1.4 12 5.4 VI > V2

hC8 (50X) hM8 (50X) 5.1 _+ 1.3 5.2 I0.8 1.1 15 0.2 VI = V3

hC1 (100X) hC8 (100X) 6.5 I 1.8 4.9 + 0.9 1.4 13 2.6 V1 > '42

hC 1 (IOOX) hM8 (1 00x) 6.5 k 1.8 4.5 t 1.2 1.5 16 3.0 VI > V2

hC8 (100X) hM8 (100X) 4.9 2 0.9 4.5 A 1.2 1.1 17 0.9 VI = V2

Table 3-1: t-Test for antibody assay of hurnan MBP. The Students t-Test results for the antibody assay for hurnan MBP (Fig. 3-2) are presented here. In this analysis the 'tg values were calculated for two samples, assuming unequal variances. Based on the degrees of freedom. a critical t value (0.05) of 1.8 was used. In al1 of the tests based on this critena, hC 1 probed with primary antibody (anti-bovine MEiP IgG) had significantly higher labelling then hC8 or hMC8. regardless of dilution. In the three dilution tests where the primary antibody binding to hC8 was compared to MC8 it was found that there was no significant difference. The terms that were used to define the variables are as follows: hC 1, hC8, and MC8 - C 1, C8, and MC8 charge isomers of human MBP; 10X, 50X, and lOOX refer to the dilution of MBP placed on the niîrocellulose membrane. 1O4 antibody bound hC1 significantly more than both hC8 and hMC8, regardless of dilution.

The t-Test also determined that the antibody bound hC8 and hMC8 equally. regardless of dilution.

Since the anti-bovine IgG was shown to bind al1 three human MBP isomers it was then used to probe the lipid monolayers of PSIG,, (4: 1) which had been incubated with hC1, hC8, hMC8 (Figs. 3-3, 3-4, 3-5, and 3-6). Close examination of these images showed that the protein A gold had bound to the MBP single particle complexes (Fig. 3-

7), thus showing the correlation between the single particles and MBP. To provide

Mersupport for this correlation the nurnber of gold parùcles per unit area (0.5 pm X

0.5 pm) were added up for positive and negative control experiments for MBP-antibody binding (Fig. 3-8). Exarnination of Fig. 3-8 clearly demonstrates that the positive controls for hC 1, hC8. and WC8 have the highest degree of protein A gold binding. These results were further quantified by the Students t-Test (Table 3-2)' which showed that in al1 three trials (hC 1, hC8, hMC8) that the protein A gold had significantly higher binding in the positive controls, compared to the negative controls. When the protein A gold signals were compared using the t-Test between the three positive controls it was found that the hMC8 had the highest binding, followed by hC1, and finally hC8. In each case the t-Test determined that the differences in gold binding were significant.

The immuno-gold data for MBP indicated that the three charge isomers showed different amounts of gold particles per unit area. However, it was established the primary antibody (anti-bovine Cl/MBP IgG) bound the three charge isomers with different affinity. Therefore, one could not definitely state that the three isomers bound the PSIG,, Figure 3-3: Immuno-gold electron microscopy on PS/Gw[l lipid monolayers. Fields of view of PS/GM~lipid monolayers that were probed with protein A gold (prA) are shown here. The lipid monoiayer was probed with four different anti body -gold combinations. These were as foilows: (a) only prA, (b) non-immune LgG and prA. (CI non-specific monoclonal antibody and prA, and (d) antibovine IgG and pr A. In Figs. (a- d) there was little if any gold label. Scale bar 100 nm. Figure 3-4: Immuno-gold electron microscopy of human Cl on PS/Gkl1 lipid monolayers. Fields of view of PS/GMl lipid monolayers that were probed with protein

A gold (prA) are shown here. The lipid monolayer was probed with four different antibody-gold combinations. These were as follows: (a) only prA. (b) non-immune IgG and prA, (c) non-specific monoclonal antibody and prA, and (d) anti-bovine IgG and pr

A. In Figs. (a-c) there was little if any gold label, whereas in Fig. (d) there was considerable gold labelling. The black arrow points to a single gold particle in (d).

Scale bar 100 nm. Figure 3-5: Immuno-gold electron microscopy of human C8 on PS/GM1 lipid monolayers. Fields of view of PS/GM~lipid monolayers that were probed with protein

A gold (prA) are shown here. The lipid monolayer was probed with four different antibody-gold cornbinations. These were as follows: (a) only prA, (b) non-immune IgG and prA, (c) non-specific monoclonal antibody and prA, and (d) anti-bovine IgG and pr

A. In Figs. (a-c) there was little if any gold label, whereas in Fig. (d) there was considerable gold labelling. The black arrow points to a single gold particle in (d).

Scale bar 100 nm. Figure 3-6: Immuno-gold electron microscopy of human MC8 on PS/Gwl lipid monolayers. Fieids of view of PS/GMI lipid monolayers that were probed with protrin

A gold (prA) are shown here. The lipid monolayer was probed with four different antibody-gold combinations. These were as follows: (a) only prA, (b) non-immune IgG and prA, (c) non-specific monoclonal antibody and prA, and (d) anti-bovine IgG and pr

A. In Figs. (a-c) there was little if any gold label, whereas in Fig. (d) there wns considerable gold labelling. The black arrow points to a single gold particle in (d).

Scale bar 100 nm. Figure 3-7: Immuno-gold MBP complexes. High magnification images of protein A gold (prA) bound to lipid layers which were incubated with hCI (a). hC8 (b), and hMC8

(c) are shown here. Unlike the other high magnification images in this thesis. rhese images were not recorded over a hole in the plastic film. Therefore, there is a considerable amount of background noise from the plastic support film that obscures the viewing of the single particle that is attached to the lipid monolayer and prA. In this figure the prA clearly appears as large black spots, and the single particle is difficult to see against the background film. In some cases the single particle may have been partially or totally covered by the gold particle. Scale bar 10 nm. 1234 1233 1234 1234 Buffer S hC 1 hC8 hCM8 Sample Type

Figure 3-8: Immuno-gold counts for human MBP. A sirnilar analysis to that camed out in Fig. 3-12, was appIied to the immuno-gold data for human 1MBP (Figs. 3-3,343-5, and 3-6). In this figure the mean number of gold particles per unit area (0.5 pm X 0.5 prn) was plotted for sixteen different sample preparations. These data sets can be divided up into four primary groups; the first set consisted of buffer S incubated with a lipid monoiayer, the second. third and fourth sets consisted of human CI, CS. and iMC8. respectively, suspended in buffer S with a lipid monolayer. AI1 samples were incubated at

32"~.Each of the four groups were probed with 4 different antibody-gold cornbinations

(labeled 1-4). These combinations are: (1) only protein A gold (prA), (2) preirnmune IgG and prA, (3) non-specific monoclonal antibody and prA. and (4) anti-bovine IgGand prA.

The mean +/- standard deviation are plotted in this figure, and these values were then analyzed by the Students t-Test. The results of this analysis are presented in Table 3-2. Table 3-2: t-Test for immuno-gold counts of human MBP. The Students t-Test results for the immuno-gold data for hurnan MBP (Figs. 3-3, 3-4, 3-5, 3-6, and 3-8) are presented here. In this analysis the 't' values were calcdated for two samples, assuming unequal variances. Based on the degrees of fieedom, a critical t value (0.05) of 1.7 was used. In the fim three blocks. when three isomers (hC1, hC8, and hMC8) were probed with primary antibody. and protein A gold there is significantly more gold labelling in these positive controls, then in the negative controls.

The last three rows of this table compare the gold binding between the three positive controls.

In this case the MC8 > Cl > C8, with respect to the arnount of gold labelling. The terms that were used to define the variables are as follows: hC 1, hC8, and hMC8 - C 1. C8. and MC8 charge isomers of human MBP, respectively; B - baer S; 1 - primary anti-bovine MBP IgG; A -protein

A go1d;N - non-immune IgG; M - monoclonal antibody that is not specific for MBP. - - Variable 1 Variable 2 Y,k s, Y: + s SI-: d f t Rrsult I

hC1. 1. A

hC8. 1. A

hMC8. 1. A

hCI. 1. A

hMC8. 1, A hMC8. 1. A 113

(4: 1) lipid rnonolayer with different affinity.

Since the immuno-gold data did identify the particles to be composed of MBP (at least a component of the particles) the next logical experirnent was to count the nurnber of particles per unit area (0.3 Fm X 0.3 pm) when the different charge isomers interacted with the PS/GM,(4: 1) lipid monolayer (Fig. 3-9). This expenment was similar to the one conducted with bovine C 1, where the PSIG,, (4: 1) lipid monolayer was incubated with buffer-S done and buffer-S and MBP. This experiment differed by the fact that the three isomers were compared at equimolar concentrations. and also a time course experiment was conducted with samples taken der a 0.5 h, 1.0 h, 2.0 h, and 6.0 h incubation period at 32°C (Fig. 3-10). In each preparation the samples showed some variation over time. however the differences between the four samples (buffer-S; buffer-S + hC 1: buffer-S + hC8; buffer-S + hMC8) showed the greatest variation. For cornparison each sarnple type was then added up to produce a global mean value of particles per unit area regardless of tirne of incubation. When the global mean numbers of particles per unit area were compared using the Students t-Test (Table 3-3), in al1 cases the MBP-buffer-S samples had significantly more single particles than the buffer-S aione. Furthermore. the t-Test demonstrated that the most particles were produced in the hMC8 preparation. followed by hC1, and finally hC8. In al1 cases the t-Test indicated that these differences were significant, thus showing that the different charge isomers interacted differently with the lipid monolayer. The next step in this investigation was the structural analysis of the single particles that were being produced on the PS/G,, (4:l) lipid monolayer by cornputer image analysis in 2D, and the 3D. This will be the topic of the following Figure 3-9: Electron micrographs (TEM) of Iipid layers and human MBP. Shown here are charactenstic fields of view of PS/%l lipid monolayers that have been incubated with buffer S (a), buffer S and hC1 (b), buffer S and hC8 (c), and buffer S and hMC8 (d). Al1 the incubations were at 32 OC, for 0.5 hour, and the concentration of MBP was constant. It can be seen that hMC8 had the highest amount of single particle nucleation followed by hCI, hC8, and finally buffer S alone. In these images the particles were stained with 2% uranyl formate. and appear as light particles with a dark halo of stain around them. Several single particles have been circled. Scale bar 100 nm. Figure 3-10: Particle çounts for human MBP. A sirnilar analysis to that carried out in Fig.

2- 10 was applied to the single particle data for human MBP (Fig. 3-9 ). MBP - lipid monolayer interactions were quantified by counting the number of single particles in a unit area (0.3 pm X

0.3 Pm). Four sample preparations were compared, one with buffer S and lipid, and three others containing either human Ci. C8 or MC8. AI1 three preparations were suspended in buffer S and lipid. in al1 four preparations the numbers of single particles were counted after a

0.5, 1.0, 2.0 and 6.0 hour incubation rimes at 32 OC. The mean +/- standard deviation are plotted in this figure. Panel (a) of this figure shows the individual mean values for each sample set. and panel (b) shows the total mean value for each of the four populations. where al1 the values from the time course experiment have been summed together. These data roughly correlate with the particle sizes for the three MBP isomen (Figs. 3-27 and 34). The hMC8 produced the largest particles, and had the highest nurnber of single particles per unit area. hC1 was intermediate for both values. and hC8 produced the smallest particles, and had the lowestsingle particle counts. The summed values for these data (b) were analyzed by the

Students t-Test. The results of this analysis are presented in TabIe 3-3.

Variable 1 Variable 2 Y,f s, Y, + s, SI-z Result hC 1 B 146.3 + 30.6 4.2 f 5.1 31.4 83 -9 hC8 B 34.8 + 15.4 4.2 f 5.1 11.8 38.8 VI > V2 hMC8 B 251.3 k23.4 4.2f5.1 15.9 hC 1 hC8 146.3 t 30.6 34.8 + 15.4 23.3 hMC8 hC 1 25 1.3 + 23.4 146.3 I30.6 27.5 hMC8 hC8 251.3 + 23.4 34.8 + 15.4 19.0

Table 3-3: t-Test for particle counts of human MBP. The Students t-Test results for the single particle data for human MBP (Figs. 3-9, and 3-10) are presented here. In this analysis the 't' values were caiculated for two samples, assuming unequd variances. Based on the degrees of fieedom, a cntical t value (0.05) of 1.7 was used. The data presented here represent the single particle counts for the surnrned data fiom the time course experiment (Fig. 3- 10 (b)). In the tests based on this criteria, when MBP single particle counts were compared to counts from buffer S alone, MBP always had a significantly higher number of particles. In the experiments where the particle counts for the different charge isorners were compared, it was found that hMC8 > hC1

> hC8, indicating, that the different charge isomers have significantly different interactions with the lipid monolayer, with hMC8 having the greatest degree of interaction, and hC8 the lest. The terms that were used to define the variables are as folIows: hC1, hC8, and hMC8 - Cl. C8. and

MC8 charge isomers of human MBP, B - baer S. sections.

3.4.2: Two-dimensional analysis of human myelin basic protein TEM images

Transmission electron microscopy of the uranyl formate negatively stained hurnan

MBP particles yielded the images appearing in Fig. 3-9. Upon initial exarnination. the negatively stained particles viewed on the phosphatidylserine and rnonosialoganglioside

GM, lipid layer look quite homogeneous in nature. Each data set for the three MBP charge isomers was imaged at liquid nitrogen temperature in the TEM. and negarively stained with aqueous 2% uranyl formate. The particles al1 appear to be the sarne size, and there appears to be little particle aggregation; thus these spreads are ideal for single particle cornputer andysis.

Two-dimensional image analysis of the particles that were produced by the interaction of buffer-S with the PS/GM,(4: 1) lipid monolayer is shown in Fig 3-1 1. The class averages were in general circular. However, they do not appear to have the distinct stain filled pores that are visible in the single particle averages that were produced when

MBP was incubated with the PS/G,, (4: 1) lipid monolayer (Figs. 3- 12, 3- 13. and 3- 14).

Two-dimensional image analysis of the hC 1 (6172 particles) data set yielded the class averages shown in Fig. 3- 12. The class averages of hC 1 are in general circular with a 7-9 nrn diameter, most of which have a stain-filled pore, or cavity in the center.

The results of the 2D image analysis of the hC8 (5017 particles) data set are show in Fig. 3-13. The hC8 averages appear more compact than the hC1 averages. and many are prolate in shape with a minor axis of 5-6 nrn, and a major axis of 9 nm. As Figure 3-11: Class averages of particles produced on lipid monolayers of PS/Gk[l and high salt buffer. PS/GM~lipid monolayers were incubated with buffer S and were imaged by TEM, at liquid nitrogen temperature, and stained with 2% uranyl formate

(Fig. 3-9 (a)); 416 single particles were used in the computer analysis. Scale bar 10 nm. Figure 3-12: TEM class averages of human Cl produced by MSA. C 1 was imaged by

TEM, at liquid nitrogen temperature. and stained with 2% uranyl formate (Fig. 3-9(b)):

6172 particles were used in the cornputer analysis. In this figure some of the common groups of class averages have been circled. These groups are the; C (grey circles). M

(white circles), and prolate (black circles) groups. Scale bar IO nm. Figure 3-13: TEM class averages of human C8 produced by MSA. C8 was imaged by

TEM, at liquid nitrogen temperature, and stained with 2% uranyl formate (Fig. 3-9 (c)):

5017 particles were used in the cornputer analysis. In this figure some of the comrnon groups of class averages have been circled. These groups are the; C (grey circles). prolate (white circles), and small three fold (black circles) groups. The white arrows point to the knob structure, and the black arrows point to the pore structure. Scale bar 10 nm. Figure 3-14: TEM class averages of human MC8 produced by MSA. Marburg type

C8 was imaged by TEM, at liquid nitrogen temperature, and stained with 28 uranyl formate (Fig. 3-9 (d)); 5266 particles were used in the cornputer analysis. In this figure some of the common groups of class averages have been circled. These groups are the: prolate (grey circles), big three fold (white circles), and small three fold (black circles) groups. Scale bar 10 nm. 123 was the case with the hC1 averages, many of the hC8 averages have a central stain filled pore. In many of the prolate classes the pore is associated with a knob-like protuberance on one end of the major axis of the particle. In other prolate classes there is no pore. but the knob is present at one end of the major axis of the particle.

The results of the 2D image anaiysis of the hMC8 (5266 particles) data set are show in Fig. 3-14. The hMC8 class averages appear to be circular with a 8-10 nm diameter. In many of the class averages of MC8there appears to be a suggestion of three-fold symmetry. Many of the class averages have three gain-filled pores. Other averages have only one centrai pore, but the stain excluding protein mass around it also shows signs of three-fold syrnrnetry.

Cornparison of selected class averages of hC1. hC8, and hMC8 to their raw noisy aiigned class members shows a reasonable amount of correlation between the images and the averages (Figs. 3- 15, 3-16, and 3-1 7). The resolution of the 2D aligned images was caiculated by the Fourier ring correlation method, using the canonical significance threshold of three standard deviations as suggested by van Heel (1987b). The resolutions were estimated to be 0.9 nrn, 1.6 nm, and 1.7 nrn for the hC1, hC8. and hMC8 2D data. respectiveiy (Fig. 3- 18).

3.4.3: Three-dimensional analysis of human myelin basic protein TEM images

The iterative quaternion-assisted angular determination process (Farrow and

Ottensmeyer, 1992, 1993) was applied independently to the three populations of MBP class averages (300 averages in each population). The 3D reconstructions computed using Figure 3-15: TEM class members and averages of Cl produced by MSA. C 1 MBP was imaged in the TEM and analyzed by single particle analysis. Sample classes 50 (a) and 92 (b) from Fig. 3-12 are shown here. In both figures the individual class members are shown, and the last image in each figure is the average of al1 the ciass members.

Scde bar 10 nm. Figure 3-16: TEM class members and averages of CS produced by MSA. C8 MBP was imaged in the TEM and analyzed by single particle analysis. Sainple classes 50 (a) and 66 (b) from Fig. 3-13 are shown here. In both figures the individual class memben are shown. and the 1st image in each figure is the average of al1 the class members.

Scale bar 10 nm. Figure 3-17: TEM class members and averages of MC8 produced by MSA. MC8

MBP was imaged in the TEM and analyzed by single particle analysis. Sample classes

6 (a) and 50 (b) from Fig. 3-14 are shown here. In both figures the individual class members are shown, and the last image in each figure is the average of al1 the class members. Scale bar IO nm. sigma

rsigma

Pixel

Figure 3-18: TEM resolution caiculation of 2D averages of human myelin basic protein. An estimate of the spatial resolution of the 2D aligned images of human C 1

(Fig. 3- 12) (a), CS (Fig. 3- 13) (b). and MC8 (Fig. 3- 14) (c) were calculated by Fourier ring correlation (FRC).Image (a) was scanned at 0.223 nrn/pixel, whereas (b) and (c) were scanned at 0.276 nrdpixel. Using a a value of 3, the resolution was estimated to be

0.9 nm, 1.6 nrn, and 1.7 nm for hC 1, hC8, and hMC8, respectively. Figure 3-19: 3D reconstruction of human Cl from TEM images. 300 hurnan C 1 2D class averages (Fig. 3-12) of TEM images were used to produce this reconstruction by angular reconstitution. The reconstruction is illustrated here rotated at 90° angles to the vertical axis of this page (a-ci). the top and bottom views are shown in ce) and (0 respectively. The reconstruction is shown here in stereo as a shaded surface representation. The CI reconstruction, as determined by TEM appears to have a hollow crucible type shape with a cap. The scale bar corresponds to 5 nm. however when the particles are back-projected the volume of the reconstruction was reduced. Therefore. the size of the reconstmctions are srnaller then the initial 2D data. In al1 subsequent 3D figures the term "scale bar corresponds to" refen to the scale of the voxels, and not the reconstruction. Figure 3-20: 3D reconstmction of human CS from TEM images. 300 human C8 2D class averages (Fig. 3- 13) of TEM images were used to produce this reconstruction by angular reconstitution. The reconstruction is illustrated here rotated at 90° angles to the vertical mis of this page (a-d), the top and bottom views are shown in (e) and (0 respectively. The reconstruction is shown here in stereo as a shaded surface representation. The C8 reconstmction, as determined by TEM appears to have a hollow crucible type shape, but is more compact and then the C 1 reconstruction (Fig. 3- 19). and has a prominant cap. The scale bar corresponds to 5 nm. Figure 3-21: 3D reconstruction of human MC8 from TEM images. 300 human MC8

2D class averages (Fig. 3-14) of TEM images were used to produce this reconstruction by angular reconstitution. The reconstruction is illustrated here rotated at 90° angles to the verticai axis of this page (a-d), the top and bottom views are shown in (e) and (0 respectively. The reconstruction is shown here in stereo as a shaded surface representation. The MC8 reconstruction, as detemined by TEM appears to have a hollow crucible type shape with a distinct cap, with some three fold symmetry which is can be seen in (e). The scale bar corresponds to 5 nm i ' -.- . -.O 1 . .- t i creasing ' - 1 . .. ! - . . 4..--- '4 90...... - . '..' i - '.*-..:. : - .C'j .-. . .. I -' - - ,-- .'f .- .--.

Figure 3-22: Angular orientations of myetin basic protein TEM 3D reconstructions, plotted cartographically as points on a hemispherical surface. The angular orientations of human C 1 (a), C8 (b), and MC8 (c) 3D reconstructions (Figs. 3- 1 9. 3-20. and 3-21, respectively) are shown here. The reconstructions for (a-c) were produced from 300 class averages which are plotted here. Figs. 3- 12, 3- 13. and 3- 14, respectively show 100 class averages each, for the TEM data. The square points represent angular orientations in the foreground hemisphere, and the crosses show angular orientations in the background hemisphere. In general these plots indicate that there is little clustering of the particles around a specific orientation. It seems that the particles tend to lie in bands dong regions of the plot, indicating preferred adsorption dong an axis of the

MBP-lipid cornplex. 132 filtered back-projection are shown in stereo in Figs. 3-19. 3-20. and 3-21. The distributions of the Euler angles on the unit sphere are show in Fig. 3-22. for the three reconstructions. Overall, the three reconstructions are similar in that they have a

"crucible" type shape, with a "cap" on the top of each reconstruction. The cap region of the reconstruction from this point onward will be arbitrarily referred to as the top of the reconstruction. Thus images viewed from this perspective will be referred to as a top view, and images viewed perpendicuiar to the cap will be referred to as a side view. The hC8 reconstruction (Fig. 3-20) is the most compact of the three reconstructions. and the hMC8 reconstruction (Fig. 3-2 1) appears to be the most extended, with its apparent three- fold symmetry .

In order to test the quality of the Euler angle assignment (Fig. 3-22). the initial class averages were compared to the forward projections. and shaded surface representations of the 3D reconstructions. Figs. 3-23, 3-24, and 3-25 show the first 40 of 300 images fiom the 2D data sets (100 averages are shown in Figs. 3-12. 3-13. and 3-

14) compared to their corresponding 3D representations from hC1. hC8. and hMC8 reconstructions. Overall the correspondence between the images is fair. but not exact.

In general the forward-projections appear more compact then their corresponding class averages, and a fair amount of the density has been smoothed out giving the forward- projections a ring-like appearance. The best correspondence appears to be in the hMC8 population, where the characteristic three-fold appearance shows up in the fonvard- projections.

Resolution estimates caicdated by the Fourier shell correlation method for the ca

f P

ss

- Filpre 3-23: TEM Cl 3D reconstruction; class averages, forward-projections, and shi ided surface representations. The first fony of the 300 2D class averages (c :a) ( Fig.

3- 12), are shown here with their corresponding forward-projections (fp), and sh; ided

SUIface representations (ssr). For each triplet set, the images correspond to one an iothc:r in thait they al1 share the same set of Euler angles (Fig. 3-22 (a)) that were used to pro(luce the: 3D reconstruction of C 1 (Fig. 3- 19). This figure provides a visual cornparisoi1 foi: the

"qilality of fit" between the initial class averages and the final reconstmction. Scale : bar

10 nm. Figure lS, imd shaded 4 (1 Fig.

3- 13), shaded surface )the that the ~rod the 3D for the

"qualit! :aie bar

IO nm. Figure 3-25: TEM MC8 3D reconstruction; class averages, forward-projections, and shaded surface representations. The first forty of the 300 2D class averages (ca) (Fig.

3-14), are shown*here with their corresponding fonvard-projections (fp). and shaded surface representations (ssr). For each triplet set. the images correspond to one another in that they al1 share the same set of Euler angles (Fig. 3-22 (c)) that were used to produce the 3D reconstruction of MC8 (Fig. 3-21). This figure provides a visual cornparison for the "quality of fit" between the initial class averages and the final reconstruction. Scale bar 10 nm. FS C -..- sigma

Pixel

sigma

3 0.8 LI 4 c -3 0.6 '53 \ C -13 0.4 s I Ë o., S

Pixel Figure 3-26: TEM resolution calculation of 3D averages of human myelin basic

protein. An estirnate of the spatial resolution of the 3D reconstruction of human C 1

(Fig. 3- 19) (a), C8 (Fig. 3-20) (b), and MC8 (Fig. 3-2 1) (c) were calculated by Fourier

ring correlation (FRC). Image (a) was scanned at 0.223 ndpixel, whereas (b) and (c)

were scanned at 0.276 ndpixel. Using a o value of 3, the resolution was estimated to

be 2.2 nm, 2.8 nm, and 2.1 nm for hC 1, hC8, and hMC8, respectively. Figure 3-27: Cornparisons of TEM reconstructions of human myelin basic protein.

The human Cl (left). C8 (center), and MC8 (right) 3D reconstructions are cornpared here. in this figure side (a.b), and top (c,d) view orientations are shown, as both full

(a.c), and cut away (b,d) representations. When the reconstructions are compared this way it is apparent that the reconstructions are similar in that they are hollow. The 3D reconstmctions are dso quite different from one another, with the C8 being the most compact, and the MC8 being the most extended with potential three fold symrnetry. The scale bar corresponds to 5 nm. 138 human MBP 3D reconstructions produced an estimated resolution of 2.2 nrn. 2.8 nm. and

2.1 nm for the hC1, hC8 and hMC8 reconstructions, respectively (Fig. 3-26). Fig. 3-27 compares al1 three reconstnictions (hC1, hC8, and MIIC8) directly as Ml, and cut away views; the reconstructions are viewed from the top and side. These final cornparisons are in agreement with the initiai 2D averages (Figs. 3-12. 3-1 3, and 3-1 4). In 3D the reconstructions al1 have a distinct surface shell with a central cavity. which corresponds to the "stain-filled pore" present in the 2D averages. The side on view. and top view of the hMC8 reconstruction both show the apparent three-fold syrnmetry which was present in the 2D class averages of hMC8.

3.4.4: Two-dimensional analysis of human myelin basic protein STEM images

Scanning transmission electron microscopy of the methylamine vanadate stained

MBP induced single particles on PS/GM,(4: 1) lipid monolayers yielded the images show in Fig. 3-28. In these images the majority of the electron scattering events are caused by the stain. When the scattered electrons are collected by the small angle annula dark field detector the stain produces the majority of the signal. The STEM images have reverse contrast cornpared to the TEM images (Fig. 3-9), in these images the stain appears as a white halo surroundhg the dark MBP single particle. As was the case in the TEM data

(Fig. 3-9), the particles are well separated, and are ideal for single particle analysis.

2D images of the hC 1 data set (5439 particles) produced the class averages shown in Fig. 3-29. Although similar in overall dimensions to the hC 1 TEM class averages (Fig.

3-12) the images look dramatically different. The STEM averages tend to look crisper. Figure 3-28: Electron micrographs (STEM) of human myelin basic protein Three fields of view of human MBP are shown here. C 1, C8, and MC8 are shown in figures (a).

(b), and (c) respectively. MBP was imaged by STEM at 500 K magnification. at liquid nitrogen temperature, with a pixel size of 0.33 ndpixel. The sample was negatively stained with 28 methylamine vanadate. In these images the protein appears as a dark structure surrounded by a bright halo of negative stain. The MBP is shown here adsorbed to a lipid monolayer of GM~and phosphatidylsenne. Several single particles have been circled. Scale bar 10 nm. Figure 3-29: STEM class averages of human Cl produced by MSA. Cl was irnaged by STEM, at liquid nitrogen temperature, and stained with 28 methylamine vanadate;

5420 particles were used in the computer analysis. One hundred class averages are shown here. In this image the density has been inverted from Fig. 3-28, therefore, the protein is white, and the negative stain appears dark . Scaie bar 10 nm. 141 and granular in comparison to the TEM data, which look smooth and blobby. The major difference between the two data sets (TEM vs. STEM) is absence of the stain-filled pore in the STEM class averages. Similar resuits were generated for the hC8 data (Fig. 3-30:

6346 particles), and the hMC8 data (Fig. 3-3 1; 5349 particles), al1 were lacking the distinct central pore. In both populations the averages have a crisp and granular appearance. The hC8 class averages in general appear to be prolate, in agreement with the TEM data (Fig. 3-13), with a "knobl'apparent on some of the averages. The MC8 averages dso demonstrate a fair degree of similarity with their TEM counterparts (Fig.

3- 14). In Fig. 3-3 1 many of the classes have a triangular appearance, in accord with the apparent three-fold symmetry apparent in the TEM data.

Figs. 3-32, 3-33, and 3-34 show the correlation between the individual aligned class rnembers, and their average. In general, there is a reasonûble agreement between the noisy raw images, and their averages. Fourier ring correlation was used to estimate the resolution of the aligned single particles (Fig. 3-35). The resolution was estimated to be 0.9 nrn, 0.9 nm, and 1 .O nrn for the hC 1, hC8, and hMC8 2D populations, respectively.

3.4.5: Three-dimensional analysis of human myelin basic protein STEM images

The exact sarne process of iterative quaternion-assisted angular determination

(Farrow and Ottensmeyer, 1992, 1992) was applied to the STEM hC 1, hC8, and MC8 data sets, as was applied to their twin TEM data sets. The filtered back-projections are shown in stereo in Figs. 3-36, 3-37, and 3-38 for hC1, hC8, and hMC8, respectively. The distributions of the Euler angles on the unit sphere are shown in Fig. 3-39. Compared to Figure 3-30: STEM class averages of human CS produced by MSA. C8 was imaged by STEM, at liquid nitrogen temperature, and stained with 2% methylarnine vanadate:

6346 particles were used in the computer analysis. One hundred class averages are show here. In this image the density has been inverted from Fig. 3-28, therefore, the protein is white, and the negative stain appears dark . Scale bar 10 nm. Figure 3-31: STEM class averages of human MC8 produced by MSA. MC8 was imaged by STEM, at liquid nitrogen temperature, and stained with 2% methylamine vanadate; 5439 particles were used in the computer analysis. One hundred class averages are shown here. In this image the density has been inverted from Fig. 3-28, therefore, the protein is white, and the negative stain appears dark . Scale bar IO nm. Figiure 3-32: STEM class members and averages of Cl produced by MSA. C1

MB P was imaged by STEM and analyzed by single particle analysis. Sample classc:s 29

(a) and 81 (b), from Fig. 3-29 are shown here. In both figures the individual class mer nbers are shown, and the last image in each figure is the average of al1 the cIass mer nbers. Scale bar 10 nm. Figure 3-33: STEM class members and averages of C8 produced by MSA. C8

MBP was imaged by STEM and analyzed by single particle analysis. Sample classes 12

(a) and 29 (b), from Fig. 3-30 are shown here. In both figures the individual class

members are shown, and the last image in each figure is the average of al! the class

rnembers. Scale bar 10 nm. Figure 3-34: STEM class members and averages of MC8 produced by MSA. MC8

MBP was imaged by STEM and analyzed by single particle analysis. Sample classes 3

(a) and 41 (b), from Fig. 3-31 are shown here. In both figures the individual class members are shown, and the Iast image in each figure is the average of ail the class members. Scale bar 10 nm. -.. -..

*. -. .-..--.--.

O t 2 3 J 5 6 7 3 9 10 11 17 13 14 15 16 17 iY 19 20 Pixel

Pive 1

1.4 1

Pixel Figure 3-35: STEM resolution calcdation of 2D averages of human myelin basic protein. An estimate of the spatial resolution of the 2D aligned images of human CI

(Fig. 3-29) (a), C8 (Fig. 3-30) (b), and MC8 (Fig. 3-3 1) (c) were calculated by Fourier ring correlation (FRC).These images were scanned at 0.33 nmlpixel. Using a o value of

3, the resolution was estimated to be 0.9 nm, 0.9 nm, and 1.0 nm for hC1, hC8, and hMC8, respectively, representing an approximate 1.5 fold increase in resolution over the

2D TEM data (Fig. 3- 18). Figure 3-36: 3D reconstruction of human Cl €rom STEM images. 300 human C 1 2D class averages (Fig. 3-29) of STEM images were used to produce this reconstruction by angular reconstitution. The reconstruction is illustrated here rotated at 90° angles to the vertical axis of this page (a-d), the top and bottom views are shown in (e) and (fl respectively. The reconstruction is shown here in stereo as a shaded surface representation. The Cl reconstruction as determined by STEM is similar to the TEM reconstruction of Cl (Fig. 3-19) in overall geometry, but the STEM reconstruction has more structural detail. The scale bar corresponds to 5 nm. Figure 3-37: 3D reconstruction of human C8 from TEM images. 300 human C8 2D class averages (Fig. 3-30) of STEM images were used to produce this reconstruction by angular reconstitution. The reconstruction is illustrated here rotated at 90° angles to the vertical axis of this page (a-d). the top and bottom views are shown in (e) and (0 respectively. The reconstruction is shown here in stereo as a shaded surface representation. The C8 reconstruction as deterrnined by STEM is similar to the TEM reconstruction of C8 (Fig. 3-20). with the exception of the large "cap" present on the top of the C8 TEM reconstmction. As was the case with Cl, the STEM reconstruction has more structural detail than the TEM reconstruction. The scale bar corresponds to 5 nm. Figure 3-38: 3D reconstruction of human MC8 from STEM images. 300 human iMC8 2D class averapes (Fig. 3-31) of STEM images were used to produce this reconstruction by angular reconstitution. The reconstruction is illustrated here rotated at

90° angles to the vertical axis of this page (a-d), the top and bottom views are shown in

(e)and (f) respectively. The reconstruction is shown here in stereo as a shaded surface representation. The MC8 reconstruction as determined by STEM is similar to the TEM reconstruction of MC8 (Fig. 3-21), however the three fold syrnmetry and "caps"differ between the two reconstructions. Consistent with dl the reconstructions, the STEM reconstruction has more structural detail than the TEM reconstruction. The scale bar corresponds to 5 nm. 'Ç. &' $ . / . & .- . .

Figure 3-39: Angular orientations of myelin basic protein STEM 3D reconstructions, plotted cartographically as points on a hemisphericai surface. The angular orientations of the human Cl (a), C8 (b), and MC8 (c) 3D reconstructions (Figs.

3-36. 3-37, and 3-38, respectively), are shown here. The reconstmctions for (a-c) were produced frorn 300 class averages which are plotted here. Figs. 3-29. 3-30, and 3-3 1 show 100 STEM 2D class averages for each MBP isomer population: these 2D averages were used as the input for the angular reconstitution process. The square points represent angular orientations in the foreground hemisphere, and the crosses show angular orientations in the background hemisphere. In general these plots indicate that the particles tend to lie in bands along regions of the plot. This indicates preferred adsorption along an axis of the MBP-lipid cornplex. These resulrs agree with the TEM data (Fig. 3-22) which produced similar banded plots. 152 the TEM reconstnictions (Figs. 3- 19, 3-20, and 3-2 1) the STEM reconstructions display an increased amount of visible fine structure. The STEM reconstructions are fiiled with small cavities and out-croppings compared with the relatively smooth TEM reconstructions. Consistent with the TEM data the hC8 reconstruction is the most compact (Fig. 3-37), and the MC8 reconstruction appears to be the most extended.

Euler angle assignment (Fig. 3-39) using the class averages (2D data). fonvard- projections (3D data), and shaded surface representations (3D data) for hC1. hC8. and hMC8 (Figs. 3-40, 3-41, and 3-42, respectively) was carried out for the STEM data. As was the case for the TEM data (Figs. 3-23, 3-24, and 3-25) the correlation of the 2D data to the 3D data was fair, with the hMC8 data (Fig. 3-25) showing the highest degree of consistency. Fourier shell correlation resolution estimates of the 3D reconstructions of hC 1. hC8. and hMC8 were calculated to be 2.4 nm, 2.5 nm. and 2.4 nm. respectively

(Fig. 3-43). These results are comparable to the values for the 3D reconstnictions of human MBP fiom the TEM data (Fig. 3-26). Fig. 3-44 compares ail three (hC 1. hC8. and hMC8) STEM reconstructions directly as full. and cut away views; the reconstnictions are viewed fiom the top and side. Sirnilar to the TEM reconstructions. the STEM reconstructions compare favourably with their 2D class averages (Figs. 3-29. 3-30. and

3-31). The major difference between the reconstructions (TEM vs. STEM) lies in the intemal mass distribution. In this case, the STEM 3D reconstructions al1 have a solid interior, consistent with the 2D class averages which also exhibited a sotid interior. The

TEM reconstructions (Fig. 3-27) al1 exhibited a hollow interior. Although these results seem irreconcilable, one must remember that what were produced were 3D reconstnictions 153 of a metal casting of the MBP-lipid complex. Therefore. the densities represent a surface topology of the complex, and not the intemal mass distribution. Figure IS, iand shaded a) (:Fig.

3-29), i sha.ded surface ~he!r in that the ~rodluce the 3D for the

"quaiity :ale bar

10 nm. Fi gure ZS, and shiaded a) (Fig.

3- 301, shaded

SU rface xher in th;at the roduce tht :3D : for the

"quaiity :ale bar

IO1 nm. Figiire !ctions, and sh ;es (Ca)

(Fig . 3- shiaded surf;ace )th(er in that the )rO(iuce the 3D SOI1 for the "qu Skale bar 10 i FSC -.. - sigma- 157

Pixel

Pixel

O 1 1 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 10 Pixel Figure 3-43: STEM resolution calculation of 3D reconstmctions of human rnyelin basic protein. An estimate of the spatial resolution of the 3D reconstructions of human

Cl (Fig. 3-36) (a), C8 (Fig. 3-37) (b), and MC8 (Fig. 3-38) (c) were calculated by Fourier shell correlation (FSC).These images were scanned at 0.33 nmfpixel. Using a G value of 3, the resolution was estimated to be 2.4 nm, 2.5 nm, and 2.4 nm for hC1, hC8, and hMC8, respectively. Figure 3-44: Cornparisons of STEM reconstructions of human myelin basic protein.

The human Cl (left), C8 (center), and MC8 (right) 3D reconstmctions are compared here.

In this figure side (a,b), and top (cd) view orientations are shown, as both full @.CL and cut away (b,d) representations. When the reconstructions are compared this way it is apparent that the reconstructions differ considerably from the TEM reconstructions (Fig.3-

77) which were essentially hollow. The 3D STEM reconstructions have intemal mas.

The 3D reconstructions are also quite different from one another, with C8 being the most compact, and C 1 and MC8 being more extended. The scale bar corresponds to 5 nm. 159

3.5: Discussion

High resolution TEM and STEM were used in concert to ascertain the huictional interactive and structural differences and similarities between different human iMBP charge isomers (Wood and Moscarello, 1989, 19%; Wood et al., 1996) when they interacted with a PS/GMl (43) lipid rnonolayer. The analyses in this chapter were similar to those conducted in Chapter 2. However, they have been expanded to study human MBP charge isomers C 1, C8, and MC8 (Wood and Moscarello, 1989, 1996; Wood et al.. 1996). The use of low dose STEM improved the structural resolution (Fig. 3-35) of the images compared to the TEM images (Fig. 3-18).

In the introductory section of this thesis the biochemical properties of MBPs. with respect to their involvement in MS were discussed. MBP isolated from MS afflicted individuals was less cationic than that from age-matched controls and from patients with other neurological diseases (Moscarello et al., 1994). This snidy directly investigated the degree of interaction and structural properties of three purified human MBP charge isomers in situ when they interacted with a PSIG,, (4:l) lipid rnonolayer.

One of the major stumbling blocks in myelin research lies in the physical nature of myelin. In the native in vivo state, myelin-forming cells are in contact with several other types of cells (Morell, 1977). This intimate contact gives myelin its unique membranous structure, permitting it to perform its fûnction as the insulator for the axon that it has enrobed. Presently it is not possible to mimic myelin structure faithfully in vitro. The only way to study myelin structure and fünction is via the analysis of the isolated myelin components in vitro. In this study it was necessary to isolate the different 160 charge isomers of MBP. The only method by which one can separate these isomers is by the use of acid extraction, followed by chromatography in 2 M urea (Cheifetz and

Moscarello, 1985; Wood and Moscarello, 1989). Unfomuiately in doing this the native structure of MBP will be perturbed. Studies on isolated MBP have indicated that the conformation of MBP is predominantly a random coi1 in aqueous solution (Chapman and

Moore, 1976; Epand et al., 1974; Gow and Smith, 1989; Krigbaum and Hsu. 1975). The random coi1 structure may arise as an artefact of isolation.

Many spectroscopic studies using circular dichroism (CD) or FTIR have revealed that MBP undergoes a significant conformational change in the presence of lipid vesicles or detergent micelles, as well as afier phosphorylation (Anthony and Moscarello. 197 1:

Deibler et al., 1990; Keniry and Smith, 1979. 198I ; Lees and Brostoff. 1984: Mendz et al., 1988, 1991; Moore, 1982; Ramwani et al., 1989; Smith, 1977b; Stuart. 1996:

Surewicz et al., 1987), adopting some a-helical and P-sheet structure. Therefore it has been postdated that in vitro the structure of MBP when associated with lipid is probably more representative of its structure in vivo. The microheterogeneity of MBP gives rise to structural variation and changes in the protein's ability to interact with lipids (Cheifetz and Moscarello, 1985; Jones and Epand, 1980; Wood and Moscarello, 1989). The goal of this sîudy was to determine if the hC1, hC8, and hMC8 charge isomers which are predominant in normal, chronic MS, and Marburg type acute MS individuals. respectively. behave differently, and have different tertiary and quaternary structures. If so, one could postulate that the in vitro behaviour (degree of interaction and structural) of the charge isomers may reflect their roles in vivo in normal and MS afflicted individuals. 161

3.5.1 : Interaction of myelin basic protein with PS/G,,, lipid monolayers

The fvst stage of this investigation was to provide positive identification that the particles seen on the EM contained MBP. In this study equimolar amounts of hC 1. hC8. and hMC8 were applied to the lipid monolayer (Fig. 3-9). The biochemical purity of the three isomers had been confirmed by SDS-PAGE and Western blot analysis (Fig. 3-1).

The SDS-PAGE gel indicated that there was no proteolytic degradation. In the Western blot the anti-MBP IgG only bound the 18.5 kDa bands, and not the molecular weight markers. thus providing a positive identification for MBP. In order detect MBP in the

EM, immuno-gold electron microscopy was used to probe the PS/G,, (4:l) lipid monolayers. Since a polyclonal anti-bovine MBP/Cl IgG was being used on human charge isomers of MBP, its binding efficiency to the three human charge isomers was first tested and analyzed by the Students t-Test (Fig. 3-2, Table 3-1). This analysis showed that the antibody bound al1 three isomers, but it bound hC1 with higher efficiency than hC8 and hMC8, which it bound equally well.

Immuno-gold electron microscopy was used to probe the PS/GM, (4:l) lipid monolayers (Figs. 3-3, 3-4, 3-5, 3-6). Usually immuno-gold electron microscopy is applied to cellular structures, in order to localize a specific epitope on a cellular component. In doing this, the ce11 structure is weakiy stained, and serves as a backdrop for the irnmuno-gold particles. This process has been applied to the localization of MBP in the myelin membrane (e.g., McLaurin et al., 1993). In this investigation randomly oriented and randomly spaced single particles were probed with immuno-gold on a lipid monolayer. As such, there was no background structure to serve as a reference. In the 162 immuno-gold experiment the gold particle counts per unit area (0.5 pm X 0.5 pm) were counted and averaged for both positive and negative controls for al1 three isomers of MBP

(Fig. 3-8).

The immuno-gold counts fiom Fig. 3-2 were analyzed by the Students t-Test

(Table 3-1). The test indicated that for dl three charge isomers the positive controls had significantly more labelling than the seven negative controis. Since four of the seven negative controls were buffer-S and the PSIG,, (4: 1) lipid monolayers, the particles that showed up without MBP did not significantly label with gold. Fig. 3-7 shows some close up views of the protein A gold particles bound the lipid monolayer. Due to the thickness of the carbon-plastic support it is difficult to see the single particles that the protein A gold has bound to. Some particles are faintly visible, but in row (c) for hMC8 one cm see sorne of the charactenstic three-fold particles that were unique to the MC8 preparation (Fig. 3-14). The irnmuno-gold data provided mersupport for the presence of MBP as a constituent in the single particles being investigated. When the gold particle counts for the three charge isorner positive control expenments were compared. the hMC8 treatrnent had more particles than hC1, which had more particles than hC8. This is the same pattern that was produced in the single particle counts for the three charge isomers on PS/GMl(4: 1) lipid monolayers (Fig. 3- 10).

The second stage of this investigation defined the interaction of MBP with PS/GM,

(41) lipid monolayers. A tirne course expenment was run where the different isomers were incubated with the lipid monolayer, as well as a control in which buffer-S alone was incubated with the monolayer. The number of single particles per unit area (0.3 pm X 163

0.3 pm) on the PS/GM,(4:l) lipid monolayers were counted in electron micrographs; Fig.

3-10 shows a plot of these results. Since the different isomer treatments showed different numbers of single particles per unit area it was determined that the appearance of these particles must be indicative of an interaction occurring on the PS/GM, (4:l) lipid monolayers with MBP.

In order to assess the significance of these differences the Students t-Test was applied to the data (Table 3-3). In Fig 3- 10 (a) the mean particle count per unit area is displayed, and each population has some degree of fluctuation over time. However. the data in most instances seem to fluctuate within a standard deviation of their neighbouring samples. In order to compare the populations. the data fiom each treatment were pooled. and in Fig 3-10 (b) the pooled data were plotted. Table 3-3 shows the results of the

Students t-Test analysis of this data. In al1 cases the addition of MBP produced significantly more particles per unit area (hC1 146.3 + 30.6; hC8 34.8 + 15.4: MC8

25 1.3 + 23 -4) than the buffer-S treatment (4.2 + 5.1). Secondly the hMC8 treatment had more particles than the hC1, which had more particles than the hC8. This data indicates that the different charge isomers of MBP each interact differently with lipids (Boggs er ai.. 1997; Cheifetz and Moscarello, 1985; Jones and Epand, 1980; Wood and Moscarello.

1989). Since these particle counts do not correlate with the net positive charge of these isomers (Cl = +18, C8 = +12, MC8 = +L), other factors than net positive charge must be responsible for these results.

In a recent investigation by Boggs et al. (1997) the interactions of different charge isomers (CLC8) of bovine MBP with acidic lipid vesicles were studied. In this 164 experirnent the MBP isomers were isolated using the same procedure that was used in this thesis, and different lipids, monovalent salt concentrations (KCI), and lipid-to-protein ratios were tested. They found that salt concentration affected the way the isomers bound the lipid, and the different isomers demonstrated different lipid aggregation abilities.

In the results presented in this thesis the isomers were incubated with a divalent salt (10 mM CaCIJ and a monovalent salt (150 mM NaCI). When comparing these results to those of Boggs et al. (1997) for the same monovalent salt concentration (150 mm, one sees a similar pattern. In both cases there is always a higher association of C 1 with the lipid compared to C8, with the other charge isomers taking up intermediate amounts of lipid. Boggs et al. (1997) postulated some possible reasons that could account for these differences, based on the post-translational modifications of MBP. The proposed mechanisms were: (i) reduced binding due to the reduced net positive charge. in this case the binding of MBP was considered to be solely due to isomer net charge. In the second mode1 (ii) a conformational change was also considered to be a probable factor involved in the interaction of MBP with the lipids. The mechanism which is electrostatic (i) was considered to be the most probable. The concurrence between their data with vesicles

(lipid bilayer) and this data (lipid monolayer) also indicates that MBP potentially interacrs with lipid bilayers and monolayers in a similar fashion. Since MBP is an extrinsic membrane protein, it is believed to reside on the surface of the myelin membrane. This may explain why MBP demonstrates sirnilar activity with lipid bilayers and monolayers.

Interestingly the unique hMC8 isomer did not behave in a rnanner which agrees with mechanism (i), since it showed higher single particle production, suggesting greater 165 association of protein with lipid. In the case of MC8 it has 10s 18 positive charges compared to the loss of 6 in hC8. therefore one wodd have predicted based on charge that hMC8 should have had lower single particle production. However the data indicate that MC8 may have possibly undergone a structural change (mechanism (ii)) which has increased lipid binding. This change in structure perhaps has led to an increase in exposed hydrophobic surface area, or charged groups which are made available to interact with the lipid due to a structural change, which is causing higher binding of hMC8 with the lipid monolayer.

3.5.2: 2D computer image analysis of TEM and STEM images of human myelin basic protein

The previous section demonstrated that the choice of charge isomer had a dramatic effect on the degree of single particle production. Irnmuno-gold electron microscopy further correlated the presence of the single particles with the presence of MBP.

Throughout that section the terminology MBP isomer vs. single particle was clearly delineated. The reason for this can be seen in Fig. 3-9, which showed that single particles were produced under al1 test conditions, including when buffer-S alone was incubated with the PS/GMl (4:l) lipid monolayer. In this section the single particles that were produced on the PS/GM, (4:1) lipid monolayer were stained with uranyl formate, and imaged by transmission electron microscopy (TEM) at liquid nitrogen temperature.

Analysis of the number of single particles per unit area of the TEM images indicated that the MBP treatments had significantiy more particles per unit area (hC1. 166 146.3 t 30.6; hC8 34.8 + 15.4; hMC8 251.3 i 23.4) than the buffer-S treatment (4.3 t

5.1 ) (Fig. 3- 10, Table 3-3). These data indicated that MBP has caused a drarnatic increase in single particle fiequency per unit area of 7-fold to 60-fold. depending on the charge isomer applied to the lipid monolayer. One of the goals of this investigation was the study of MBP with lipid monolayers, and the subsequent structural determination of MBP in a lipid environment where MBP has been show to adopt some a-helical and P-sheet secondary structure (Anthony and Moscarello, 1971; Deibler et al.. 1990: Keniry and

Smith, 1979, 198 1; Lees and Brostoff, 1984; Mendz et al., 1988. 1991 : Moore. 1982:

Rarnwani et al., 1989; Smith, 1977b; Stuart, 1996; Surewicz et al., 1987).

Four separate populations of TEM images of single particles were digitized and analyzed by 2D single particle cornputer image analysis (Frank. 1996: Haraur and

Boekema 1992; Yada et al., 1995). The four populations were composed of buffer-S single particles (Fig. 3- 1 1), and the single particles produced when hC 1 (Fig. 3- 12). hC8

(Fig. 3-13), and Fig. 3-14) interacted with lipid monolayers of PS/GM,(4:l). Visual inspection of the class averages indicates that the MBP treated particles tend to have unique cornmon structural similarities for each isomer. with distinct stain filled pores.

The buffer-S particles tend not to have such distinctive pools of stain, and there appears to be some size heterogeneity between class averages (e.g. Classes 2. 5, 22 vs. Classes 6.

7, 23, 25).

The hC 1 classes (Fig. 3-12) are composed of particles of approxirnately 7 - 10 nm in diameter. There appear to be three common structural groups, including the M class

(e.g. Class averages (CA'S); 13,43, 69, and 78) and C class (e.g. CA'S; 6, 8, 10, and 17). 167 both of which were also were present in the bovine C 1 data set (Fig. 2- 1 3). .A third class. the prolate class (e-g. CA'S; 7, 29, 31 and 41) also are visible in this data set. The hC8 class averages (Fig. 3-13) in general appear smaller then the hC 1 class averages. and the prolate classes tend to dominate this population (e.g. CA's: 4. 15. 16. and 78). and many of the prolate classes have a distinct knob at one end of the particle. There are still a few

C classes present in this population (e.g. CA'S 49, 54, and 99). Also, a new grouping of compact particles with apparent three-fold symmetry are visible (e-g. CA's 7. 13. 91. and

92). On average the small particles have a diameter of 7 nm. and the prolate particles have a minor axis of 7 nm, and a major axis of 8-10 nrn.

The hMC8 class averages differ dramatically fiom the hC I and hC8. In these data sets the apparent three-fold symmetric classes are predominant in the population. There are two primary groups of apparent three-fold syrnmetric class averages. One set is the small compact set (e.g. CA'S; 3, 5, 25, and 86) with one central pore surrounded by three stain excluding regions. The second group is larger and typified by three distinct stain filled pores in the particle (e.g. CA'S; 4, 6, 15, and 38). There are also a few prolate class averages in this population (e.g. CA's 9, 50, and 92). In general these particles are

9 - 11 nm in diameter, which make them the largest of the three data sets. When the class averages for the three MBP treated samples are compared to their averages there appears to be a good fit between class members, and their average surn (Figs. 5-1 5, 3- 16.

3- 17).

This structural cornparison has dernonstrated that the particles derived from the four different treatments differ from one another. The presence of single particles in the 168 bufTer-S and PS/GMl(4: 1) lipid monolayer, although of a lower titre, must be considered in this structural investigation. Bedu-Addo and Huang (1995) demonstrated that when

GMlinteracted with phosphatidylcholine (PC), depending on sdt concentration, the PC/GMl mixtures existed in several States ranging from a lamellar state to a micellar phase. In their investigation the micelles were of the same size as the single particles observed here.

Thus potentially the single particles are lipids fonning micelles. When MBP which is known to have a high affinity for G,, (Sela and Bach, 1984; Ong and Yu, 1984: Bach and Sela, 1985; Chan et al., 1990; Smith, 1992) interacts with the PS/GM, (4: 1) lipid monolayer, micelles may be formed. It is known that there are two specific sites on the

18.5 kDa isoform of MBP which bind GM, (Caamaiïo and Zand, 1989: Smith. 1992).

Boggs et al. (1977) used differential scanning calorimetry to show that MBP preferentially binds acidic phospholipids in vesicles, and that one molecule of MBP binds 34 molecules of PS. It has been demonstrated that charge isomer selection af3ects the degree of particle production (Fig. 3- 10). Therefore the single particles which are being observed probably have both a lipid and MBP component. The MBP component was identified by immuno- gold electron microscopy. The potential MBP vs. lipid component of this structure will be addressed in the last section of this discussion where 3D structures of these complexes are considered. The 2D structurai investigation has revealed that each population of single particles has a unique appearance depending on which charge isomer is added to the

PS/G,, (4: 1) lipid monolayer. This demonstrates that charge isomer selection is having an effect on particle structure. The three MBP isomers appear to behave differently when they interact with the PS/GMl(4: 1) lipid monolayer at the molecular level. 169

In a separate and independent STEM investigation the structures of hC 1. hC8. and hMC8 were studied when they interacted with the PS/GM,(4:l) lipid rnonolayer. In this investigation the particles were negatively stained with the neutral methylamine vanadate stain, and imaged at a dose of 17 eiectrons/A2 at liquid nitrogen temperature (Fig. 3-28).

These single particles were then analyzed by 2D image analysis as was done with the

TEM images (Figs. 3-29, 3-30, 3-31). The STEM averages were different from the TEM averages. The TEM images had distinct domains of stain versus stain excluding areas. which sometimes had a puddle or blob-like appearance. The STEM averages on the whole tend to have a stippled granular appearance, and appear to be more solid than the

TEM images (i.e., no nain filled pores in the center of the particle). Although distinct

M, C, prolate and three-fold classes are not readily apparent in this data there are similarities to the TEM results. The entire population the hC 1 averages (Fig. 3-29) appear circular with a little elongation the hC8 averages (Fig. 3-30) tend to be mainly prolate. and many of the hMC8 averages (Fig. 3-30) have a triangular appearance. These results agree quite well with the TEM results where the hC1 averages were pnmarily C or M shaped, the hC8 were mainly prolate, and the MC8 were mainly trianguladthree-fold symrneûical. As was the case for the TEM data the class averages of the STEM data match up favourably with the individual class members for each set (Figs. 3-32. 3-33, and

3-34). When Fourier ring correlation was used to estimate the resolution of the data sets it was found that the average resolution of the TEM images was 1.4 nm (hC1 = 0.9 nrn; hC8 = 1.6 nm; hMC8 = 1.7 nm, (Fig. 3-18)). The STEM images had an average estimated resolution of 0.9 nrn @Cl = 0.9 nm; hC8 = 0.9 nrn; hMC8 = 1.0 nm, (Fig. 3- 170

35))-

One of the probable reasons for the difference between the TEM data and the

STEM data lies in the contrasting agents used in the analysis, and their application to a protein-lipid cornplex. The uranyl formate staui is ionic, and thus it exists as a multitude of charged species (positive and negative), which exhibit preferential staining of phosphate, carboxyl, and hydroxyl groups (Harris. 1997; Miller et al., 1990). As such. one mut consider that the preferential staining will affect al1 the results. In the case of preferential staining, the macromolecule being stained will have common areas of stain pooling and stain exclusion.

When the TEM images are compared to the non-ionic methylamine vanadate contrasted STEM averages, it is evident that the overall particle morphologies are conserved (hC 1 circular, hC8 prolate. hMC8 triangular). but the fine structural details Vary considerably. When structures are compared, especially in 3D. one must consider the effect of stain on the structure being investigated. The benefit of comparing two physically different stains is that it will provide insight into the location of the preferentially stained charged groups which an ionic stain will emphasize compared to a non-ionic stain, which will not.

Finally. one mut always consider that the stained structure visualised in a heavy metal stained preparation reflects primarily a surface metal cast of the structure. Due to incomplete stain penetration, interpretation of the interior structure may not retlect the intemal mass distribution within the biological sample. The intemal mass only reflects the degree and intensity of stain penetration in the particle. This stain penetration will 171 have implications in the interpretation of a 3D reconstruction of stained images.

In the previous section the differential single particle counts were addressed. and it was proposed that (i) reduced binduig due to the reduced net positive charge. and (ii) conformational changes, could account for what was observed. Biochernical data

(Moscarello et al., 1994; Wood and Moscarello. 1996; Wood et al.. 1996) which has identified charge differences between the isomers of MBP can account for proposal (i).

The 2D data presented in this section indicates that there is a structural alteration in the

MBP-Iipid particles being studied, thus providing evidence for proposai (ii). The final section of this thesis takes the 2D stnicturd data sets and computes 3D reconstructions for the MBP-lipid complexes that were imaged by STEM and TEM.

3.5.3: 3D cornputer image analysis of TEM and STEM images of hurnan myelin basic protein

The final study presented in this thesis was the angular reconstitution of the ZD data sets for the TEM and STEM data using IQAD (Farrow and Onensmeyer. 1992.

1993). The reconstructions for the TEM data (Figs. 3-19,340. and 3-21, 3-27)1 and the angular assignments of their 2D data sets (Fig. 3-22) show these reconstructions to have a hollow crucible type shape, with a cap on top. The MC8 reconstruction appears to be the most extended and the hC8 reconstruction appears to be the most compact. When viewed fiom the top or bottom the hC1 (Fig. 3-19 (e,f)) and hMC8 reconstructions (Fig.

3-21 (e,f)) appear to show some degree of three-fold syrnmetry. For the hC8 data three- fold syrnmetry dong this mis was not apparent (Fig. 3-20 (e,f). The hMC8 shows the 1 73 highest degree of apparent three-fold symmetry, which is in good agreement with the 2D data (Fig. 3-14) which had similar potential symmetry. The hC8 reconstruction had a definite long axis when viewed from the side (Fig. 3-20 (a-d)). compared to when it was viewed from the top or bottom (Fig. 3-20 (e,f)). This is in agreement with the particle having prolate (side), and compact three-fold (top) views.

Figures 3-23, 3-24, and 3-25 show exact cornparisons of the initial 2D images compared to fonvard-projections and shaded sudace representations as viewed from the exact Euler angles that were assigned to the individual 2D images (Fig. 3-22). When the

initial averages are compared to the forward projections and surface representations. the

3D data appear to be smaller; this is a result of the back-projection process. In al1 cases the fonvard projections appear somewhat blob-like. with a low density mass in the interior of the reconstruction. One of the primary reasons for the reconstruction looking this way

was that the uranyl formate ZD images appeared to have heavy stain pooling on the

surface. When reconstnicted there was little apparent interior mass signal. thus the

forward-projections have a hollow bal1 appearance.

The 3D reconstructions from the 2D STEM data produced slightly different results

(Figs. 3-36, 3-37, 3-38, and 3-44). The STEM reconstructions al1 have a mottled surface

cornpared to the relatively smooth surface appearance of the TEM reconstructions (Figs

3- 19, 3- 10, 3-21). The major structural difference between the two reconstructions is that

the STEM reconstructions had a solid intenor (Fig. 3-44). As was the case for the TEM

data, the MC8 reconstruction was the most extended, the hC1 was intermediate, and the

hC8 was the most compact. When the Euler angle assignment (Fig. 3-39) was used to 173 compare the 2D data to the fonvard-projections and shaded surface representations (Figs.

3-40, 3-41, and 3-42) a better correlation was found for the STEM data. In these cases although the forward-projections look more compact, they share a closer resernblance to their individual 2D input images. This correlation is the best for the hMC8 data where the distinct triangular shape makes many of the 2D and fonvard-projections compare

favourably.

When the resolution of the 3D reconstructions were estimated by the Fourier shell correlation method, the resolutions were similar for the TEM (hC1 = 2.2 nm; hC8 = 2.8 nm; hMC8 = 2.1 nrn, (Fig. 3-26)) and STEM (hC1 = 2.4 nm: hC8 = 2.5 nrn: hMC8 = 2.4

nm, (Fig. 3-43)) reconstructions. Since the images were of the sarne approximate resolution it appears that the stain must be responsible for the difference in the interior of the reconstructions.

In section 3.2 of this chapter where negative stain properties were discussed stain density was mentioned as a crucial factor to consider. Protein has a density of 0.84 D&"

(Stark et al., 1995), whereas uranyl formate has a density of 2.2 DA-'(Hayat and Miller.

1WO), and methylarnine vanadate has a density of approximately 1.O DA''(personal communication with J. Hainfeld). Based on the relative densities of the stains compared

to that of protein, it is obvious that the uranyl formate electron images will be biased by

uranyl formate structural information. Since the methylamine vanadate and protein have

fairly similar densities, the electron images for these samples will have more protein-

produced contrast, compared to the uranyl formate images. In the 3D reconstructions

presented here it appears that the uranium has potentially produced the "hollow" core as 1 74 an artifact of staining.

A direct comparison of the TEM and STEM reconstructions (Fig. 3-45) shows the

STEM reconstnictions as a solid surface paired with the TEM reconstructions overlaid as a wire mesh. In al1 cases the hC1 (Fig. 3-45 (qb)), hC8 (Fig. 3-45 (c-d)). and hMC8

(Fig. 3-45 (e,f)) reconstructions show a close fit with their paired reconstruction.

The hC1 reconstructions (Fig. 3-45 (ab)) show the closest fit. with the envelopes of both reconstn>ctions in close proximity to one another. Both hC8 reconstructions (Fig.

3-45 (c,d)) are the most compact: the lower sections of the reconstruction show a tight fit

(Fig. 3-45 (c)). The top half of the subunits differ due to the prominent cap present on the TEM reconstruction, which is absent on the STEM reconstruction. This could indicate that potentially there are charged species present (phosphate. carboxyl. and hydroxyl groups (Harris, 1997; Miller, 1990)) at this location in the hC8 complex. and the ionic uranyl formate stain is preferentially staining this region of the complex. In the methylamine vanadate stained STEM image no preferential staining would occur. The

MC8 reconstructions (Fig. 3-45 (e,£))show a reasonable fit. with a large diameter in the upper half of the reconstruction, and a fairly flat base at the bottom of the reconstruction

(Fig. 3-45 (e)). In both reconstructions there is a prominent cap; in the TEM reconstruction it is attached to the reconstniction, and in the STEM reconstruction the cap appears detached.

In general, this direct comparison of the outer surfaces has shown that both the

TEM and STEM reconstructions compare favourably as representations of the MBP-lipid complexes being studied. Figs. 3-46, 3-47, 3-48, 3-49, 3-50, and 3-51 are used to Figure 3-45: Overlaying TEM and STEM reconstmctions. The TEM (mesh) and

STEM (solid) reconstructions have been overlaid here to show the similarities between the reconstructions. In this figure the Cl (a,b). C8 (c,d), and MC8 (e,f) reconstructions are shown from side (a,c,e) and top (b,d,f) views. The reconstructions are shown here in stereo to provide 3D visualization. From these images the TEM and STEM reconstnictions compare quite favorably, with the greatest variance between the 3D reconstructions being in the "cap" region. Figure 3-46: Probing the internai structure of the hurnan CI TEM reconstmction.

The internai mass distribution of the human Cl TEM reconstruction (Fig. 3-19) is illustrated here. In (a) and (b) the reconstmction is shown in stereo at an oblique angle. in (c-e), the reconstruction is shown frorn the top, side, and bottom views. respectively.

In the above figures the reconstruction has a high density threshold (the shaded surface in

(a-e)) to show the interior structural elements. In figure (a) the exterior surface of the reconstruction is included as a wire mesh frame around the solid interior structures. In

(c-e) the cap (c). annular ring (ar), and base (b) have been labeled. Figure 3-47: Probing the interna1 structure of the human Cl STEM reconstruction.

The intemal mass distribution of the human Cl STEM reconstruction (Fig. 3-36) is illustrated here. In (a) and (b) the reconstruction is shown in stereo at an oblique angle. in (c-e), the reconstruction is shown from the top, side, and bottom views. respectively.

In the above figures the reconstniction has a high density threshold (the shaded surface in

(a-e)) to show the intenor stmcturai elements. In figure (a) the exterior surface of the reconstruction is included as a wire mesh frame around the solid interior structures. In

(c-e)the cap (c), annular ring (ar), and base (b) have been labeled. Figure 3-48: Probing the internal structure of the human C8 TEM reconstruction.

The internal mass distribution of the human CS TEM reconstruction (Fig. 3-20) is illustrated here. In (a) and (b) the reconstruction is shown in stereo at an oblique angle. in

(c-e). the recnnstruction is shown from the top, side, and bottom views. respectively. In the above figures the reconstruction has a high density threshold (the shaded surface in

(a-e)) to show the interior structural elements. In figure (a) the exterior surface of the reconstruction is included as a wire rnesh frame around the solid interior structures. In (c- e) the cap (c),annular ring (ar), base (b), and pore (p) have been labeled . Figure 3-49: Probing the interna1 structure of the human C8 STEM reconstruction.

The intemal mass distribution of the human CS STEM reconstruction (Fig. 3-37) is

illustrated here. In (a) and (b) the reconstruction is shown in stereo at an oblique angle. in

(c-e), the reconstruction is shown from the top, side, and bottom views. respectively. In the above figures the reconstruction has a high density threshold (the shaded surface in

(a-e)) to show the interior structural elements. In figure (a) the exterior surface of the reconstruction is included as a wire mesh frame around the solid interior structures. In (c- e) the annula ring (ar), base (b), and pore (p) have been labeled. Figure 3-50: Probing the internal structure of the human MC8 TEM reconstruction. The internal mass distribution of the human MC8 TEM reconstruction

(Fig. 3-21) is illustrated here. In (a) and (b) the reconstruction is shown in stereo at an oblique angle, in (c-e), the reconstruction is shown from the top, side, and bottom views. respectively. In the above figures the reconstruction has a high density threshold (the shaded surface in (a-e)) to show the intenor structural elements. In figure (a) the exterior surface of the reconstruction is included as a wire mesh frame around the solid interior structures. In (c-e) the cap (c), annular ring (ar), base (b), and pore (p) have been labeled. Figure 3-51: Probing the internal stmcture of the human MC8 STEM reconstruction. The internal mass distribution of the human MC8 STEM reconstruction

(Fig. 3-38) is illustrated here. In (a) and (b) the reconstruction is shown in stereo at an oblique angle, in (c-e), the reconstruction is shown from the top, side. and botroni views. respectively. In the above figures the reconstruction has a high density threshold (the shaded surface in (a-e)) to show the interior structural elements. In figure (a) the exterior surface of the reconstruction is included as a wire mesh frame around the solid interior structures. In (c-e) the cap (c), annular ring (ar), base (b), and pore (p) have been labeled. 182 facilitate this comparison. and Mer probe these reconstructions in search of any common structural motifs. In al1 six figures the reconstructions were shown with the outer envelope of the reconstruction represented as a wire mesh in (a). The interior densities of the reconstnictions were illustrated as a solid surface (a-e). Only one interior surface was illustrated, this was generated by selecting a higher density threshold than was used for the exterior surface. At first, a comparison of the TEM reconstructions (Figs.

3-46, 3-48, and 3-50) to the STEM reconstructions (Figs. 3-47. 4-49, and 3-51) revealed that the TEM reconstructions at a high threshold appear to look like "plates" surrounding a transparent intenor. The STEM reconstructions were the opposite when viewed at a high threshold. they have a dense interior filled with intncate protuberances and channels.

As discussed above the probable reason for these discrepancies lies in the differences in the contrasting stains used, compared to the sarnple being investigated. The results presented here show that the methylamine vanadate data have a less stain biased representation of the reconstruction's interior, whereas the uranyl formate reconstructions tend to be stain biased, and represent a metal cast of the exterior of the reconstruction.

Therefore, the methylamine vanadate stained reconstructions show a closer representation of the biological structure under investigation.

Even though the uranyl formate stained reconstructions are stain biased. this bias is extremely usefùl in the interpretation of the structure. Since the uranyl formate stain has specific staining for charged groups, the repetitive staining of these groups helped to identiQ common structures. The most obvious example in this investigation can be seen in the 2D analysis of the hMC8 single particles which showed a strong tendency for 183 apparent tbree-fold symmetry, especially in the uranyl formate images (Fig. 3-14).

When the hC1 reconstructions viewed with a high threshold fkorn the TEM (Fig.

3-46), and STEM (3-47) data are compared several features are apparent. Both reconstructions have three structures, referred to as the cap, annular ring, and base. In figure (c) of both figures (Figs. 3-46, and 3-47) there appears to be some degree of three- fold syrnmetry down the vertical axis of the reconstruction.

When the hC8 reconstructions viewed with a high density threshold from the TEM

(Fig. 3-48), and STEM (3-49) data are compared the cap annular ring, and base are present, with the base having a pore in it. However, both reconstructions do not have al1 these features. The TEM reconstruction has a prominent cap, and weak annular ring. whereas the STEM reconstruction has no cap at al1 but a prominent ring. Both reconstructions have a base with a pore in it. In figure (c) of Figs. 3-48. and 3-49 there is no apparent three-fold syrnrnetry in either reconstruction, although the TEM reconstruction has four dominant lobes.

When the MC8 reconstructions viewed with a high threshold frorn the TEM (Fig.

3-50), and STEM (3-51) data are compared both reconstructions have a distinct cap. annular ring, and base, which has a prominent pore. The TEM reconstruction shows some degree of three-fold symmetry in Fig. 3-50 (c); this syrnrnetry is even more apparent in

Fig. 3-27 (c) where in the last column the hC1 TEM reconstruction is shown fiom the same vantage point, at a different threshold. The STEM reconstruction in Fig. 3-51 (c) shows potential three-fold symmetry, but this is weak.

As previously mentioned in this thesis, the data indicate that the single particles 184 that were imaged when the charge isomers of MBP interacted with the PSIG,, (4: 1) lipid monolayer are composed of both lipid and MBP. The structural investigation of these single particles has shown that when hC 1, hC8, and hMC8 interact with the PSIG,,, (4:1) lipid monolayer three distinct particles are produced. This indicates that the three charge isomers upon interaction with the lipid monolayer have unique structural features. The primary debate then becomes, to what extent are these particles composed of lipid vs.

MBP. The fust mode1 (i) is that the particle is primarily Iipid (PWG,,). associated with a single molecule of MBP. The second model (ii) is that the particle is composed primarily of MBP; in this case the lipid is a minor constituent. and the MBP is an oligomer. If one considers only the TEM data the particle is hollow. and thus an oligomer seems unlikely, and the model (i) is favoured. However since the STEM data rnust also be considered, and since it is a better representation of the particle. model (ii) cannot be ruled out. The apparent three-fold symmetry in the data supports mode1 (ii). since symmetry is a characteristic of an oligomer.

Studies on MBP association have indicated that MBP self-associates at pH values above 6. MBP also has a wide range of interactions with arnphipathic and hydrophobie compounds (Smith, 1992). MBP has been shown to lose some of its random coi1 structure derinteracting with the above compounds, as well as derphosphorylation (Anthony and

Moscarello, 1971 ; Deibler et al., 1990; Keniry and Smith, 1979, 198 1; Lees and Brostoff.

1984; Mendz et al., 1988, 1991; Moore, 1982; Ramwani et al., 1989; Smith. 1977b;

Stuart, 1996; Surewicz et al., 1987), adopting some a-helical and P-sheet structure. Two primary complexes are proposed in the literature for MBP self-association. The first is 185 that MBP forms dimers (Braun, 1977; Smith, 1977a). These dimers are believed to help compact the cytoplasmic surfaces of the myelin membrane. In investigations by Smith

(1985) and Gow and Smith ( 1989) when purified MBP was interacted with myristoyl- lysophosphatidyIcholine (1 ysoDMPC) and analyzed by sedimentation equilibriurn (SE) and circular dichroism (CD) several key results were found. MBP was found to exist in a monomer-hexamer equilibriurn, without significant accumulation of intermediates. In earlier studies (Smith, 1980, 1982) a weak association of dimers was found. but in the

1985 study by Smith there was no support for the dimer data. The initiai CD data indicated that upon hexarner formation MBP developed P-pleated sheet conformation. however these results were later not deemed repeatable (Gow and Smith. 1989). Ln the first study (Smith, 1985) the MBP was isolated under the same conditions as in this thesis.

Although the pnmary study was with Cl. the other charge isomers of MBP were reported to exhibit the same monomer-hexamer equilibrium as C 1. In the Gow and Smith study

(1989) MBP was purified by both the denaturing method as used previously, and a procedure that avoids denaturants (by this method the charge isomers cannot be separated). This study found that under both conditions the MBP could exist in the monorner-hexarner equilibriurn, and that both the structures in aqueous solution existed as a random coil.

When the structural data presented in this thesis are compared to the data of Smith

(1985), the three-fold symmetry that is seen in the data could be interpreted potentially as a hexamer. This information supports model (ii), however model nurnber (i) can not be ded out based on the potential correlation with symmetry. The lipid monolayer 186 system in which MBP was imaged was dynarnic, and thus al1 possible results must be considered. Boggs et al. (1977) demonstrated that 34 molecules of PS could bind one molecule of MBP in a lipid vesicle. Based on this data a rnonomer of MBP (18.5 kDa) bound with 34 PS molecules would have a molecular weight of 45.3 kDa and a hexarner of MBP (1 1 1 kDa) would have a molecular weight of 271 -5 kDa. This second value is unlikely since a hexarner will probably not bind as many PS molecules as a monomer.

When al1 the 2D and 3D data are considered for the complexes that were produced when the hC1. hC8 and hMC8 interacted with the PS/G,, (4: 1) iipid monolayer. three consensus models were produced which reflect an amalgarnation of the data. This model. shown in Fig. 3-52, is composed of the primary structural units referred to as the cap. annular ring, base, and pore. The annular ring is displayed as a set of six spheres indicative of the potential hexameric structure.

Figures 3-53, 3-54 and 3-55 show the hC1, hC8, and hMC8 reconstructions, respectively with no symrnetry (a), imposed C3 symrnetry (b), and imposed C6 symrnetry

(c). The C3 and C6 symrnetry are the two possible ways in which a hexamer could arrange itself symmetrically within the constraints of the existing EM data (i.e. radial syrnmetry dong the long axis of the reconstructions). The reconstruction without symmetry is included for cornparison, and to emphasize that the hexamer model is one of two possible proposed structures. Based on molecular weight alone one cm predict the size that a specific protein will have, assuming an average protein density of 0.8 D~/A'. and that the protein will be a perfect sphere (Henderson, 1995). Based on these assumptions a monomer of MBP (1 8.5 kDa) would have a diameter of 3.5 nm, and Figure 3-52: Mode1 of human myelin basic protein complex. Consensus models of the

MBP complexes are proposed here for the TEM and STEM structural data for hC 1 (a), hC8

(b), and hMC8 (c). The left colurnn shows the models from the top view. and the nght colurnn shows the models from the side view. The prorninent structures shown here are the cap, annular ring, base, and pore. These are the most conserved structural features found in the six 3D reconstructions of human MBP from the TEM and STEM data. in (a), (b), and (c) the cap, annular ring, and base structures are extended or contracted depending on the charge isomer. The base is the proposed site for the P superbarrel (Fig. 3-56). [n (b) the left column has two figures to show this perspective with and without a cap. as was the case for the TEM and STEM reconstructions, respectively. In (c) the annular ring has been rnodified to show the prominent three-fold symmetry of hMC8. Annular Ring

Base

Annular Ring Pore

Base

Annular Ring Pore

Base Figure 3-53: Syrnrnetrised human Cl STEM reconstructions. The hC1 STEM reconstruction is shown here with no symmetry (a), C3 symmetry (b), and with C6 symmetry (c). In each figure (a-c) the reconstruction is shown from the top, at tilted angle and from the side. The reconstructions are shown here with a density threshold which emphasizes boih the symmetry, and the cap. annular ring and base structures. C

Figure 344: Symmetrised human C8 STEM reconstructions. The hC8 STEM reconstruction is shown here with no symmetry (a), C3 symmetry (b), and with C6 symmetry (c). In each figure (a-c) the reconstmction is shown from the top, at tiIted angle and from the side. The reconstructions are shown here with a density threshold which emphasizes both the symmetry, and the annular ring and base structures. Figure 3-55: Symmetrised human MC8 STEM reconstructions. The hMC8 STEM reconstruction is shown here with no symmetry (a), C3 symmetry (b), and with C6 symmetry (c). In each figure (a-c) the reconstruction is shown from the top, at tilted angle and from the side. The reconstructions are shown here with a density threshold which emphasizes both the symmetry, and the cap, annula ring and base structures. 192 a hexamer (1 1L kDa) would have a diarneter of 6.5 nm. The STEM reconstructions are considered to be more representative of the actual volumes of the MBP complexes. although they are slightly bigger than a 6.5 nrn sphere. This increase in size is probably due to the lipid component in these complexes (Boggs et al.. 1977; Caamaiio and Zand.

1989: Smith. 1992). Based on this one can conclude that a hexamer could fit into the volumes reconstructed. However, a single monorner associated with lipid could also fit: again neither of the structural models cm be ruled out.

As mentioned above, MBP is pnmarily random coi1 in solution, and in association with lipid develops some a-helical and P-sheet structure. One possible argument against the hexamer mode1 could be that if MBP has secondary structure when it associates with lipid, then perhaps it could not fit in the reconstructions presented in this thesis. Although the secondq structure of MBP is unknown, several predictions of secondary structure based on amino acid sequence have been made (Martenson, 198 1. 1986: Stoner. 1984.

1990). The models of Stoner (1 984, 1990) are simple ones, and Martenson (198 1. 1986) incorporated potential 3D folds in the B-structure. Stoner (1984) predicted that MBP would be composed of an antipardlel J3-sheet, with five strands. along with two a-helices

(Fig. 3-56 (a)). Since the tertiary structure of MBP is not known. yet alone that of a hexamer, the Brookhaven Protein Data Bank (PDB) which contains 3D X-ray crystailographic structures was searched. The data base was searched for hexameric antiparaIle1 P-sheet structures, and the structure of neuraminidase solved to 2.2 A resolution fitted this description (PDB accession number l nn2; Varghese and Colman.

1991), since it has a motif known as a superbarrel. In Fig. 3-56 the entire molecule (b). Figure 3-56: Neuraminidase and MBP, possible structural correlations The predicted secondary structure of MBP is shown (Stoner. 1984) in (a), and this structure is compared to the 3D structure of neurarninidase solved to 0.22 nrn (Brookhaven Protein

Data Bank accession number lnn2; Varghese and Colman, 199 1). The antiparallel B- sheets are being compared. to demonstrate the potential hexameric packing of MBP. In

(b) the entire neuraminidase molecule is shown, and in (c) only the antiparellel P-sheets are shown. This structure has six sets of P-sheets which fonn a motif called a superbarrel.

In (b) and (c) the three fold symmetry of this structure is identified; this symmetry was also seen in the electron images along the sarne axis of the reconstructions. In (d) the hC8 reconstruction has been superimposed over the neuraminidase X-ray structure to demonstrate the fit of these two structures. In (b-d) the left column shows the structures from the side, and the right column shows the structures from the top. 194 and the superbarrel (c) alone are shown; also the hC8 reconstruction is shown supenmposed over the superbarrel (d). When viewed from the top (end on view of barrel) in (b) and (c) the barrel has a triangular appearance that is similar to the viangular

(three-fold symmetric) shapes found in the EM data (especially the hMC8). When the superbarrel was fitted in 3D with the syrnmetrised hC8 reconstruction it was found that the two fitted together favourabiy. Based on this fit, one cannot mle out potential hexamer packing within the 3D structures presented in this thesis.

In summary the 3D reconstructions presented in this thesis show distinct differences between the hC1, hC8, and hMC8 complexes that were produced when they were incubated with a lipid monolayer of PS/GM,(4: 1). At present there are wo models for the composition of these structures. either (i) a single molecule of MBP associated with a major lipid component, or (ii) a hexamer of MBP associated with a minor lipid component. The three-fold symmetry that appears in some of the data sets makes model

(ii) the favoured prediction, but this evidence is not conclusive enough to rule out model

(0-

The data presented in this thesis have demonstrated that the different charge isomers of purified MBP exhibited different behaviour in the in vitro mode1 system that was used. The hMC8 isomer produced the greatest number of single particles, and these particles produced the largest 3D reconstructions. The hC8 isomer produced the smallest number of single particles, and these particles produced the smallest 3D reconstructions.

In both cases hC1 produced intermediate results. This indicates that the MBP isorners which predominate in chronic MS (C8K1 ratio = 2.5) and acute Marburg type MS 195

(MCIICI ratio = 6.7), behave differentiy than Cl which predominates in normal

individuals (C8/C1 = 0.82). Based on net charge alone (hC 1 +19. hC8 +12. hMC8 + 1 ) one would not anticipate this result (one would predict MC8 to interact less then C8 with the lipid monolayers). It appears that the structural differences observed in this study may be accountable for these differences.

Chronic MS is a slow debilitating disease often charactensed by initial periods of remission and relapse. Marburg type MS is very rare and behaves totally differently. It has a rapid onset in apparently normal individuals, and death often occurs within weeks of diagnosis. In Marburg type MS individuals a unique isomer of MBP (18.5 kDa) was recently discovered (Wood et al., 1996), that has never been found in nomat or chronic

MS individuals.

MBP is an extrinsic membrane protein, which is believed to be responsible for the compaction of the myelin sheath. This in vitro investigation of MBP has demonstrated that hC8 and hMC8 behave differently than Cl with a PS/GM, (4:l) lipid monolayer.

Based on this difference one can postulate that any variation in structure and behaviour of MBP could have a disniptive effect on the myelin sheath. hC8 and hMC8 behave in an opposite fashion to one another. These differences, coupled with the relative amounts of each isomer present in chronic MS and MS/Marburg individuals cornpared to Cl suggest that these two diseases of myelin have different effects on MBP and myelin disruption.

In Fig. 3-57 the three STEM reconstructions of MBP are shown fitted into a scale mode1 of the myelin sheath. In this figure two possible models are presented for the 196 fitting of the reconstructions into the major dense line (MDL) of the myelin sheath. In model (i) the reconstructions are shown in an side-on orientation in which the reconstructions align their long axis parallel to the lipid bilayer. In model (ii) the reconstructions are presented in an end-on orientation with their long axis perpendicular to the lipid bilayer. In this investigation MBP was imaged in association with a lipid monolayer; as such it is not possible to state exactly how these reconstructions would pack in the space between two lipid bilayers in the myelin sheath. The two models presented here are possible interpretations of how the MBP reconstructions may fit into the myelin sheath. In both models it is apparent that some degree of bilayer penetration will occur for the reconstructed MBP-lipid complexes. As previously stated in this thesis. MBP is believed to be responsible for myelin compaction. The tight fit i~.the myelin sheath presented here for the three different reconstructions (hC 1, hC8, and hMC8), coupled with their structural variability may have a putative effect on myelin compaction and stability. MDL

IPL

MDL

Figure 3-57: Fitting MBP reconstructions into the myelin sheath. The STEM

reconstmctions of hC1 (a), hC8 (b), and hMC8 are shown here fitted into a 3D mode1 of

the myelin sheath with the major dense line (MDL), and intraperiod line (IPL) indicated.

In this figure two possible interpretations of how MBP fits into the myelin sheath are

shown. In the upper MDL the reconstmctions have been placed in a side-on orientation, and in the lower MDL the reconstructions have been placed in the end-on orientation. In both cases the reconstructions have some degree of penetration into the lipid bilayer. 198

3.6: Conclusions

Myelin basic protein (MBP) is an extrinsic membrane protein that exists as several different post-translationally modified charge isomers. Individuals with multiple sclerosis

(MS) have a decrease in the amount of cationic MBP. In this chapter the Cl. C8 and

MC8 charge isomers of MBP were investigated by TEM and STEM interacting with a

PS/GM,(4:l) lipid monolayer. These studies showed that the CS isomer produced the lowest arnount of single particles, Cl displayed an intermediate amount. and MC8 produced the highest arnount of single particles. Imrnuno-gold electron microscopy identified MBP as a component of the single particles. The 3D structures of the MBP- lipid complexes were solved to approximately 2.5 nm resolution by computer image analysis. C8 produced the most compact structure, while Cl produced an intermediate structure, and MC8 produced the most extended structure, which showed signs of apparent three-fold symrrietry. Al1 three reconstructions shared three cornrnon structural elements refe~~edto as the cap, annular ring, and base. However. depending on the stain used different structures were enhanced. If one interprets Cl as the "normal" structure. then

C8 appears to be a contracted version of the Cl structure, and MC8 appears to be a relaxed fonn of the structure.

The various reconstructions have a diameter of 8-10 nm; this size indicates that the complex being reconstmcted is composed of more than one MBP monomer of 18.5 ma. There are two possible interpretations of the data, either the reconstruction is composed of one monomer of MBP with a large lipid component, or the reconstruction is an oligomer of MBP with a minor lipid component. The data indicate that there is 199 some degree of three-fold syrnmeûy in particles being reconstructed. Since syrnmetry tends to be a feature of oligomers, the data favour the second hypothesis. It is speculated that the complex may be composed of a hexarner of MBP associated with the PS/GM,

(4: 1) lipid monolayer.

In conclusion this study has demonstrated that the different charge isomers of MBP interact with lipids, and are stmcturdly different fiom each other. when they associate with PS/GM, (4:l) lipid monolayers. Since this is an in vitro study of MBP which was isolated under conditions that do not favour the retention of secondary structure. one can not make a direct cornparison to native MBP in the myelin sheath. Based on the results presented in this chapter one cmtheorize the following. Since MBP is believed to be a structural protein in the myelin sheath, the data presented indicates that the different charge isomers may have a disruptive effect on myelin structure. CHAPTER 4

THESIS CONCLUSION

This thesis has been an investigation of the behaviour of isolated charge isomers of the extrinsic membrane protein myelin basic protein (MBP) by electron microscopy.

MBP is one of the major proteins of the myelin sheath. Presently, there is no known biochemical function for MBP. It is believed that MBP plays a fundamental structural role in the maintenance of the compaction of the rnyelin sheath.

Multiple sclerosis (MS) is a disease of the nervous system, in which the rnyelin sheath is degraded. Environmental and genetic factors are believed to be involved in MS. however, to date the cause of MS is unknown. It has been found that the MBP in MS individuals differs from that of healthy individuals by an abundance of a citnillinated form of MBP (C8). In healthy individuals the Cl isomer is the predominant species. and C8 is still present (CK8 ratio = 0.82), but not in the quantity seen in MS (ClK8 ratio =

2.5) (Wood et al., 1996). Although it is not known if these modifications are the cause of MS, or a secondary effect of the disease, they are still associated with the developmenr of MS. It is believed that the modifications of MBP lead to the inability of MBP to compact the lipid bilayers of the myelin sheath.

In this thesis the purified Cl and C8 charge isorners were studied, as well as the unique highly citnillinated form of MBP (MC8) found in acute Marburg type MS

(MC8KI ratio 6.7), (Wood et al., 1996). In order to isolate MBP, an acid extraction and column chromatography in 2M urea were used. The use of such conditions is not 20 1 conducive to the retention of ordered structure, however this is the only way by which the charge isomers can be separated. Since the modified charge isomers appear to play a crucial role in MS, they had to be isolated by this method. It has been found that isolated MBP which has a random coi1 structure in solution takes up some a-helical and

B-sheet structure when it interacts with various lipids and detergents.

In this investigation MBP in association with lipid monolayers of phosphatidylserine and monosialoganglioside GM,(PS/GMi) (4: 1) were studied in order to see how the different isomers behaved. It was found that the C8 isomer produced the lowest activity of single particle production when it interacted with the PS/GM,(4: 1) lipid monolayers. CI displayed an intermediate interaction, and MC8 demonstrated the highest activity. Immuno-gold electron microscopy identified MBP as a component of the single particles. Computer image analysis of the single particles produced 2D averages which were used as input to determine the 3D structures by angular reconstitution solved to approximately 2.5 nm resolution. This anaiysis determined that C8 produced the most compact structure. Cl produced an intermediate structure, and MC8 produced the most extended structure, which showed signs of apparent three-fold syrnmetry. Al1 three reconstructions had three cornmon structurai elements referred to as the cap. annular ring, and base. The reconstructions based on the three isomers interacting with the PS/GMi

(4: 1) lipid monolayer appear to be variations of a common structure. If one interprets C 1 as the "normal" structure, then C8 appears to be a contracted version of the Cl structure, and MC8 appears to be a relaxed form of the structure.

The reconstructions presented in this thesis are too large to be of a single monomer 202 of MBP aione. There are two possible interpretations of the structural results. Either the

reconstructions are a single MBP associated with major lipid cornponent, or a hexamer

of MBP associated with a minor lipid component. MBP has been shown to exist in equilibrium between a monomers and hexamers (Gow and Smith, 1989; Smith. 1985.

1992), thus supporting both interpretations. The three-fold symrnetry that was detected

in the MC8 data favours the second interpretation. since symrnetry tends to be indicative of an oligomer. The three-fold symmetry couId indicate C3 symrnetry in a hexarnenc cornplex.

In conclusion this study has demonstrated that the C8 and MC8 charge isomers of

MBP are stmcturally different from Cl when they interact with a PS/GM, (4:l) lipid

monolayer. In vivo, myelin has a ordered structure composed of compacted lipid

membranes (bilayer). The structures presented here were produced in vitro with purified

MBP on simple lipid monolayers, therefore the conditions Vary considerably to that of native myelin. Myelin has an ordered compacted structure, and any variations in its structural components could putatively disrupt this membranous structure. Sine MBP is believed to be an extrinsic structural membrane protein, the results presented here provide putative evidence that the modifications to MBP may have a potential dismptive effect on myelin structure. 203 FUTURE RESEARCH

This investigation has produced information on the behaviour of different charge isomers of MBP interaction with PS/GM, (4:l) lipid monolayers. Based on these encouraging results several new issues cm now be addressed in future research: (1 ) the

Merstructural investigation of the particles, (2) the interaction of MBP with other lipid monolayers, (3) the interaction of MBP with lipid bilayers, and (4) the interaction of MBP with PLP.

In order to address the first and major issue. two preferred approaches exist. In both cases the particles must be imaged by EM, unstained, and under the conditions described in this thesis. In the first case the particles should be irnaged by cryo-TEM in vitreous ice. This will address the issues of staining raised in this thesis, and preserve the native structure sample. Secondly the particles should be imaged. freeze-dried by STEM.

This will facilitate the mass calculation of the particles, to assist in the determination of the nurnber of MBP molecules present in the reconstmction.

The use of diflerent lipid monolayers will provide information on the lipid component of the present reconstmctions. This is based on the possibility that different lipids may interact differently with MBP. This will also serve as test of the interaction of the different MBP charge isomers with other lipids. In using vesicles to produce lipid bilayers, it will be possible to detemine the structure of the different charge isomers when they complex with bilayers of lipid, and compare them with the data presented here with monolayers. Finally, since PLP is the other major component of the myelin sheath, the above mentioned experiments can then be applied to it. Also MBP and PLP can be imaged together to see how they interact under varied conditions. REFERENCES

Adamson-Sharpe, KM., Ottensmeyer, F.P. (1 98 1) Spatial resolution and detection sensitivity in microanalysis by electron energy Ioss selected imaging. J. Microsc. 122, 309-3 14.

Afshar-Rad, T., Bailey, A.I., Luclcham, P.F., MacNaughtan, W.. Chapman. D. (1987) Forces between proteins and mode1 polypeptides adsorbed on mica surfaces. Biochim. Biophys. Acta 915, 10 1- 11 1.

Agrawal, H.C., Burton, RM., Fishman, M.A., Mitchell, R.F., Prensky, A.L. (1972) Partiai characterization of a new myelin protein component. J. Neurochem. 19. 2083- 2089.

-4grawa1, R.K., Penczek, P., Grassucci, R.A., Li, Y., Leith, A.. Nierhaus, K.H., Frank. J. (1996) Direct visualization of the A-, P-, and E-site transfer RNAs in the Escherichia coli ribosomes. Science 271, 1000- 1002.

Aita, J.F., Snyder, D.H., Reichel, W. (1974) Myasthenia gravis and multiple sclerosis: an unusual combination of diseases. Neurology. 24, 72-75.

Akey. C.W., Rademacher, M. (1993) Architecture of the Xenopus nuclear pore complex revealed by three-dimensionai electron microscopy. J. Ce11 Biol. 122. 1-19.

Akiba, T.. Toyoshirna, C., Matsunaga, T., Kawamoto, M., Kubota. T., Fukuyama. K.. Narnba, K., Matsubara, H. (1996) Three-dimensional structure of bovine cytochrome bc, complex by electron cryornicroscopy and helical image reconstruction. Nature Struct. Biol. 3, 553-561.

Allen, I.V.. Millar, J.H.D., Hutchinson, M.J. (1978) General disease in 120 necropsy- proven cases of multiple sclerosis. Neuropathol. Appl. Neurobiol. 4. 279-284.

Andrews, D. W., Walter, P., Ottensmeyer, F.P. (1 987) Evidence for an extended 7SL RNA structure in the signal recognition particle. EMBO J. 6, 3471-3477.

Anthony, J.S., Moscarello, M.A. (1971) A conformation change induced in the basic encephaiitogen by Iipids. Biochim. Biophys. Acta 243, 429-433.

Aoyama, K., Ogawa, K., Kimura, Y., Fujiyoshi, Y. (1995) A method for 2D crystallization of soluble proteins at liquid-liquid interface. Ultramicroscopy 57. 345-354.

April, R.S., Vansomenberg, E. (1976) A case of neuromyelitis optica (Devic's syndrome) in systemic lupus erythematosus. Clinicopathologic report and review of the literature. Neurology 26. 1066- 1070.

Bach, D., Sela, B. (1985) Interaction of myelin basic protein with gangliosides and ganglioside-phospholipid mixtures. Biochim. Biophys. Acta 819, 225-230.

Banik, N.L., Davison, A.N. (1974) Lipid and basic protein interaction in myelin. Biochem. J. 143, 39-45.

Barnes, D., Munro, P.M.G., Youl, B.D., Prineas, J. W., McDonald. W.I. ( 199 1 ) The long standing MS lesion. A quantitive MN and elecîron rnicroscopic study. Brain 114. 1271-1280.

Barré, J.A., Morin, V., Draganesco, S., Reys, L. (1926) Encephalite periaxiale diffuse. Rev. Neurol. 2, 541-557.

Barylko, B., Dobrowolski, 2. (1984) CaL+-calmodulin-modulatedregulation of F-actin - myelin basic protein interaction. Eur. J. Ce11 Biol. 35, 327-335.

Baumeister, W. (1982) Towards higher resolution in biomoIecular electron microscopy. Ultramicroscopy 9, 15 1- 158.

Bauer, R. (1 988) Electron spectroscopic imaging: an advanced technique for imaging and analysis in transmission electron microscopy. Meth. Microbiol. 20, 1 13- 146.

Bazett-Jones, D. P., Ottensmeyer, F.P. ( 1 98 1 ) Phosphorus distribution in the nucleosome. Science 2 11, f 69- 1 70.

Bazett-Jones, D.P., Locklear, L., Rattner. J. (1988) Electron spectroscopic imaging of DNA. J. Ultrasmct. Mol. Stnict. Research 99, 48-58.

Bazett-Jones, D.P. (1993) Empirical basis for phosphorus mapping in DNA: protein complexes by electron spectroscopic imaging. Microb. Anal. J. 2, 69-79.

Bazeît-Jones, D.P., Mendez, E., Czarnota, G.J., Ottensmeyer, F.P., Allfiey. V.G. (1996) Visualization of unfolded nucleosomes associated with transcribing chromatin. Nucl. Acids Res. 24, 321-329.

Bedu-Addo, F.K., Huang, L. (1996) Effect of matrix lipid chain length on liposomes containing cholesterol and ganglioside GM,: implications in drug delivery. J. Pham. Sci. 85, 714-719.

Bellini, T., Rippa, M., Matteuzi, M., Paolino, E., Dallochio, F. (1 987) Polymenzation of myelin basic protein. Ital. J. Biochem. 35, 13 1A- 1 32A. Beniac, DR, Harauz, G. (1993) Visualisation of E. coli ribosomal RNA in situ by electron spectroscopic imaging and image averaging. Micron 24, 163- 171.

Beniac, D.R., Harauz, G. (1995a) Structures of small subunit ribosomal RNAs in situ fiom Escherichia diand Thermomyces lanuginosus. Mol. Cell. Bioc hem. 148. 165-181.

Beniac. D.R., Harauz, G. (1995b) Remarkable moments in single particle electron image anaiysis. Optik 99, 62-74.

Beniac, D.R., Luckevich, M.D., Czarnota, G.J., Tompkins. T.A.. Ridsdale. R.A.. Ottensmeyer, F.P., Moscarello, M.A., Harauz, G. (1997a) Three-dimensional structure of Myelin Basic Protein 1. Reconstruction via angular reconstitution of randomly oriented single particles. J. BioI. Chem. 272, 4261-4268.

Beniac, D.R., Czarnota, G.J., Rutherford, B.L., Ottensmeyer, F.P., Harauz G. ( 1997b) Three-dimensional architecture of Thermomyces lanuginosus small subunit ribosomai RNA. Micron, 28, 13-20.

Beniac, D.R., Czarnota, G.J., Rutherford, B.L., Ottensmeyer, F.P., Harauz. G. (1997~)The in situ archetecture of Escherichia coli ribosomal RNA derived by electron spectroscopic irnaging and three-dimensional reconstruction. J. Microsc. 188. 34- 3 5.

Beniac, D.R., Czarnota, G.J., Rutherford, B.L., Ottensmeyer, F.P.. Harauz. G. (1 997d) Probing ribosomal RNA by electron spectroscopic imaging and three-dimensional reconstruction. Microsc. Today 97, 10.

Beniac, D.R., Czarnota, G.J., Bartlett, T.A., Ottensrneyer, F.P., Harauz. G. (1 997e) Challenges of the three-dimensional reconstruction of ribonucleoprotein complexes from electron spectroscopic images - reconstmcting ribosomal RNA. Scan. Microsc. Suppl. 11, in press.

Boggs, J.M., Moscarello, MA., Papahadjopoulos, D. (1977) Phase separation of acidic and neutral phospholipids induced by human myelin basic protein. Biochemistry 16, 5420-5426.

Boggs, LM., Moscarello, M.A. (1978a) Effect of basic protein from human central nervous system rnyelin on lipid bilayer structure. J. Membr. Biol. 39, 75-96.

Boggs, J.M., Moscarello, M.A. (1978b) Structural organization of the human myelin membrane. Biochim. Biophys. Acta 515, 1-2 1.

Boggs, LM., Wood, D.D., Moscarello, M.A. (198 1) Hydrophobic and electrostatic interactions of myelin basic protein with lipid: participation of N-terminal and C- terminal portions. Biochernistry 20, 1065-1 073.

Boggs, J.M., Moscarello, M.A., Papahadjopoulos, D. ( 1982) Structural organization of myelin - role of lipid-protein interactions determined in mode1 systems. In: Jost. P.C.. Griffith, O.H. (eds.) Lipid-Protein Interactions. Wiley. New York. pp 1 -5 1 .

Boggs, J.M., Yip, P.M., Ebngaraj, G., JO, E. (1997) Effect of posttranslational modifications to myelin basic protein on its ability to aggregate acidic lipid vesicles. Biochemistry 36, 5065-507 1.

Boisset, N., Taveau, J.-C., Lamy, N., Wagenknecht, T., Rademacher, M.. Frank. I. (1990) Three-dimensional reconstmcticn of native Androctonus australis hemocyanin. J. Mol. Biol. 216, 743-760.

Boisset, N., Grassucci, R., Motta, G., Lamy, L, Radermacher, M., Taveau, J.C.. Liu. W.. Frank, J., Lamy, J.N. (1992) Three-dimensional irnmunoelectron microscopy of Androctonus australis hemocyanin: the location of monoclonal Fab fragments specific for subunit Aa6. Proc. 10th Eur. Congr. on Elect. Microsc. 1. 407-108.

Boisset, N., Radermacher, M., Grassucci, R., Traveau, J-C., Liu, W., Lamy, J., Frank. I.. Lamy, J.N. ( 1993) Three-dimensional irnrnunoelectron microscopy of scorpion hemocyanin labeled with a monoclonal Fab fragment. J. Struct. Biol. 111.234-244.

Boublik, M., Oostergetel, G.T., Frankland, B., Ottensmeyer, F.P. (1 984) Topographical mapping of ribosomal RNAs in situ by electron spectroscopie imaging. In Bailey G.W. (ed), Proc. 42nd Annual Meeting of the Electron Microscopy Society of America. San Francisco Press, San Francisco. pp 690-69 1.

Boublik, M., Mandiyan, V., Tumminia, S., Hainfeld, J.F., Wall, J.S. (1990) Structural analysis of ribosomai particles fiom Escherichia coli by scanning transmission electron microscopy. In Peachey, L.D., Williams, D.B. (eds), Proc. of the XIIth International Congress for Electron Microscopy. San Francisco Press. San Francisco. Volume 3, pp 134-135.

Boublik, M., Wall, J. (1992) Structure of rRNA in the ribosome. In Bailey, G.W.. Bentley, J., Small, J.A. (eds), Proc. of the 50th Annuai Meeting of the Electron Microscopy Society of America. San Francisco Press, San Francisco. pp 462-463.

Boylan, K.B., Ayres, T.M., Popko, B., Takahashi, N., Hood, L.E., Prusiner, S.B. (1990) Repetitive DNA (TGGA) 5' to the human myelin basic protein gene: a new form of oligonucleotide repetitive sequence showing length polymorphism. Genomics 6, 16-22. Brady, G.W., Birnbaurn, P.S., Moscarello, M.A., Papahadjopoulos, D. (1979) Liquid difiaction analysis of the mode1 membrane system: egg lecithin + myelin protein (N-2). Biophys. J. 25, 23-42.

Brady, G. W.. Murthy, N.S., Fein, D.B., Wood, D.D.. Moscarello. M.A. ( 198 1) The effect of basic myelin protein on multilayer membrane formation. Biophys. J. 34. 345- 350.

Brain, R., Wilkinson, M. (1957) The association of cervical spondylosis and disseminated sclerosis. Brain 80, 456-478.

Branden, C., Tooze, J. (1991) Introduction to protein structure. Garland Publishing. London.

Braun, P.E. (1977) Molecular architecture of myelin. In Morell, P. (ed.), Myelin. pp. 91- 115, Plenum, New York.

Bremer, A., Hem, C., Engel, A., Baumeister, W., Aebi, U. (1992) Has negative staining still a place in biomacromolecular electron microscopy? Ultramicroscopy 46. 85- 111.

Brenner, S., Home, R. W. (1959) A negative staining method for high resolution electron microscopy of viruses. Biochim. Biophys. Acta 34, 103-1 10.

Brimacombe, R., Gruer, B., Mitchell, P., Osswald, M., Rinke-Appel, J., Schüler. D.. Stade, K. (1990) Three-dimensional structure and hction of Escherichia coli 16s and 23s rRNA as studied by cross-linking techniques. In Hill, W.E.? Dahlberg, A.. Garrett, RA, Moore, P.B., Schlessinger, D., Wamer, J.R. (eds). The Ribosome: Structure, Function, and Evolution. Amencan Society for Microbiology. Washington. DC, pp 93-1 06.

Brimacombe, R. (1995) The structure of ribosomal RNA: a three-dimensional jigsaw puzzle. Eur. J. Biochem. 230, 365-383.

Brosnan, C.F., Selmaj, K., Raine, C.S. (1988) Hypothesis: a role for nimor necrosis factor in immune-mediated demyelination and its relevance to multiple sclerosis. J. Neuroimmunol. 18, 87-94.

Bmer, C., Lassmann, H., Waehneldt, T.V., Matthieu, LM., Linington, C. (1989) Differential ultrastructural localization of myelin basic protein, myelin/oligodendroglial glycoprotein, and 2',3'-cyclic nucleotide 3'- phosphodiesterase in the CNS of adult rats. J. Neurochem. 52, 296-304.

Caarnaiio, C.A., Zand, R. (1989) Homologous sequences in cholera toxin A and B subunits to peptide domains in rnyelin basic protein. FEBS Lett. 252. 88-90.

Cammer, W., Bloom, B.R., Norton, W.T., Gordon, S. (1978) Degradation of basic protein in myelin by neutral proteases secreted by stimulated macrophages: a possible mechanism of inflammatory demyelination. Proc. Natl. Acad. Sci. USA 75. 1554- 1558.

Campagnoni, A.T. (1 988) Molecular biology of myelin proains from the central nervous system. J. Neurochem. 51, 1-14.

Capel, M.S., Kjeldgaard, M., Engleman, D., Moore, P. ( 1988) Positions of S2. S 13. S 16. S17, S19 and S21 in the 30s ribosomai subunit of Escherichia cofi. J. Mol. Biol. 200, 65-87.

Carnegie, P.R. (1971) Amino acid sequence of the encephatilogenic protein of human myelin. Biochem. J. 123, 57-67.

Carswell, R. (1838) Illustrations of the elementary forrns of disease. London:Longman. Orme, Brown, Green, Longman.

Cavatorta, P., Giovanelli, S., Bobba, A., Riccio, P.. Szabo, A.G., Quagliariello. E. ( 1994) Myelin basic protein interaction with zinc and phosphate: fluorescence studies on the water-soiuble fom of the protein. Biophys. J. 66, 1 174- 1 179.

Chan, C.K., Ramwani, J., Moscarello, M.A. (1987) Myelin basic protein binds GTP at a single site in the N-terminus. Biochem. Biophys. Res. Commun. 152. 1468- 1473.

Chan, K.-F.J., Robb, N.D., Chen, W.H. (1990) Myelin basic protein: interaction with calmodulin and gangliosides. J. Neurosci. Res. 25, 53 5-544.

Chapman, B.E., Moore, W.J. (1976) Conformation of myelin basic protein in aqueous solution from nuclear magnetic resonance spectroscopy. Biochem. Biophys. Res. Commun. 73, 758-766.

Cheifetz, S., Moscarello, M.A. (1985) Effect of bovine basic protein charge microheterogeneity on protein-induced aggregation of unilarnellar vesicles containing a mixture of acidic and neutral phospholipids. Biochemistry 24. 1909- 1924.

Chiu, W. (1986) Electron microscopy of fiozen, hydrated biological specimens. Annu. Rev. Biophys. Biophys. Chem. 15, 237-257.

Chiu, W. (1993) What does electron cryomicroscopy provide that X-ray crystallography and NMR spectroscopy cannot? Annu. Rev. Biophys. Biomol. Struct. 22,233-253. Chou, F.C.-H., Chou, C.-H.J., Shapira, R., Kibler, R.F. (1976) Basis of microheterogeneity of myelin basic protein. J. Biol. Chern. 251, 267 1-2679.

Chou, C.-H.J., Shapira, R., Fritz, R.B. (1983) Encephalitogenic activity of the small form of mouse myelin basic protein in the SJL/J mouse. J. Irnmunol. 130. 2 183-2 186.

Compston, D.A.S. (1988). The 150th anniversary of the first depiction of the lesions on multiple sclerosis. J. Neuroi. Neurosurg. Psychiat. 5 1, 1 249- 1 252.

Compston, D.A.S., Kellar-Wood, H.F., Robertson, N., Sawcer, S., Wood, N. W. ( 1995) Genes and susceptibility to multiple sclerosis. Acta. Neurol. Scand. 161, 43-5 1.

Czarnota, G.J., Andrews, D.W., Farrow, N.A., Ottensmeyer, F.P. (1994). A structure for the signal sequence binding protein SRP54: 3D reconstruction from STEM images of single molecules. J. Stnict. Biol. 113, 35-46.

Czarnota, G.J. Ottensmeyer, F.P. (1996) Structural States of the nucleosome. J. Biol. Chem. 271, 3677-3683.

Dean, G. (1967) Annual incidence, prevalance and mortality in white south afncan-born and white immigrants to south africa. Br. Med. J. 2, 724-730.

Dean, G., Grimaldi. G., Kelly, R., Karhausen, L. (1979) Multiple sclerosis in southern Europe 1: prevalence in Sicily in 1975. J. Epidemiol. Commun. Hlth. 33. 107- 1 10.

Deber, C.M., Reynolds, S.J. (1991) Centrd nervous system myelin: structure. function. and pathology. Clin. Biochem. 24, 1 13- 1 34.

Deibler, G.E., Stone, A.L., Kies, M.W. (1990) Role of phosphorylation in conformational adaptability of bovine myelin basic protein. Proteins: structure. function, and genetics 7, 32-40.

Deibler, G.E., Stone, A.L., Kies, M.W. (1990) Role of phosphorylation in conformational adaptability of bovine myelin basic protein. Proteins 7. 32-40.

Deibler, C.M., Reynolds, S.J. (1991) Central nervous system myelin: structure. function. and pathology. Clin. Biochem. 24, 1 13-1 34.

Demel, R.A., London, Y., Guerts van Kessel, W.S.M., Vossenberg, F.G.A., van Deenen. L.L.M. (1973) The specific interactions of myelin basic protein with lipids at the air-water interface. Biochim. Biophys. Acta 311, 507-519.

Detels, R., Visscher, B.R., Malmgren, R.M., Coulson, A.H., Lucia, M.V., Dudley, J.P. (1977) Evidence for lower susceptibility to multiple sclerosis in Japaneese- Amencans. Am. J. Epidemiol. 105, 303-3 10.

Dobrowolski, Z., Osinska, H., Mossakowska, M., Barytko, B. (1986) ~a~~*calmodulin- dependent polymerization of actin by myelin basic protein. Eur. J. Ce11 Biol. 42. 17-26.

Dube, P., Tavares, P., Lurtz. R,van Heel, M. (1993) The portal protein of bacteriophage SPPI : a DNA pump with 13-fold symmetry. EMBO J. 12. 13034309.

Dyer, C.A., Benjamins, I.A. ( 1989) Organization of oligodendroglial membranes 1. J. Neurosci. Res. 24, 20 1-2 1 1.

Ebers, G.C., Bulman, D.E., Sodovink, A.D., Paty, D. W., Warren, S., Hader. W.. Murray. T.J., Seland, T.P., Duquette, P., Grey, T., Nelson, R., Nicolle. M.. Brunet. D. (1986) Population based studies of multiple sclerosis in twins. New Engl. J. Med. 315, 1636-1642.

Edwards, A.M., Ross, N.W., Ulmer, J.B., Braun, P.E. (1989) Interaction of myelin basic protein and proteolipid protein. J. Neurosci. Res. 22, 97- 102.

Egerton, R.F. (1 988) A simple pararneterization scheme for imer-shell cross sections. In Bailey, G.W. (ed), Proc. of the 46th Annual Meeting of the Electron Microscopy Society of Amerka. San Francisco Press, San Francisco. pp 532-533.

Egerton, R.F. (1996) Electron energy-loss spectroscopy in the electron microscope. Plenum Press, New York.

Elian, M., Nightingale, S., Dean, G. (1990) Multiple sclerosis among United Kingdom- born children of immigrants fiom the Indian subcontinent. AInca and the West Indies. J. Neurol. Neurosurg. Psychiat. 53, 906-91 1.

Engel, A., Meyer, J. (1980) Preparation of unstained protein structures for rnass determination by electron scattering. J. Ultrastruct. Res. 72, 2 12-222.

Engel, A., Hoenger, A., Hefii, A., Hem, C., Ford, R.C., Kistler, J., Zulauf, M. (1992) Assembly of 2-D membrane protein crystals. J. Stmct. Biol. 109, 2 19-234.

Eiigel, A., Colliex, C. (1993) Application of scanning transmission electron microscopy to the study of biological structure. Curr. Opin. Biotechnol. 4, 403-41 1.

Epand, R.M., Moscarello, M.A., Zierenberg, B., Vail, W.J. (1974) The folded conformation of the encephalitogenic protein of the human brain. Biochemistry 13. 1264- 1267. Eylar, EH., Bronstoff, S., Hashim, G., Caccam, J., Burnett, P. (1971) Basic Al protein of the myelin membrane. J. Biol. Chem. 246, 5770-5784.

Fantelli, F.J., Mitsumoto. H., Sebek, B.A. (1978) Multiple sclerosis and malabsorption. Lancet (Letter) i, 1039- 1040.

Farrow, N.A. Ottensmeyer, F.P. (1992) A posteriori determination of relative projection directions of arbitrarily oriented macromolecules. I. Opt. Soc. Amer. A 9. 1 749- 1760.

Farrow, NA., Ottensmeyer, F.P. ( 1993) Automatic 3D alignment of projection images of randomly oriented objects. UItramicroscopy 52. 14 1- 156.

Fidelio, G.D., Maggio, B., Curnar, F.A. (1982) Interaction of soluble and membrane proteins with rnonolayers of glycosphingolipids. Biochem. S. 203. 71 7-725.

Forrester. C., Lascelles, R.G. (1979) Association between polyneuritis and multiple sclerosis. J. Neurol. Neurosurg . Psychiat. 42, 864-866.

Fowler, W.E., Aebi, U. (1983) Preparation of single molecules and supramolecular complexes for high resolution metal shadowing. J. Ultrastruct. Res. 83. 3 19-334.

Frank, J. (1989) Image analysis of single macromolecules. Elect. Microsc. Rev. 2. 53-74.

Frank, J., Penczek, P., Grassucci, R., Srivastava, S. ( 1991 ) Three-dimensional reconstruction of the 70s Escherichia coli ribosome in ice: the distribution of ribosomal RNA. J. Ce11 Biol. 115, 597-605.

Frank, 1. ( 1992) Electron tomography : three-dimensional imaging with the transmission electron microscope. Plenum Press, New York.

Frank, J., Penczek, P. (1995) On the correction of the contrast transfer function in biological electron microscopy. Optik 98, 125- 129.

Frank, J., Zhu, J., Penczek, P., Li, Y., Srivastava, S., Verschoor, A., Radermacher. M.. Grassucci, R., Lata, R.K., Agrawal, R.K. (1995a) A model of protein synthesis based on cryo-electron microscopy of the E. coli ribosome. Nature 376, 441 -444.

Frank, J., Verschoor. A., Li, Y., Zhu, J., Lata, R.K., Radermacher, M., Penczek. P.. Grassucci, R., Agrawal, R.K., Srivastava, S. ( 1995b) A model of the translational apparatus based on a three-dimensional reconstruction of the Escherichia coli ribosome. Biochem. Ce11 Biol. 73, 757-765. Frank, J. (1 996) Three-dimensional electron microscopy of macromolecular assemblies. Academic Press, New York.

Fraser, P.E., Deber, C.M. (1985) Structure and function of the proline-rich region of myelin basic protein. Biochemistry 24, 4593-4598.

Fraser, P.E.. Rand, R.P., Deber, C.M.(1 989) Bilayer stabilizing properties of myelin basic protein in dioleoylphosphatidylethanolamine systems. Biochim. Biophys. Acta 983. 23-29.

Glasel, J.A., Deutscher, M.P. (eds.) (1 995) Introduction to biophysical methods for protein and nucleic research. Academic Press, San Diego.

Goldenberg, P.Z., Kwon, E.E., Benjamins, J.A., Whitaker, J.N., Quarles. R.H.? Prineas. J. W. ( 1989) Opsonization of normal myelin by anti-myelin antibodies and normal senun. J. Neuroimrnunol. 23, 157-166.

Golds, E.E.. Braum, P.E. (1978) Protein associations and basic protein conformation in the myelin membrane. J. Biol. Chem. 253, 8 162-8 170.

Golubovich, V.P., Kirnaskh, L.I., Gaiaktionov, S.G. (1989) Theoretical conformational anaiysis of the molecule of an encephalitogenic peptide and study of the "structure-activity" relations in a series of its analogues. Biophysics (translated from the Russian Biofizika 34, 368-37 1, 1989) 34, 395-399.

Gorodezky, C., Najera, R., Rangel, B.E., Castro, L.E., Flores, J.. Velazquez. G.. Granados, J., Sotelo, J. (1986) Imrnunogenic profile of multiple sclerosis in Mexicans. Hum. Imrnunol. 16, 364-374.

Gould, R.M., London, Y. (1972) Specific interaction of central nervous sy stem myelin basic protein with lipids. Biochim. Biophys. Acta 290, 200-2 18.

Gow, A., Auton, W., Smith, R. (1990) Interactions between bovine myelin basic protein and zwitterionic lysophospholipids. Biochemistry 29. 1 142- 1147.

Gow, A., Smith, R. (1989) The thrmodynarnicaily stable state of myelin basic protein in aqueous solution is a flexible coil. Biochem. J. 257, 535-540.

Graham, D.I., Lantos, P.L.eds. (1997) Greenfield's Neuropathology. Vol. 1. Arnold, New York, U.S.A.

Grigorieff, N., Ceska, T.A., Downing, K.H., Baldwin, J.M., Henderson, R. (1996) Electron crystailographic refinement of the structure of bactenorhodopsin. J. Mol. Biol. 259, 393-421. Hainfeld, J.F.. Safer. D.. Wall, J.S., Simon, M.. Lin. B.. Powell. R.D. (1994) Methylamine vanadate (Nanovan) negative stain. Proc. of the 52nd Annal Meeting of the Microscopical Society of America San Francisco, USA. 132- 133.

Harauz, G., van Heel, M.G. (1986) Exact filters for general geometry three dimensional reconstruction. Optik 73, 146- 156.

Harauz, G., Stoffler-Meilicke, M., van Heel, M. (1987) Characteristic views of prokaryotic 50s nbosomal subunits. J. Mol. Evol. 26, 347-357.

Harauz, G., Boekema, E.J., van Heel, M.G. (1988) Statistical image anaiysis of electron micrographs of nbosomal subunits. Meth. Enzymol. 164, 35-49.

Harauz, G., Boekema E.J. (1992) Processing and analysis of electron images of biomoiecules. Ch. 1 1 in Image analysis in biology. Hader, D-P (ed.). CRC Press. London.

Harauz, G. (1992) Electron microscopical projections of the large nbosomal subunit from Thermomyces lunuginosus. Biochim. Biophys. Acta 1132, 58-66.

Harauz, G., Flannigan, D.T.(1992) Characteristic electron microscopical projections of the small ribosomal subunit from Thermomyces lanuginosus. Biochim. Biophys. Acta 1130, 289-296.

Haraw, G., Letvenuk, L., Flannigan, D.T. (1992) Electron image analysis of ribosomal subunits fiom Thermus aquaticus. Biochim. Biophys. Acta 1129. 207-111.

Harauz, G., Cicicopol, C., Hergl, R., Cejka, Z., Goldie, K., Santarius. U., Engel. A.. Baumeister, W. (1996) Structural studies on the 2.25-MDa homomultimenc phosphoenolpynwate synthase fiom Staphylothermus marinus. J. S tmct. Bio 1. 116. 290-3 O 1.

Harris. J.R. ( 1997) Negative staining and cryoelectron microscopy. BIOS Scientific PubIishers, Oxford.

Hart, M.N., Earle, K.M. ( 1975) Haemorragic and penvenous encephditis: n clinicd- pathological review of 3 8 cases. J. Neurol. Neurosurg. Psychiat. 38, 585-59 1.

Hayat, M.A., Miller, S.E. (1990) Negative staining. McGraw-Hill Publishing Company. Toronto.

Hayer-Hartl, M., Brophy, P.J., Marsh, D., Watts, A. (1993) Interaction of two complementary fiagrnents of the bovine spinal cord myelin basic protein with phosphatidylglycerol bilayers, studied by 'H and "P NMR spectroscopy. Biochemistry 32, 9709-97 13.

Hefher, R.R., Porro, R.S.. Olson, M.E.. Earle, K.M. (1976) A demyelinating disorder associated with cerebrovascular amyloid angiopathy. Arch. Neurol. 33. 50 1-506.

Henderson, R., Baldwin, J.M., Ceska, T.A., Zemlin, F., Beckmann. E., Downing, K.H. (1990) Mode1 for the structure of bacteriorhodopsin based on high-resolution electron cryo-microscopy. J. Mol. Biol. 213: 899-929.

Henderson, R. (1995) The potentiai and limitations of neutrons, electrons and X-rays for atomic resolution microscopy of unstained biological molecules. Quart. Rev. Biophys. 28, 171 - 193.

Heng, Y-M.,Simon, G.T., Boublik, M., Ottensmeyer, F.P. (1 990) Experimentai ionization cross-sections of phosphorus and calcium by electron spectroscopic imaging. J. Microsc. 160, 161 - 171.

Hinshaw, J.E., Carragher, B.O., Milligan, R.A. (1992) Architecnire and design of the nuclear pore complex. Ce11 69, 1 133- 1 141.

Hughes, D.W., Stollery, J.G., Moscarello, M.A., Deber, CM. (1982) Binding of myelin basic protein to phospholipid micelles. J. Biol. Chem. 257. 4698-4700.

Inouye, H., Kirschner, D.A. (1 984) Effects of ZnCl, on membrane interactions in myeiin of normal and shiverer mice. Biochim. Biophys. Acta 776, 197-208.

Inouye, H., Kirschner, D.A. (1 991) Folding and function of the myelin proteins fiom primary sequence data. J. Neurosci. Res. 28, 1- 17.

Israelachvili, J., Marra, J. (1986) Direct methods for rneasureing conformational water forces (hydration forces) between membrane and other surfaces. Methods Enzmol. 127, 353-360.

Itoyama, Y., Stemberger, N.H., Webster, H.D., Quarles, R.H., Cohen. S.R.. Richardson. E.P. (1980) Irnmunocytochemical observations on the distribution of myelin- associated glycoprotein and myelin basic protein in multiple scierosis lesions. Am. Neurol. 7, 167-177.

Jap, B.K., Zulauf, M., Scheybani, T., Hefti, A., Baurneister, W., Aebi, U.. Engel. A. (1 992) 2D crystallization: from art to science. Ultramicroscopy 16. 45-84.

Jones, A.J.M., Epand, R-M. (1980) Effect of rnicroheterogeneity on the structure and function of the myelin basic protein. Biochim. Biophys. Acta 625, 165- 178. Juba, A. (1939) Die maligne exacerbation der multiplen sklerose. Arch. Psychiat. 109. 727-743.

Kalwy, S.A., Smith, R. (1994) Mechanisms of myelin basic protein and proteolipid protein targeting in oligodendrocytes. Mol. Membr. Biol. 11. 67-78.

Keniry, M.A., Smith, R. (1979) Circular dichroic analysis of the secondary structure of myelin basic protein and derîved peptides bound to detergents and to Iipid vesicles. Biochim. Biophys. Acta 578, 3 8 1-39 1.

Keniry, M.A.. Smith, R. (1981) Dependence on lipid structure of the coil-to-helix transition of bovine myelin basic protein. Biochim. Biophys. Acta 668. 107- 1 18.

Kesselring, J., Miller, D.H., Robb, S.A., Kendall. B.E., Moseley, I.F.. Kingsley. D.. du- Boulay, E.P., McDonald, W.I. (1990) Acute disseminated encephalomyelitis: MN findings and the distinction fkom multiple sclerosis. Brain 113. 29 1-302.

Kies, M. (1965) Chernical studies on an encephalitogenic protein from guinea pig brain. Ann. N.Y. Acad. Sci. 122, 161-170.

Kinney, E.L. Berdoff, R.L. Rao, N.S., Fox, L.M. (1979) Devic's syndrome and systemic lupus erythematosus. A case report with necroscopy. Arch. Neurol. 36. 643-644.

Kirschner, D.A., Ganser, A.L. (1980) Compact myelin exists in the absence of basic protein in the shiverer mutant mouse. Nature 283, 207-210.

Kirschner, D.A., Inouye, H., Ganser, A.L., Mann, V., ( 1989) Myelin membrane structure and composition correlated: a phylogenetic study. J. Neurochem. 53. 1 599- 1609.

Knauer V., Hegerl R., Hoppe W. (1 983) Three-dimensional reconstruction and averaging of 30s ribosomal subunits of Escherichia coli fkom electron micrographs. J. Mol. Biol. 163, 409-430.

Kolodziej, S.., Penczek, P.A., Schroeter, J.P. Stoops, J.K. (1996) Structure-function relationships of the Saccharomyces cerevisiae fatty acid synthase. Three- dimensional structure. J. Biol. Chem. 271, 28422-28429.

Korn, A.P. (1980) Dark field electron microscopy of the 30s subunit of the E. coli ribosome. Ultrarnicroscopy 5, 5 13-520.

Korn, 1\.P., Spitnik-Elson, P., Elson, D., Ottensmeyer, F.P. (1983) Specific visualization of ribosomal RNA in the intact ribosome by electron spectroscopie imaging. Eur. J. Ce11 Biol. 31, 334-340. Koshy, KM., Hashim, G.A., Boggs. J.M. (1996) Proton NMR study of peptides fiom myelin basic protein: evidence for ~ys'~-~is"interaction revealed fiom histidine line broadenuig. Biochim. Biophys. Acta 1293, 23-30.

Krigbaum, W.R., Hsu, T.S. (1975) Molecular conformation of bovine Al basic protein. a coiling macromolecule in aqueous solution. Biochemistry 14. 2542-2516.

Kubalek, E.W., Le Grice, S.F.J., Brown, P.O. (1994) Two-dimensional crystallization of histidine-tagged HIV- 1 reverse transcriptase promoted by a novel nickel-chelating lipid. L Struct. Biol. 113. 1 17- 123.

Kühlbrandt, W., Unwin, P.N.T. (1982) Distribution of RNA and protein in crystalline eukaryotic ribosomes. L Mol. Biol. 156, 43 1-448.

Kiihlbrandt, W. (1992) Two-dimensional crystallization of membrane proteins. Quart. Rev. Biophys. 25, 1-49.

Kurdi, A., Ayesh, I., Abdallat, A., Maayta, U. (1977) Different B lymphocyte alloantigens associated with multiple sclerosis in Arabs and North Europeans. Lancet 28. 1 123- 1125.

Kurowia, Y., Schibasaki, H., Ikeda, M. (1983) Prevalence of multiple sclerosis and its north south gradient in Japan. Neuroepidemiology 2, 62-69.

Kyle, KM., Harauz, G. (1992) Electron microscopie visualisation of the 5s rRNA-YL3 complex £Yom Saccharomyces cerevisiae. Mol. Cell. Biochem. 117. 1 1-2 1.

Laernmli, V. (1970) Cleavage of structural proteins during the assembly of the head of the bacteriophage T4. Nature 227, 680-685.

Lake, J.A., Henderson, E., Clark, M. W., Scheinman, A., Oakes, M.I. (1986) Mapping evolution with three-dimensional ribosome structure. Sys. Appl. Microbiol. 7. 13 1- 136.

Lampe, P. D., Nelsestuen, G.L. ( 1 982) Myelin basic protein-enhanced fusion of membranes. Biochim. Biophys. Acta 693, 320-325.

Lassmann, H., Budka, H., Schnaberth, G. (1981) Inflamrnatory demyelinating polyradiculitis in a patient with multiple sclerosis. Arch. Neurol. 38, 99-102.

Lata R.K., Agrawal R.K., Penczek P., Grassucci R., Zhu J., Frank J. (1996) Three - dimensional reconstruction of the Escherichia coli 305 ribosomal subunit in ice. J. Mol. Biol. 262. 40-52. Leberman, R. ( 1965) Use of uranyl formate as a negative stain. J. Mol. Biol. 13. 606.

Lees, M.B., Brostoff, S. W. (1984) Proteins of myelin. In: Morell, P. (ed.), Myelin. 2nd ed., Plenum Press, New York, pp. 197-224.

Linnington, C., Bradl, M., Lassmann, H., Brunner, C., Vass. K. (1988) Augmentation of dernyelination in rat acute allergic encephalomyelitis by circulating mouse monoclonal antibodies directed against myelin/oligodendrocyte glycoprotein. Amer. J. Pathol. 130, 443-454.

Liu, W., Boisset, N., Frank, J. (1995) Estimation of variance distribution in three- dimensional reconstruction. II. Applications. J. Opt. Soc. Am. A 12, 2628-3635.

Liu, W., Frank, J. ( 1995) Estimation of variance distribution in three-dimensional reconstruction. 1. Theory. J. Opt. Soc. Amer. A 12, 2615-2627.

MacNaughtan, W., Snook, KA., Caspi, E.. Franks, N.P. (1985) An X-ray diffraction analysis of oriented lipid multilayers containing basic proteins. Biochim. Biophys. Acta 818, 132-148.

Maggio, B., Yu, R.K. (1989) Interaction and fusion of unilamellar vesicles containing cerebrosides and sulfatides induced by myelin basic protein. Chem. Phys. Lipids 51, 127-136.

Maggio, B., Yu, R.K. (1992) Modulation by glycosphingolipids of membrane-membrane interactions induced by myelin basic protein and melittin. Biochim. Biophys. Acta 1112, 105-114.

Malhotra, A., Gabb, H.A., Harvey, S.C. (1993) Modeling large nucleic acids. Curr. Opin. Struct. Biol. 3, 241-246.

Malhotra, A., Harvey, S.C. (1994) A quantitative mode1 of the Escherichia coli 16s RNA in the 30s ribosomal subunit. J. Mol. Biol. 240, 308-340.

Marburg, 0. (1 906) Die sogenannte 'akute multiple skierose' (Encephalamyelitis periaxialis scleroticans). Jahrb. Psychiat. Neurol. (Leipzig) 27' 2 13-3 1 1.

Marrosu, M.G., Muntoni, F.. Murru, M.R., Spinicci, G.. Pischedda, M.P.. Goddi. F.. Cossu, P., Pirastu, M. (1988) Sardinian multiple sclerosis is associated with HLA- DR4: a serologic and molecular analysis. Neurology 38, 1749-1753.

Martenson, R.E. ( 198 1) Prediction of the secondary structure of myelin basic protein. J. Neurochem. 36, 1543- 1560. Martenson, R.E. (1986) Possible hydrophobie region in myelin basic protein consisting of an orthogonally packed P-sheet. J. Neurochem. 46, 16 12- 1 622.

Martin, R, McFarland, H.F., McFarlin, D.E. (1 992) Imrnunological aspects of demyelinating diseases. Annu. Rev. Imrnunol. 10, 1 53 - 1 87.

Mastronardi, F.G., Ackerley, C.A., Arsenault, A.L., Roots, B.I.. Moscarello. M.A. (1993) Demyelination in a transgenic mouse: a mode1 for multiple sclerosis. J. Neurosci. Res. 36, 3 15-324.

Mateu, L., Lmti, V., Vonasek, E., Borgo, M., Lachapelle, F. (1 996) Order-disorder phenornena in myelinated nerve sheaths. VI. The effects of quaking, jimpy. and shiverer mutations: an X-ray scattering study of mouse sciatic and optic nerves. J. Mol. Biol. 256, 3 19-329.

Matthews, W.B. (1968) The neurological complications of ankylosing spondylitis. J. Neurol. Sci. 6, 561-676-

May, R.P., Nowotny, V.. Nowotny, P., Voss, H., Nierhaus, K.H. (1992) Inter-protein distances within the large subunit from Escherichia coii ribosomes. EMBO J. 11. 373-378.

McGough, A., Way, M., DeRosier, D.J. (1994) Determination of the alpha-actinin binding site on actin filaments by cryo-electron microscopy and image analysis. I. Ce11 Biol. 126, 433-443.

McLaurin, J., Ackerley, C.A., Moscarello, M.A. (1 993) Localization of basic proteins in human myelin. J. Neurochem. Res. 35, 61 8-628.

McLaurin, J., Hashim, G., Moscarello, M.A. (1992) An antibody specific for component 8 of myelin basic protein reacts strongly with component 8 fiom multiple sclerosis brain. J. Neurochem. 59, 14 14- 1420.

McRee, D.E. (1993). Practical protein crystallography. Academic Press, San Diego.

McWilliarns, R.A., Glitz, D.G. (1991) Localization of a segment of 16s RNA on the surface of the small nbosomal subunit by immune electron microscopy of complementary oligodeoxynucleotides. Biochimie 773, 9 1 1-9 1 8.

Mendz, G.L., Barden, J.A., Martenson, R.E. (1995) Conformation of a tetradecapeptide epitope of myelin basic protein. Eur. J. Biochem. 231, 659-666.

Mendz, G.L., Miller, D.J., Jarnie, I.M., White, J.W., Brown, L.R., Ralston, G.B., Kaplin. LJ. (1991) Physicochemical characterization of dodecylphosphocholine / palmitoyllysophosphatidic acid / myelin basic proteir. complexes. Biochemistry 30. 6509-65 16.

Men4 G.L., Moore, W.J. (1982) Proton NMR evidence for secondary and tertiary structure in myelin basic proteins. Biochem. Biophys. Res. Commun. 105, 1333- 1340.

Mendz, G.L., Moore, W.J. (1985) N.M.R. studies of myelin basic protein. Conformation of a peptide that is an antigenic determinant for B-ce11 reactivity. Biochem. J. 229. 305-313.

Mendz, G.L., Moore, W.J., Kaplin, I.J. ( 1988) Characterization of dodecylphosphocholine/myelinbasic protein complexes. Biochemistry 27.3 79-3 86.

Menon. N.K., Williams, R.E., Kampf, K., Campagnoni, A.T. (1 990) An analysis of the region of myelin basic protein that binds to PC. Neurochem. Res. 15, 777-783.

Miller, D.H., Homabrook R.W., Dagger, J., Fong, R. (1986) Ethnic and HLA patterns related to multiple sclerosis in Wellington, New Zealand. J. Neurol. Neurosurg. Psychiat. 49, 43-46.

Miller, D.H.,Ormerod, I.E.C. Gibson, A., du-Boulay, E.P., Rudge, P., McDonald. W.I. (1987) MR brain scanning in patients with vasculitis: differentiation from multiple sclerosis. Neuroradiology 29, 226-23 1.

Miller, D.H., Kendall, B.E., Barter, S., Johnson, G., MacManus, D.G., Logsdail- S.J.. Ormerod, LE., McDonald, W.I. ( 1988) Magnetic resonance imaging in central nervous system sarcoidosis. Neurology 38, 3 78-3 83.

Miller, D.H., Hammond, S.R., McLeod, J.G. ( 1990) Multiple sclerosis in Australia and New Zealand: are the determinants genetic or environmantal? J. Neorol. Neurosurg. Psychiat. 53, 903-905.

Monferran, C.G., Maggio, B., Cumar, F.A. (1986) Effect of chernical modifications of myeiin basic protein on its interaction with lipid interfaces and fusion ability. Mol. Cell. Biochem. 70, 13 1- 139.

Moore, P.%. (1988) The ribosome returns. Nature 331, 223-227.

Moore, W.J. (1982) Interaction of basic protein and peripheral nerve P2 protein with lipids. Biophys. J. 37, 59-61.

Morell, P.M. (1977) Myelin, Plenum Press, New York. Morgan, D.G., Owen, C.. Melanson, L.A., DeRosier, D.I. (1 995) Structure of bacterial flagella filaments at 1 1A resolution: packing of the alpha helices. J. Mol. Biol. 249, 88-1 10.

Mory, C., Collies, C., Revet, B., Delain, E. (1981) Improved visualization of single and double stranded nucleic acid by STEM. Ultramicroscopy. 7, 16 1- 1 68.

Moscarello, M.A., Brady, G.W., Fein, D.B.,Wood, D.D.. Cruz, T.F. (1986) The role of charge microheterogeneity of basic protein in the formation and maintenance of the multilayered structure of myelin: a possible role in multiple sclerosis. J. Neurosci. Res. 15, 87-99.

MoscareIIo, M.A., Pang, H., Pace-Asciak, C.R., Wood. D.D. (1992) The N terminus of human myelin basic protein consists of Cz, C,, C,, and C, alkyl carboxylic acids. J. Biol. Chem. 267, 9779-9782.

Moscarello, M.A., Wood, D.D.,Ackerley, C., Boulias, C. (1994) Myelin in multiple sclerosis is developmentally immature. J. Clin. Invest. 94. 146- 1 54.

Moscarello, M.A. ( 19%) in ce11 biology and pathology of rnyelin: evolving biological concepts and therapeutic approaches (Devon, R. M., Doucette. R.. luurlink. B .H.J.. Nazarali, AL, Schreyer, D.J.,Verge, V.M.K.) Plenum Press, New York.

Moskaitis, J.E., Shriver, L.C., Campagnoni, A.T. (1987) The association of myelin basic protein with itself and other proteins. Neurochem. Res. 12, 409-417.

Mosser, G., Brisson, A. ( 1991) Conditions of MO-dimensional crystallization of cholera toxin B-subunit on lipid films containing ganglioside GMI. J. Stnict. Biol. 106. 191-198.

Mueller, F., Doring, T., Erdemir, T., Greuer, B., Jünke. N., Osswald. M.. Rinke-Appel. J., Stade, K., Thamm, S., Bnmacombe, R. (1995) Getting closer to an understanding of the three-dimensional structure of ribosomal RNA. Biochem. Ce11 Biol. 73, 767-773.

Mumford, C.J. Wood, N. W., Kellar-Wood, H.F., Thorpe, J. W,. Miller, D.H.,Compston, D.A. (1 994) The British Ides survey of multiple sclerosis in twins. Neurology 44. 11-15.

Nag, B., Wada, KG., Deshpande, S.V., Passmore, D., Kendrick, T., Sharma, S.D., Clark. BR, Mccomell, H.M. (1993) Stimulation of T cells by antigenic peptide complexed with isolated chains of major histocompatibility complex class II molecules. Proc. Natl. Acad. Sci. USA 90, 1604-1608. Naito, S., Kuroiwa Y., Itoyama, T., Tsubaki, T., Horikawa, A., Sasazuki. T., Noguchi. S., Ohtsuki, S., Tokuomi, H., Miyatake, T., Takahata, N., Kawanarni. S. (1978) HLA nad Japanese MS. Tissue Antigens 12, 19-24.

Napolitano, L., Lebaron, F., Scaletti, J. (1967) Preservation of myelin lamellar structure in the absence of lipid: a correlated chernical and morphological study. J. Cell Biol. 34, 817-826.

Nezil, F.A., Bayerl, S., Bloom, M. (1992) Temperature eruptions of vesicles in mode1 membranes studied by NMR. Biophys. J. 61, 1413-1426.

Nierhaus, K.H. (199 1) The assembly of prokaryotic ribosomes. Biochimie 73. 739-75 5.

Orlova, E.V., Serysheve, LI., van Heel, M., Hamilton, S.L., Chiu. W. (1996) Two structural configurations of the skeletal muscle calcium release channel. Nat. Struct. Biol. 3. 547-552.

Oettl, H., Hegerl, R., Hoppe, W. (1983) Three-dimensional reconstruction and averaging of 50s ribosomal subunits of Escherichia Coli from electron micrographs. J. Mol. Biol. f 63, 43 1-450.

Ong, R.L., Yu, R.K. (1984) Interaction of ganglioside GM,and myelin basic protein studied by carbon-13 and proton nuclear magnetic resonance spectroscopy. J. Neurosci. Res. 12. 3 77-393.

Ottensmeyer, F.P., Andrew, J.W. (1980) High resolution microanalysis of biological specimens by electron energy loss spectroscopy and by electron spectroscopic imaging. J. Ultrastruct. Res. 72, 336-348.

Ottensmeyer, F.P. (1982) Scattered electrons in microscopy and microanalysis. Science 215, 461-466.

Ottensmeyer, F.P . ( 1984) Electron spectroscopic imaging : parallel energy filtenng and microanalysis in the fixed-beam electron microscope. J. Ultrastmct. Res. 88. 13 1 - 134.

Ottensmeyer, F .P. ( 1986a) Elemental mapping by energy filtration: advantages. limitations, and compromises. Am. N.Y. Acad. Sci. 483, 339-353.

Ottensmeyer, F.P. (1986b) Scattered electrons in biological structure determination. In: Examining the subrnicron world. Feder, R., MacGowan, W. (eds.) Plenum, New York, pp 137-151.

Pdi, T., Ebert, B., Horvath, L.I. (1987) ESR imaging of myelin basic protein induced vesicle aggregation. Biochim. Biophys. Acta 904, 346-352.

Payan, H., Levine, S. (1963) Hydrocephalus due to membranous obstruction of the founh ventricle in a case of multiple sclerosis. Acta Neuropathol. (Berlin) 3. 202-205.

Peiffer, J. (1982) Gerstmann-Straussler's disease, atypicd multiple sclerosis and carcinomas in a farnily of sheep breeders. Acta Neuropathol. (Berlin) 56. 87-93.

Penczek, P., Grassucci, RA.. Frank, J. (1994) The ribosome at improved resolution: new techniques for merging and orientation refinement in 3D cryo-elecaon microscopy of biological particles. Ultramicroscopy 53. 25 1-270.

Peters, G. (1958) Multiple sklerose; Encephalomyelitis disseminata non purulenta acuta (akute multiple sklerose). In : Lubarsch, O., Henke, F, Rossle, R. eds. Handbuch der Speziellen Pathologischen Anatomie und Histologie. 13, Part ZA. Berlin: Spnnger Verlag. 525-629.

Petit, H., Wattel, F., Caron, C.C. (1969) Polyradiculonevrite aigue avec dissociation albuminocytologique au cours de I'evolution d'une sclerose en plaques. Lille Med. 14, 708-710.

Petrescu, A. (1969) Histochemistry of lipids in multiple sclerosis. Wein Z. Nervenheilk Grenzgeb 2 (Suppl. 2), 38-52.

Pereyra, P.M., Horvath, E., Braun. P.E. (1988) Triton X-100 extractions of central nervous system myelin indicate a possible role for the minor myelin proteins in the stability of lamellae. Neurochem. Res. 13, 583-595.

Pérez-Gil, J.. Keough, K.M.W. (1994) Structurai similarities between myelin and hydrophobic surfactant associated proteins: protein motifs for interacting with bilayers. J. Theoret. Biol. 169, 221-229.

Perkins, G.A., Bergsma-Schutter, W., Keegstra, W., Amberg, A.C., Cou, O., Scherrer. K. (1994) The 1.5-MI projection structure of HeLa ce11 prosome-MCP (proteasorne) provided by two-dimensional crystals. J. Struct. Biol. 113. 124- 134.

Persaud, R., Boggs, J.M., Wood, D.D., Moscarello, M.A. (1989) Interaction of glycosylated hurnan myelin basic protein with lipid bilayers. Biochemistry 28. 4209-42 16.

Pirollet, F., Derancourt, J., Haiech, J., Job, D., Margolis, R.L. (1992) Ca2--calmodulin regulated effectors of microtubule stability in bovine brain. Biochemistry 31.8849- 8855. 225

Price, W.S., Mendz, G.L., Martenson, R-E. ( 1988) Conformation of a heptadecapeptide comprising the segment encephalitogenic in rhesus monkey. Biochemistry 27. 8990-8999.

Prineas, J. W., Kwon, E.E., Sternberger, N.H., Lennon, V.A. (1984) The distribution of myelin-associated glycoprotein and myelin basic protein in actively demyelinating multiple sclerosis lesions. J. Neuroirnmunol. 6. 25 1-264.

Quarles, R.H., Everly, J.L.. Brady, R.O., (1972) Demosntration of a glycoprotein which is associated with a purified myelin fraction fiom a rat brain. Biochem. Biophys. Res. Commun. 47, 49 1-497.

Quarles, R.H., Everly, J.L., Brady, RO. (1973) Myelin-associated glycoprotein: a developmental change. Brain Res. 58, 506-509.

Radermacher, M., Wagenknecht, T., Verschoor, A., Frank, J. (1987) Three-dimensional structure of the large nbosomal subunit fiom Escherichia coli. EMBO J. 6. 1 107- 11 14.

Radermacher, M., Srivastava, S., Frank, J. (1992) The structure of the 50s ribosomal subunit from E. coli in frozen hydrated preparation reconstnicted with SECRET. Proceedings of the 10th European Congress on Electron Microscopy 3. 19-22.

Radermacher, M., Rao, V., Grassucci, R., Frank, J., Timrnerman, A.P., Fleischer. S.. Wagenknecht, T. (1 994) Cryo-electron microscopy and three-dimensional reconstruction of the calcium release channeVryanodine receptor fiorn skeletal muscle. J. Cet1 Biol. 127, 41 1-423.

Ramwani, J.J., Epand, R.M., Moscarello, M.A. (1989) Secondary structure of charge isomers of myelin basic protein before and after phosphorylation. Biochemistry 28. 653 8-6543.

Rand, R.P., Fuller, N., Lis, L.J. (1979) Myelin swelling and measurement of forces between myelin membranes. Nature 279, 258-260.

Rang, E.H., Brooke, B.N., Hermon-Taylor, J. (1982) Association of ulcerative colitis with multiple sclerosis. Lancet ii, 555.

Ranshoff, R.M. (1992) T-ce11 receptor germline genes and multiple sclerosis susceptability: an unfinished tale. Neurology 42, 7 14-7 18.

Reed, R.A., Mattai, I., Shipley, G.G. (1987) Interaction of cholera toxin with ganglioside GMlreceptors in supported lipid monolayers. Biochemistry 198, 824-832. Reid, H.M.,Bayerl, T.M. (1993) Interaction of myelin basic protein with single bilayers on a solid support: NMR. DSC and polarized infiared ATR study. Biochim. Biophys. Acta 1151, 127-136.

Reuter, A., Gaupp, R. (1932) Beitrag zur Frange der akuten multiplen sklerose. 2. ges Neurol. Psychiat. 138, 495-502.

Ribi, H.O., Ludwig, D.S.. Mercer, K.L., Schoolnik G.K., Komberg, R.D. (1988) Three- dimensional structure of cholera toxin penetrating a lipid membrane. Science 239. 1272- 1276.

Ridsdale, R.A., Beniac, D.R., Tompkins, T.A., Moscarello, M.A.. Harauz. G. ( 1997) Three-dimensional structure of myelin basic protein II. mollecuiar modelling and considerations of predicted structures in multiple sclerosis. J. Biol. Chem. 272. 4269-4275.

Robertson, N., Compston, A. (1 995) Surveying multiple sclerosis in the United Kingdom. J. Neurol. Neurosurg. Psychiat. 58, 2-6.

Roth, G.A., Gonzalez, M.D., Monferran, C.G., DeSantis, M.L., Cumar, F.A. (1993) Myelin basic protein domains involved in the interaction with actin. Neurochem. Int. 23, 459-465.

Roux, M., Nezil, F.A., Monck, M., Bloom, M. (1994) Fragmentation of phospholipid bilayers by myelin basic protein. Biochemistry 33. 307-3 1 1.

Rurnsby, M.G. ( 1978) Organization and structure in central-nerve myelin. Biochem. Soc. Trans. 6, 448-462.

Sankaram, M.B., Brophy. P.J., Marsh, D. (1989a) Spin-label ESR studies on the interaction of bovine spinal cord myelin basic protein with dimyristoylphosphatidylglycerol dispersions. Biochemistry 28. 9685-969 1.

Sankaram, M.B., Brophy, P.J., Marsh. D. ( 1989b) Interaction of two complemrntary fragments of the bovine spinal cord myelin basic protein with phospholipid bilayers. An ESR spin-label study. Biochemistry 28, 9692-9698.

Sankaram, M.B., Brophy, P.J., Manh, D. (1991) Lipid-protein and protein-protein interactions in double recombinants of myelin proteolipid apoprotein and myelin basic protein with dimyristoylphosphatidylglycerol. Biochemistry 30. 5866-5 873.

San Martin, MC, Stamford, NP., Dammerova, N., Dixon, NE., Carazo, J.M. (1995) A structural mode1 for the Escherichia coli DnaB helicase based on eiectron microscopy data. J. Struct. Biol. 114, 167-176. San Martin, M.C.,Gms, C., Carazo, J.M. (1997) Six molecules of SV40 large T antigen assemble in a propeller-shaped particle around a channel. J. Mol. Biol. 268. 15-20.

Schatz, M., Orlova, E.V., Dube, P., Jager, J., van Heel, M. (1995) Structure of Lumbricus terresîris hemoglobin at 30A resolution using angular reconstitution. J. Struct. Biol. 114, 28-40.

Scheybani, T., Baurneister, W., Yoshimura H., Nagayama, K. (1994) Two-dimensional arrays of human C-reactive protein fomed on a dense aqueous subphase undemeath a lipid layer. Supramol. Sci. 1. 11 1- 116.

Schmitt, L., Dietrich, C., Tampé, R. (1994) Synthesis and characterization of chelator- lipids for reversible irnrnobilization of engineered proteins at self-assembled lipid interfaces. J. Amer. Chem. Soc. 116, 8485-8491.

Sqdzik, J., Blaurock, A.E. (1995) Myelin vesicles: what we know and what we do not know. J. Neurosci. Res. 41, 145-152.

Sqdzik, J., Blaurock, A.E., Hochli, M. ( 1984) Lipidlmyelin basic protein multilayers. J. Mol. Biol. 174, 385-409.

Sqizïk, J., Kirscher, D.A. (1 992) 1s myelin basic protein c~stallizable?Neurochem. Res. 17, 157-166.

Sela, B.-A., Bach, D. (1984) Calorimeûic studies on the interaction of gangliosides with phospholipids and myelin basic protein. Biochim. Biophys. Acta 771. 1 77- 1 82.

Serysheva, LI., Orlova, E.V., Chiu, W., Sherman, M.B.. Hamilton. S.L.. van Heel. M. ( 1995) Electron cryornicroscopy and angular reconstitution used to visualize the skeletal muscle calcium release channel. Nature Struct. Biol. 2, 18-24.

Shapira, R., McKneally, S.S., Chou, F., Kibler, R.F. (1971) Encephalitogenic fragment of myelin basic protein. J. Biol. Chem. 246, 4630-4640.

Shaw, C-M., Alvord, E.C., hby, S. ( 198 8) Chronic remitting-relapsing experirnental allergic encephalomyelitis induced in monkeys with homologous myelin basic protein. Ann. Neurol. 24, 738-748.

Shuman H., Chang C-F., Buhle Jr. E.L., Somlyo A.P. (1986) Electron energy-loss spectroscopy: quantitation and imaging. Ann. N. Y. Acad. Sci. 483, 295-3 10.

Sixl, F., Brophy, P.J., Watts, A. (1984) Selective protein-lipid interactions at membrane surfaces: a deuterium and phosphorus NMR study of the association of myelin basic protein with the bilayer head groups of DMPC and DMPG. Biochemistry 23. Smith, R. (1977a) Non-covalent cross-linking of lipid bilayers by myelin basic protein. A possible role in myelin formation. Biochim. Biophys. Acta 470. 170- 184.

Smith, R. (1977b) The secondary structure of myelin basic protein extracted by deoxycholate. Biochim. Biophys. Acta 491, 581-590.

Smith, R (1980) Sedimentation analysis of the selEassociation of bovine myelin basic protein. Biochemistry 19, 1826- 183 1.

Smith, R. ( 1982a) Self-association of myelin basic protein: enhancement by detergents and lipids. Biochemistry 2 1, 2697-270 1.

Smith, R. (1982b) 'H-nuclear magnetic resonance shidy of the association of the basic protein of central nervous system myelin with lysophosphatidylcholine. Biophys. Chem. 16, 347-354.

Smith, R. (1985) The encephalitogenic protein of myelin forms hexamers in which the polypeptides have a pleated-sheet structure. FEBS Lett. 183, 33 1-334

Smith, R. (1992) The basic protein of CNS myelin: its structure and ligand binding. J. Nerirochem. 59, 1589- 1608.

Smith, R., Cornell, B.A., Keniry, M.A., Separovic, F. (1983) "P nuclear magnetic resonance studies of the association of basic proteins with multilayers of diacyl phosphatidylserine. Biochim. Biophys. Acta 732, 492-498.

Sogin, M.L., Gunderson, J.H., Elwood, H.J., Alonso, R.A., Peattie, D.A. (1989) Phylogenetic meaning of the kingdom concept: an unusual rRNA from Giardia lamblia. Science 243, 75-77.

Somlyo, A.V., Shuman, H., Somlyo, A.P. (198 1) In Hutchison T.E., Somlyo A.P. (eds.). Microprobe analysis of biological systems. Academic Press, New York, pp. 103- 126.

Srivastava, S., Verschoor, A., Radermacher, M., Grassucci, R., Frank, J., (1995) Three- dimensional reconstruction of mamrnalian 40s ribosomal subunit embedded in ice. J. Mol. Biol. 245, 461 -466.

Stade, K., Jünke, N., Brimacombe, R. (1995) Mapping the path of the nascent chah through the 23s rRNA in the 50s ribosomal subunit. Nucleic Acids Res. 23,237 1- 2380. Stark, H., Mueller F., Orlova, E.V., Schatz, M., Dube, P., Erdemir. T.. Zemlin. F.. Brimacombe, R., van Heel, M. (1995) The 70s Escherichia coli ribosome at î3A resolution: fitting the ribosomal RNA. Structure 3, 8 15-82 1.

Stark, H., Orlova, E.V., Rinke -Appel, J.,Mueller, F.. Rodnina, M., Wintermeyer. W.. Brimacombe, R., van Heel, M. (1997) Arrangement of tRNAs in pre- and posttranslocational ribosomes reveaied by elecuon cryomicroscopy . Ce11 88. 8 19- 28.

Staugaitis, S.M., Colman, D.R., Pedraza. L. (1996) Membrane adhesion and other functions for the myelin basic proteins. BioEssays 18, 13- 18.

Steel, R.G.D., Torrie, J.H. (1980) Principles and procedures of statistics. McGraw-Hill Book Company, Toronto.

Stoffler, G., Wittmann, H.-G. (197 1) Ribosornal proteins XXV. Irnrnunological studies on Escherichia coli nbosomal proteins. J. Mol. Biol. 62, 407-409.

Stoffler-Meilicke, M., Stoffler, G., Odom, O.W., Zim, A., Kramer, G., Hardesty. B. (1981) Localization of 3' ends of 5s and 23s rRNAs in reconstituted subunits of Escherichia coli ribosomes. Proc. Natl. Acad. Sci. U.S.A. 78, 5538-5542.

Stoffler, G., Stoffler-Meilicke, M. (1983) Ultrastructure of macromolecular complexes studied with antibodies. In Tschesche, H. (ed.), Modern Methods in Protein Chemistry, Walter de Gmyter, Berlin, pp. 409-455.

Stoffler-Meilicke, M., Stoffler, G. (1990) Topography of the ribosomai proteins from Escherichia Coli within the intact subunits as determined by irnrnunoelectron microscopy and protein-protein cross-linking. In Hill, W.E., Dahlberg, A., Garrett. R.A., Moore, P.B., Schlessinger, D., Warner, J.R. (eds.), The Ribosome. Structure. Function, and Evolution. Amencan Society for Microbiology, Washington. DC. pp. 123-133.

Stoner, G.L. (1984) Predicted folding of B-structure in myelin basic protein. J. Neurochem. 43, 433-447.

Stoner, G.L. (1990) Conservation throughout vertebrate evolution of the predicted P- strands in myelin basic protein. J. Neurochem. 55, 1404- 141 l.

Stoops, J.K., Baker, T.S., Schroeter, J.P., Kolodziej, S.J., Niu, X.-D.,Reed, L.J. (1992) Three-dimensional structure of the truncated core of the Saccharomyces cerevisiae pymvate dehydrogenase complex determined from negative stain and cryoelectron microscopy images. J. Biol. Chem. 267, 24769-24775. Stoops , J.K., Cheng, R.H., Yazdi, M.A., Maeng, C.Y., Schroeter, J.P.. Kluppelberg. U.. Kolodziej, S.J., Baker, T.S., Reed, L.J. (1997) On the unique sûmctural organization of the Saccharomyces cerevisae pymvate dehydrogenase complex. J. Biol. Chem. 272, 5757-5764.

Stuart, B.H. (1996) A Fourier transform infrared spectroscopic study of the secondary structure of myelin basic protein in reconstituted myelin. Biochem. Mol. Biol. Int. 38, 839-845.

Surewicz, W.K., Moscarello, M.A., Mantsch, H.H. (1987) Fourier transform infrared spectroscopic investigation of the interaction between myelin basic protein and dimyristoylphosphatidylglycerol bilayers. Biochemistry 26, 3 88 1-3 886.

Svergun, D.I., Pedersen, J.S., Serdyuk, LN., Koch, M.H.J. (1994) Solution scattering from 50s ribosomai subunit resolves inconsistency between electron rnicroscopic models. Proc. Natl. Acad. Sci. U.S.A. 91, 1 1826-1 1830.

Symonds, C.P.(1924) The pathological anatomy of disseminated sclerosis. Brain 47. 36- 53.

Tanphaichitir, K. ( 1980) Multiple sclerosis associated with eosinophilic vasculitis. pencarditis and hypocomplementemia. Arch. Neurol. 37, 3 14-3 15. ter Beest, M.B.A., Hoekstra, D. (1993) Interaction of rnyelin basic protein with artificial membranes. Parameters governing binding, aggregation. and dissociation. Eur. J. Biochem. 211, 689-696.

Tompkins, T.A., Moscarello, M.A. (1991) A 57-kDa phosphatidylinositol-specific phospholipase C fiom bovine brain. J. Biol. Chem. 266. 4228-4236.

Tompkins, T.A., Moscarello, M.A. (1993) Stimulation of bovine brain phospholipase C activity by myelin basic protein requires arginyl residues in peptide linkage. Arch. Biochem. Biophys. 302, 476-483.

Tompkins, T.A., Moscarello, M.A. (1994) The rnechanism of stimulation of phospholipase C by myelin basic protein involves specific interactions. Biochim. Biophys. Acta. 1206, 208-2 14.

Trent, J.D., Nimmersgen, E., Wall, J.S., Hartl, F.U., Horwich, A.L. (199 1) A rnolecular chaperone from a thermophilic archaebacteriurn is related to the eukaryotic protein T-complex polypeptide- 1 . Nature 354, 490-493.

Trotter, J., DeJong, L.J., Smith, M.E. (1986) Opsonization with anti-myelin antibody increases the uptake and intracellular metabolism of myelin in intlammatory macrophages. 47, 779-789.

Tumminia, S.J., Mandiyan, V., Wall, J.S., Boublik, M. (1991) Heterogeneity of Escherichia coZi ribosomes established by scanning transmission electron microscopy. Biochimie 73, 9 19-925.

Tumminia, S.J., Hellrnann, W., Wall, J.S., Boublik, M. (1994) Visuaiization of protein- nucleic acid interactions involved in the in vitro assembly of the Escherichia coli 50s ribosomd subunit. J. Mol. Biol. 235, 1239-1250.

Ulmer, J.B. (1988) The phosphorylation of myelin proteins. Progr. Neurobiol. 31. 241- 259.

Utz, U., McFarland, H.F. (1994) The role of T cells in multiple sclerosis. Implications for therapies targeting the T ce11 receptor. J. Neuropathol. Exp. Neurol. 53. 351-358.

Valpuesta, J.M., Fujisawa, H., h4arc0, S., Carazo. J.M., Carrascosa. J.L. (1992) Three- dimensionai structure of T3 connecter purified fiom overexpressing bactena J. Mol. Biol. 224, 103- 1 12.

Vangui, P., Lee, E., Henkart, P., Shin. M.L. (1993) Hydrolysis of myelin basic protein in myelin membranes by granzymes of large granular lymphocytes. J. Irnmunol. 150, 2431-2439. van Heel, M., Stoffler-Meilicke, M. (1985) Characteristic views of E. coli and B. stearothermophilus 30s ribosomal subunits in the electron microscope. EMBO J. 4, 23 89-23 95. van Heel, M. (1987a) Similarity measures between images. Ultramicroscopy 21, 95-1 00. van Heel, M. (1987b) Angular reconstitution: a posteriori assignment of projection directions for 3 D reconstruction. Ultramicroscopy 2 1, 1 1 1- 124. van Heel, M. (1989) Classification of very large electron microscopical image data sets. Optik 82, 114-126. van Heel, M., Harauz, G., Orlova, E.V., Schmidt, R., Schatz, M. (1996) The next generation of the IMAGIC image processing system. J. Stmct. Biol. 116. 17-24.

Varghese, J.N., Colman, P.M. (1991) Three-dimensional structure of neurarninidase of influenza virus A/Tokyo/3/67 at 2.2 A resolution. J. Mol. Biol. 221, 473-486.

Vassallo, L., Elian, M., Dean, G. (1979) Multiple sclerosis in southem Europe II: prevalence in Malta in 1978. J. Epiderniol. Commun. Hlth. 33, 1 1 1 - 1 13. Verschoor, A., Zhang, N.-Y., Wagenknecht, T., Obrig, T., Rademacher, M.. Frank. J. (1989) Three-dimensional reconstruction of mammalian 40s ribosomal subunit. J. Mol. Biol. 209, 1 15- 126.

Verschoor, A., Srivastava, S., Grassucci, R., Frank. J. (1996) Native 3D structure of eukaryotic 80s ribosome: morphological homology with the E. cdi 70s ribosome. J. Ce11 Biol. 133, 495-505.

Voges, D., Berendes, R., Burger, A., Demange, P., Baumeister, W.. Huber. R. (1994) Three-dimensional structure of membrane-bound Annexin V. A correlative electron microscopy-X-ray crystallography study. J. Mol. Biol. 238. 199-2 13.

Wagenknecht, T., Grassuccl, R., Frank, J., Saito, A., [nui, M., Fleischer. S. (1989) Three- dimensional architecture of the calcium channellfoot structure of the sarcoplasmic reticulurn. Nature 338, 167- 170.

Walleczek, J.? Schüler, D., Stoffler-Melicke, M., Brimacombe. R.. Stoffler, G. (1988) A mode1 for the spatial arrangement of the proteins in the large subunit of the Escherichia coli ribosome. EMBO J. 11, 3 57 1-3576.

Walter, M.A., Gibson, W.T., Ebers, G.C., Cox, D.W. (1991) Susceptibility to multiple sclerosis is associated with the proximal immunoglobulin heavy chain variable region. J. Clin. Invest. 87, 1266-1273.

Wang, D.-N.,Sarabia, V.E., Reithmeier. R.A.F.. Kühlbrandt, W. (1994) Three- dimensional map of the dimeric membrane domain of the hurnan erythrocyte anion exchanger, band 3. EMBO J. 13, 3230-3235.

Wekerle, H., Kojima, K., Lannes-Vieira, J., Lassmann, H., Linington, C. (1994) Animal models. Ann. Neurol. 36, S47-S53.

Werner, T., (1939) Uber gemeinsames Vorkommen von Syringomyelie und multiplen Sklerose. 2. Ges. Neurol. Psychiat. 165, 4 16-428.

Wildhaber, I., Gross, H., Engel, A., Baurneister, W. (1985) The effects of air-drying and fieeze-drying on the structure of a regular protein layer. Ultramicroscopy 16.4 1 1- 422.

Williamson, R.T. (1894) The early pathological changes in disseminated sclerosis. Manchester Med. Chron. 19, 373-379.

Wood, D.D., Bilbao, J.M., O'Comors, P., Moscarello, M.A. (1996) Acute multiple sclerosis (Marburg type) is associated with developmentally immature myelin basic protein. Am. Neurol. 40, 18-24. Wood, D.D., Bilbao, J.M., O'Connors, P.O., Moscarello, M.A. (1996) Acute multiple sclerosis (Marburg type) is associated with developmentally immature myelin basic protein. Ann. of Neurol. 40, 18-24.

Wood, D.D., Moscarello, M.A. (1989) The isolation, characterization, and lipid- aggregating properties of a citnilline containing myelin basic protein. J. Biol. Chem. 264, 5 12 1-5 127.

Wood, D.D., Moscarello, M.A. (1996) The implication of post-translationai changes in myelin basic protein for demyelinating disease. In: Russell, W. The molecular biology of muhiple sclerosis, Wiley, New York.

Yada, R.Y., Harauz, G., Marcone, M.F., Beniac, D.R., Ottensmeyer. F.P. (1995) Visions in the mist: the zeitgeist of food protein imaging by electron microscopy. Trends Food Sci. Tech. 6: 265-270.

Young, P.Re,Vacante, D.A., Waickus, C.M. (1987) Mechanism of the interaction between myelin basic protein and the myelin membrane; the role of arginine methylation. Biochem. Biophys. Res. Commun. 145, 1 1 12- 1 1 18.

Zobel, CR., Beer, M. (1961) Electron stains 1. Chernical studies on the interaction of DNA with uranyl salts. J. Biophys. Biochem. Cytol. 10, 335.

Zobel, C.R., Beer, M. (1965) The use of heavy rnetai saits and electron stains. Int. Rev. Cytol. 18, 363. APPENDIX

THE RIBOSOME: A TEST OF ANGULAR RECONSTITUTION

Authors note: Pnor to the initiation of my Ph.D. 1 studied ribosome structure for my

M.Sc. My M.Sc. was based primarily on the 2D analysis and mode1 based 3D reconstructions of Escherichia coli and Thermomyces lanuginosus small ribosomal subunits. Ail of the raw uoanalyzed electron images that were used in the 3D reconstructions presented in this appendix were from my M.Sc. work. Al1 the 3D IQAD cornputer image analysis of this data was conducted at the onset of my Ph.D. research. 235

A.l: ABSTlRACT

Angular reconstitution of electron spectroscopic images of Escherichia coli and

Thermomyces lanuginosus nbosomal subunits produced direct information regarding ribosomal RNA architecture in the nbosomd subunits of prokaryotes and eukaryotes.

This investigation elucidated the domains of rRNA within the ribosomal subunits. These results are in good agreement with the protein distribution (Cape1 et ai.. 1988) within the nbosomal subunits. Over the past few decades ribosomal structure has been intensely studied by several groups (Frank et al., 199 1, 1995a; Lake et al., 1986; Stark et ai..

1995), thus its overail architecture is well charactensed. The good agreement of this data to that in the literature provides support for the IQAD (iterative quaternion-assisted angular determination) process as a method for angular reconstitution. This control expenment using a known structure (ribosomal subunits) was useful in the assessrnent of the validity of the MBP complexes studied in this thesis. Since IQAD solved the ribosome structure reliably, one can place more confidence in the MBP complex reconstructions as a 3D interpretation of the ZD electron images. 236

A.2: INTRODUCTION

Ribosomal RNA (rRNA) has central importance in the architecture of the ribosome complex and in protein synthesis (Brimacombe, 1995; Moore. 1988; Mueller et al.. 1995).

The different rRNA species and subunits are usually denoted by their predominant sedimentation coefficients: 16s in the prokaryotic small (30s) ribosomal subunit. 5s and

23s in the prokaryotic large (SOS) ribosomal subunit. 18s in the eukaryotic small (40s) ribosomal subunit. and 5S, 5.8S, and 28s in the eukaryotic large (60s) ribosomal subunit. where S is the "Svedberg" coefficient. Many biochemical and biophysical approachs have been used to probe the structures of the ribosomal complexes and the organisation of rRNA within them. Of the most relevance here are the computationai electron image analysis and reconstruction techniques (Frank, 1992; 1996). Science does not progress simply by new ideas, but also by new techniques. Note especially the sophisticated pattern analysis algorithm created in the fiamework of "single particle electron crystallography" (Frank, 1996). This appendix fira reviews the TEM approaches to probing the tertiary structure of rRNA in situ, and then descnbes in detail work using electron spectroscopie imaging (ESI) (Beniac and Harauz, 1993; 1995a; Beniac et al..

1997b-e) to study rRNA in a prokaryote (the cornrnonly studied eubacterium Escherichia di)and in a eukaryote (the mildly thermophilic fungus Thermomyces Zanuginosus).

Electron microscopy of ribosomal RNA in intact ribosomes

Well over a decade ago, the group fiom Professor Hoppe's laboratory at the Max-

Planck-Institut tur Biochemie in Martinsried, Germany, reconstructed individual E. coli 30s and 50s nbosomd subunits from tomographic tilt series (Knauer el al., 1983: Oetti et al., 1983). The complexes were negatively stained by embedding in uranyl acetate. and regions of high density in the reconstruction were postulated to represent rRNA that had been positively stained. This work is of histoncal interest because it challenged the microscopical and computational capabilities of the time. Visual interpretation of negatively stained ribosomes (Lake et al., 1986) has led to the construction of basic models of ribosome architecture characteristic of different species (Fig. A- 1). In another approach, Kühlbrandt and Unwin (1982) applied contrast variation to distinguish the rRNA fiom the protein within crystalline ribosomes fkom [izard oocytes. Using media of variable electron scattering density to match out the contribution to image contrast of protein or nucleic acid, it was found that the rRNA formed a dense central core with extensions to the surface, especially in the region of the subunit interface. Elsewhere. the dedicated scanning transmission electron microscope (STEM) at Brookhaven National

Laboratories was used to differentiate between signals from scattered electrons at large and small angles, enabling the approximate mapping of the distribution of the phosphate backbone of the rRNA in situ in the E. coli ribosomai subunits (Boublik et al., 2990:

Boublik and Wall, 1992). Irnmunoelectron microscopy has been invaluable in localising specific rRNA sequence segments (e-g.,Stoffler-Meilicke et al., 198 1 : Mc Williams and

Glitz, 1991).

Cryoelectron microscopy has provided the best information on ribosomal structure. and the Albany group (J. Frank's laboratory) has been the vanguard of this discipline.

Specifically addressing the question of rRNA organisation, Frank and coworkers (Frank Figure A-1: Schematic of ribosomal subunits. This figure shows three different types of ribosornal subunits. In (a) and (b) prokaryotic large and srnall ribosomal subunits are shown, and in (c) an eukaryotic ribosomal subunit is shown. These types of representations of ribosomal subunits are based on visual interpretation of negativriy stained electron images of ribosomal subunits (without any computer analysis) (Lake, et al., 1986). These "classic" types of representations were and still are essential in the interpretation of ribosome structure between species. The validity of the rest of the 3D reconstmctions in this appendix lies in their correspondence to the images in this figure.

Feanires indicated on the subunit s are :CP (central protuberance), LI (LI protuberance),

S (L7lL 12 stalk), h (head), f (foot), b (beak), and bl (basal lobes). 239 et al., 1991) perforrned a three-dimensional reconstruction of the 70s E. coli ribosome encased in vitreous ice. Quantitative manipulation of the reconstruction by density discrimination yielded a structure interpretable as being ribosomal RNA. with an intriguing potential rRNA junction at the subunit interface. Most recently. the Albany and

Berlin (M. van Heel's laboratory) groups working independently have achieved high resolution reconstmctions of E. coli 70s ribosomes ernbedded in vitreous ice, and these studies have set the gold standard for the field (Frank et al.. 1995a: 1995b: Frank and

Penczek, 1995; Stark et al., 1995; Agrawd el al., 1996; Lata et al.. 1996: Verschoor et al., 1996). It is difficult to envision that techniques other than cryoelectron microscopy

might still be worthwhile to purnie in nbosomology.

Electron spectroscopic imaging

Nonetheless, it still cm help to look at biological macromolecules in a different way. Dark field EM and electron spectroscopic imaging (ESI) (Ottensmeyer and Andrew.

1980; Ottensmeyer, 1982; 1984) are two different modes of imaging macrornolecuIar complexes. Electron spectroscopic imaging uses a standard transmission electron microscope which is equipped with an energy filter below the objective lens (Ottensmeyer.

1984). By using only electrons which have lost a specific arnount of energy due to specific imer-sheil ionisation interactions with the specimen, it is possible to form an elemental rnap of the specimen. In nucleoprotein complexes, a map of the phosphoms distribution represents primarily a projection of the phosphate backbone of the nucleic acid component, since the arnount of phosphoms in (any) phosphorylated proteins is 240 udly much mialler. Since ESITsinception, important structural and fùnctional data have been yielded by the ability to visualise selectively to high resolution the nucieic acid components of nucleosomes, ribosomes, the signal recognition particle. the TFIIIA-5s rRNA complex, DNA, and sundry other protein-DNA complexes. In particular. the visualisation of rRNA in situ in the intact ribosomal subunit by ES1 was demonstrated well over a decade ago to be feasible (Boublik et al, 1984; Kom et al.. 1983).

Unfortunately, these resuits have never been fully accepted by the scientific community. in main because of the unavoidably high radiation doses inherent in the technique. measurements of which were published long ago (Bazett-Jones et al., 1988). ES1 is a microanalytical technique not just an imaging one. The higher irradiation demands of the technique are for localising and identifjring the phosphorus signal.

Electron spectroscopic imaging can be a useful and complementary technique to vitreous ice imaging of nucleoprotein complexes. In principle, the most obvious advantages of ES1 to the study of ribosome structure are a potential resolution of 0.3-0.5 nm of unstained specirnens, with a minimum detectable mass of 20 to 50 atoms of phosphorus (Adamson-Sharpe and Ottensmeyer, 198 1; Bazett-Jones, 1993; Bazett-Jones and Ottensmeyer, 198 1). Over many recent years, this lab has applied quantitative image analysis to ES images of ribosomal subunits (Beniac and Harauz, 1993; 1995a; Beniac er al., 1997b-e). A.3: MATERIALS AND METHODS

Isolation of ribosomal subunits

Escherichia coli (strain AB 264) and Thermornyces lanuginosus (ATCC 16435) ribosomai subunits were isolated by sucrose density gradient centrifugation. Al1 procedures were continuous from ce11 lysis to fieeze-drying to ensure particle integrity.

Under no circumstances were sarnples fiozen between steps. The sarnples were constmtly maintained at 4OC. and RNase and protein inhibitors were used to ensure that enzymatic degradation did not occur. After pooling of gradient fractions enriched in srnall or large ribosomal subunits, the sucrose was removed and the sait concentration of the buffer was reduced by dialysis against a 1 1 volume of dialysis beer for a 12 h period with several buffer changes. Shortcuts here were definitely deleterious, with any remaining sucrose or salts forming disturbing menisci around the pure particles. E. coli ribosomal subunits were didysed against 10 mM HEPES-KOH (pH 7.4), 60 mM KCI. 2 rnM

Mg(CH,C00)2.4H,0,10 rnM 2-mercaptoethanol. T. lanuginosus ribosomai subunits were dialysed against 10 mM HEPES-KOH(pH 7.5),0.9M KCI, 12 mM Mg(CH3C00),.4H,0.

10 mM 2-mercaptoethanol.

Preparation for electron microscopy

Ultrathin carbon film (-5-6 nrn thick) was produced by indirect evaporation ont0 mica, scored and floated off onto the surface of sterilised ddHzO (distilleci deionised water), and picked up with precleaned 1000 mesh copper or nickel grids. Both types of 342 grids were used to be able to fieeze-dry two different specimens at the same time. The coated grids were wicked dry, and allowed to stand for a 12 h period to allow the carbon to adhere firmly to the support. Omission of this 12 h eing step resulted in the carbon film detaching fiom the support grid in the subsequent liquid propane plunging step. A

25 pl drop of the dialysed nbosomal subunit enriched sarnple (adjusted to an Az,dml of

0.2) was applied to the carbon film side of a carbon film - 1000 rnesh grid. The sample was allowed to adsorb to the carbon film for 60 s and the excess sample was removed by wicking with filter paper; this step was followed instantly by application of a 25 pl droplet of stede ddH20 to rime the sample. This wicking-washing procedure was repeated several times to rime the sample thoroughly. The grid with the sample was then held in a pair of reverse tension forceps, wicked almost dry. and irnrnediately plunged into a bath of liquid propane suspended in a cup in a bath of liquid nitrogen (-196°C). The fiozen sample was then transferred into a copper tray suspended in liquid nitrogen: the copper tray (with gnds in it) was then transferred to a pre-cooled (- 196°C) cryo-stage of a Balzers 360M fieeze fracture unit. The grids were then transferred to the cryo-stage with pre-cooled forceps, the copper tray was removed, and the freeze etch unit was sealed.

A high vacuum was then drawn (10" Torr), while the sarnple was kept at - 196OC. Over a half hour penod the sample was gradually warmed to -84OC; this step was followed by a 2 h drying period at -84°C. The sample was then gradually brought up to room temperature, removed fiom the Balzers fieeze etcher, and stored under vacuum in a bel1 jar until viewed in the electron microscope. -743 Electron spectroscopic imaging

ES1 was perfomed in a Zeiss EM902 electron microscope equipped with a

Castaing-Henry type energy filter, at a magnification of 30.000 X and at an accelerating voltage of 80 keV. The energy selecting slit was set at a width of 18 eV. Two micrographs were then taken of previousiy unillurninated regions of the specimen at 150 eV and 100 eV energy Iosses. The 100 eV image produced non-specific mass density images of the subunit, and the 150 eV image represented a phosphorus enhanced image of the sample superimposed over a non-specific mass density. The 150 eV image was recorded f~stbecause it was considered to be the more important one. i. e.. a phosphorus enhanced image. The term "doublets" refers to pairs of images of the specimen recorded at different energy losses. Electron micrographs were visually exarnined and those of best optical quality (focus, astigmatism, optical density within the linear response range of the emulsion) were chosen for subsequent analysis. Electron micrographs were digitised using an Optronics rotating drum densitometer; a 1200 X 1200 pixel area kvas acquired using a 25 pm aperture, with a corresponding scanning step size of 25 Fm. During digitisation. the emulsion of the negative was kept facing into the dnun to preserve handedness in the images. Al1 micrographs were mounted so that doublet images (Le.. at 150 and 100 eV loss) would be scanned over the sarne central region of the micrograph to facilitate computer alignrnent of doublet members. Digitisation was restncted to the center of the micrograph to maintain the same energy loss range fkom image to image.

Digitised images were then transferred to an IBM RISC System/6000 PowerStation

220 for analysis within the framework of the IMAGIC-V syaem (van Heel et al., 1996). 244

The digitised ES images were pretreated with a 3 X 3 pixel median filter to suppress some of the image noise, followed by interactively selecting matching ribosomal subunits within each image, using a mouse and cursor. The centen of mass of the 200 or so matching ribosomd subunits were used as fiduciai points to calculate the rotation and translationai shifts necessary in order to bring the paired images into register with one another. Individual smdl(904 for T. lanuginosus and 927 for E. coli) and large (789 for

T. lanuginosus and 1 105 for E. coli) nbosomd subunits were then interactively selected. i.e., subimages of size 64 X 64 pixels containing a single ribosornal subunit centered within them were selected from the larger 150 eV and 100 eV loss images.

The first step in producing the net phosphorus (NetP) image from the doublet of each particle was the removal of the background density due to the photogaphic emulsion by subtraction of the average density measured fiom a corner of the negative. Next. the extrapolation of the energy loss spectnim under the ionisation edge was calculated using the relationship 1 = AE~,where 1 is the intensity, E is the energy loss. and A and R are constants. In a so-called two parameter fit, the values of A and R are calculated for each point in the image, and two reference images (below the ionisation edge) are required

(Bazett-Jones, 1993; Bazett-Jones et al., 1988; Ottensmeyer, 1986a; I986b). Here. a one parameter fit was used, with a value of R constant over the whole image and only A varying, since only one reference image was taken deliberately to limit the cumulative electron dose. The net phosphorus map was calculated by subtracting the extrapolated

100 eV loss image from the 150 eV loss image. ALthough the number of phosphorus atoms is only about 2.5% of the total number of carbon-like atoms, the inelastic scattering 245 cross-section of phosphorus is about 15 times stronger than carbon at 150 eV (Egerton.

1988; 1996; Heng et al., 1990; Ottensmeyer, I986a; 1986b) and so a large NetP signal is realised (Bazett-Jones, 1993 ; Bazett-Jones and Ottensmeyer, 198 1 ).

For a strong NetP signal such as from a ribosome, which constitutes about 20-30% of the background signal (cf, Heng et al., 1990), the one-parameter approximation would introduce a systematic error of no more than 15% while reducing the random error by

40% compared to the two-parameter fit. Obviously, for smaller elemental concentrations. more complex 3 or 4 window methods are required (Egerton, 1996). The integrated opticai densities over corresponding background areas comprising only carbon film were used to determine multiplicative factors to correct for the slightly unequal exposures of each doublet member. It was deterrnined empincally that more satisfactory NetP images were obtained by performing the above calculation for each particle separately, using the aaacent carbon film density to compute background correction factors, rather than doing the calcuiation over a large area using an average carbon film density.

2D image pretreatment

Each of the data sets (T.lanuginosus and E. coli ribosomal subunits) were reduced to 300 particles each and were treated separately. The particles were chosen solely on the bais of their structural integrity and isolation from other complexes which might interfere with the analysis. Since the 150 eV loss images contained stronger and more distinctive signals than the corresponding 100 eV loss ones, and were inherently less noisy than the corresponding NetP maps, al1 IQAD calculations were applied to the first population. 246

AI1 images had aiready undergone a 3x3 pixel median filtering. A 2x2 zoom factor was then applied to each image to make each ribosomal subunit comprise a greater number of pixels within each image, to reduce subsequent interpolation errors due to effects of digitisation at a given pixel size. The subsequent set of operations constitute a pretreatment procedure which will be referred to as "farrow-segmentation" afier its onginator (Farrow and Ottensmeyer, 1992; 1993). First, the pixel of maximum intensity was deterrnined in the center quarter of each image. Starting fiorn this point. neighbouring pixels were marked and a contiguous region was thus "grown" outwards until the integrated intensity of ail pixels within the neighbourhood were a specified fraction (here, 113) of the integrated intensity of al1 pixels within the entire image. This operation served to distinguish the particle fiom its surrounding background which represented only carbon film. Previously, we had used objective thresholding algonthms focussed on discrirninating "object" fiom "background" on the basis of grey level distributions (Beniac and Harauz, 1995b). The advantage of farrow-segmentation was the definition of a contiguous image region representing only the particle of interest.

After farrow-segmentation, al1 pixels outside the particle neighbourhood were set to a value of zero, and the value of the smallest positive gray level less 1 was then subtracted from the particle pixels. This way, the average (in the vicinity of the particle) density contribution of the carbon film was removed fiom the analysis. Then. al1 non- zero positive density values were normalised over the entire image population, being multiplied by a positive factor so that each particle had the same total intensity. By normalising al1 the images in this way, intensity differences were suppressed, enabling a better comparison of di fferent images to one another.

Three-dimensional reconstruction of ES images

The problem of determining the relative orientations of non-crystalline randomly oriented biologicai macromolecules has been a long-standing one. Here. a quaternion- assisted approach of angular reconstitution was used (Farrow and Ottensmeyer. 1992:

1993). This method exploits sinograms and sinogram correlation functions and the

Central Axis (or Projection) Theorem to determine the relative angular orientations of the ribosomai subunits from their how-segmented 150 eV Ioss images. This theorem States that projection images of a particular macromolecule at different orientations share common lines of integrated intensity. By determining these common lines it is possible to determine the relative angular orientations amongst many projection images. The initial orientation angles of the reconstructions were refined iteratively by a quaternion vector approach as described previously by Farrow and Ottensmeyer (1992: 1993). The method is robust, has been tested exhaustively with simulated images, and applied recently to structure determination of numerous proteins and nucleoproteins (e-g.. Beniac et al..

1997a-c; Czarnota et al., 1994; 1996). Mutual correfation fiuictions between sinograms were employed instead of cross-correlation functions. Reconstruction was performed using filtered back-projection with exact filters (Harauz and van Heel, 1986). 248 A.4: RESULTS AND DISCUSSION

Fields of view of the 150i9 eV loss images of quick-frozen and freeze-drird E.

ribosomal subunits were digitized and used for image analysis. The complexes are well-delineated with sharp boundaries and are non-aggregated. These images are cil quality comparable to the best freeze-dried ribosomal preparations imaged by low-dose

STEM (Tumminia et al., 199 1 ; 1994), and were superior to those of ribosomes prrparsd in other ways (Beniac and Haraw 1993; cf., Korn, 1980). Ernbedding the particles in vitreous ice (cryoEM) (Harris, 1997) was not feasible since the enerpy filtering transmission electron microscope (the Zeiss EM902) was not equipped with a cryo-stage.

On the basis of the literature, freeze-drying appeared to be the brst alternative structure- presewing preparative technique that O bviates major distortions due to dry ing from aqueous solution (Andrews et al.. 1987; Baumeister, 1982; Engel and Meyer, 1980;

Fowler and Aebi, 1983; Wildhaber el al., 1985). The local rearrangement of flexible peptide loop structures is sometimes the most serious problern that can be expected upon removal of the ice environment by sublimation (cf:, Czarnota et al., 1994). An additional benefit is the removal of water that can contribute to radical formation and radiation darnage in vitreous ice. Future studies using neutral negative stains such as methylamins vanadate, or sugary embedding media as preservatives, might be feasible providèd that these materials did not interfere with the elemental or background signals and could withnand the high electron doses (Bremer et al.. 1992; Harris, 1997).

The reproducible spatial resolutions between independent 15019 eV loss and NetP reconstructions were at least 4 nm as determined by Fourier Shell Correlation (Frank. 249 1996; van Heel, 1987a). This result is a significant improvement over our previous roughly 6 nrn resolution reconstructions which did not use quaternion-assisted angular reconstitution (Beniac and Haraq 1995a). Such resolutions are remarkable considenng that they are obtained using a technique of elemental microanalysis which necessitates a total dose of 2x10' electrondd to obtain the phosphorus signal within the ribosomal subunits. Purely structural information could have been derived at Iower doses (e.g.. using the elastic dark-field signal), but the direct relationship between the total structure and the phosphorus distribution within it would have been compromised. It should be realised that the electron doses required by this type of microanalytical imaging are favourable when compared with other microanalytical approaches, such as X-ray microanalyses, which need doses often more than one order of magnitude larger at spatial resolutions ten times worse. One of the ben results obtained by the latter technique is the dot maps for phosphorus and calcium of the sarcoplasmic reticulum. taken in a 34 hour exposure with an estimated dose of 108 electrons/nm2 at a resolution of 20 nm (Somlyo et al., 1981). In electron energy loss spectroscopy and imaging, there are instances of microanalysis at higher doses, of virus particies at 10' electrons/A2 (log electrons/nm2)

(Shurnan et al., 1986) and of other biological structures at doses of up to 10" electrons/nm2(Engel and Colliex, 1993). A cornparison of electron spectroscopie imaging with pureiy imaging techniques such as cryoelectron rnicroscopy at 10' electrons/nm2 is pointless and misleading. Nonetheless, it is acknowledged that some loss of fine structural detail due to radiation-induced structural alterations even at only 2x10' electrons/nm2 is an unavoidable price to be paid to obtain information on elemental distribution. 250

Recently, a number of molecular models of rRNA structure have been derived on the bases of biochemical data or computational results (e.g.,Brimacombe. 1995; Malhotra et al., 1993; 1994; Mueller et al., 1995). It is too large a conceptual leap to compare these near-atomic models with the results presented here. and better instead to consider the cryoEM reconstructions of ribosomal subunits to be a "gold standard" (viz.. Frank et al., 1995a; 1995b; Lata et al., 1996). Despite the similar values of the reported resolution estimates of these three-dimensional reconstructions and the cryoEM ones, the degree of structural detail discemible in the latter is far better. As mentioned above, there will be some alteration of structural detail in the ES1 data, due to the necessarily higher microanaiyticai doses. The numerical discrepancy might be due to the conservativeness of certain resolution measures compared with others (van Heel, 1987a).

The three-dimensional reconstructions of the E. coli large and small ribosomal subunits, from 150+9 eV loss and the corresponding NetP images. are shown superimposed together in Figs. A-2 and A-4, respectively. In the separate subunit reconstructions, the canonical "left- and nght-lateral" (Figs. A-2 (a, c, g. i)). "quasi- symrnetric" (or "asymmetric") (Figs. (A-2 (b, d, h, j)) "crown" (Figs. A-4 (a g)), and

"kidney" (Figs. A-4 (b, h)) views seen in negatively-stained preparations can be recognised (e-g., Harauz et al., 1987; van Heel and Stoffier-Meilicke, 1985). This consistency with previous EM work of the present reconsmictions from electron spectroscopie images is positive and supports the validity of the phosphorus and mass reconstructions. The exterior surface of the E. coli large subunit reconstnictions seems somewhat sphencal and smoother in part due to the resolution achieved, yet reveals also Figure A-2: 3D reconstmction of E. coli 30s ribosomai subunit. Surface repressntation of 3D reconstmctions from 150 eV loss and NsrP images of E. cofi srnaIl ribosomaI subunits. Green is used to represent the reconstructions from 150 eV loss. grange is used to represent the reconstructions frorn YetP images. Panels (a-d) show the small subunit reconstmction rotnting about its long a~isat 90° intervals. and ce) and < f> shcw the reconstmction t'rorn above and below. respeccively . The only differencr berw ccn

(a-f) and (g-l) is that the 150 eV loss reconstruction is ponrayed as n solid green surface in

(a-fl and as a green wire mesh in (g-li. This dual method of display facilitates viewing of the NetP (rRNi\) distribution within the mus of the subunit. as well as highlighting whcre the rRNX appears to reach the surface. Features indicated on the small subunit arc : h t head), f (foot), and ch (,channe[). Figure A-3: 3D reconstruction of E. coü 30s ribosomal subunit and 3D location ofribosomal proteins. Surface representation of 3D reconstructions from 150 eV loss and NetP images of E. coli small ribosomal subunits. Green is used to represent the reconstructions from 150 eV loss.

Orange is used to represent the reconstnictions from NetP images. The four views in each row show the small subunit reconstruction rotating about its long axis at 90° intervals. In (a). both reconstructions are represented as solid surfaces. In (b) only the NetP reconstruction is shown. In

(c). both reconstructions are represented as wire models. Spheres represent positions of proteins mapped by neutron scattering (Cape1 et al., 1988). The size of each sphere is proportional to the anhydrous volume of the protein represented. The red and yellow colours are intended only to assist visualisation. and follow the same scheme of Capel et al.. (1988), as does the numerical labelling of the proteins. In the first column, the inter-subunit interface of the small subunit is facing the viewer, which interpretation is supported by the presence of the interface protein

S20L26 (Stade et al., 1995; Stoffler-Meilicke et al., 198 1 ; StoMer-Meilicke and Stoffler, 1990).

The bottom of the small subunit is protein deficient (Cape1 et al., 1988; Stoffler-Meilicke and

Stoffler, 1990) and here is seen to be phosphorus-rich (Le.. rRNA -rich).

Figure A43D reconstmction of E.coli 50s ribosomal subunit. Surface

representation of 3D reconstructions from 150 eV loss and NetP images of E. coli large

ribosomal subunits. Blue is used to represent the reconstructions from 150 eV loss. Red

is used to represent the reconstructions from NetP images. Panels (a-1) use the same orientations, and use of solid and wire surfaces as described in Fig. A-2. Features

indicated on the large subunit are : CP (central protuberance). S (L71L12 stalk). E

(putative nascent protein exit site), PT (putative peptidyl transferase site (Stoffler and

Stoffler-~eilicke,1W)), and L 1 (L 1 protuberance).

256 substantial depressions (e-g.,Fig. A-4c) and a channel representing the putative exit site of the nascent polypeptide chah (Fig. A-4d).

The density distributions of the mass fiom the 150+9 eV loss images. and of the corresponding NetP signals, are different. The NetP is clearly not uniformly dispersed throughout either subunit. In the small subunit NetP reconstruction. a channel appears in the "neck" region between the head and the body domains. In the large subunit NetP reconstruction, a channel appears in the body of the cornplex. These features in the NetP reconstructions correspond to similar ones seen in cryoEM reconstructions of the total mass (Frank et al., 1995a; 1995b). Furtherrnore, a visual congruence is apparent with a consensus mode1 for the large subunit based on low angle X-ray scattering data (Svergun et al., 1994). A comparison of reconstructions from electron spectroscopic images with data fkom negatively-stained and vitreous ice-embedded preparations requires a degree of caution, however. The conditions of preparation and imaging vary widely in terms of degree of distortion of the sample and of electron dose. The 30s ribosomal subunit is also known to be conformationally flexible. In this light, the structural similarities amongst results from various techniques to the ones presented here are affirmations of the present results.

Fig. A-3 presents a comparison of the reconstruction of the small ribosomal subunit with resuits fiom neutron scattering and immunoelectron microscopy (Cape1 el al..

1988; Stoffler-MeiIicke and Stoffler, 1990). The bottom of the subunit, as revealed from spectroscopic electron microscopy and three-dimensional reconstruction, is phosphorus- nch (16s rRNA-rich). This resdt is wholly in accord with the neutron scattenng and 257

immunoEM studies that showed this domain to be protein-poor.

Fig. A-3 also shows two structural landmarks of the 30s subunit that are more

apparent here than in the perspectives in Fig. A-2. One of these landmarks is the

"platform", above which is the anticodon recognition site of the 16s rRNA. Another of

these landmarks is a pointed tip in the base, which corresponds to a feature called the

"extension in the foot region" by Penczek et al. (1994), the "spur"by Frank et al. ( 1995a:

1995b), and "toe" by Stark et al. (1995).

The 150i9 eV loss reconstructions of both subunits were merged computationally

to give a representation of the whole ribosome in Fig. A-5, with the large and small

subunits positioned as they would be during translation. The independent reconstructions

seem to have a naturai fit when portrayed in this way. The merger was guided using

published data on neutron scattering and imrnunolocalisation of proteins and rRNA sequence segments (Cape1 et al., 1988; May et al., 1992; Nierhaus. 1991: Stade et (il..

1995; Stoffler and Stoffler-Meilicke, 1983; Stoffler and Wittmann. 197 1 ; Stoffler-Meilicke and Stoffler, 1990; Stoffler-Meilicke et al., 1981; Walleczek et al., 1988). If one

interprets the E. coli ribosome reconstmction as a biologically functional entity (Fig. A-6). then there is a large central cavity which provides ~~cientroom to contain 7 or 3

molecules of transfer RNA (tRNA) and the necessary ancillary protein factors. The cavity is also accessible to the exterior for admission and egress of these small macromolecules.

The messenger RNA (mRNA) can pass through the hole in the neck of the NetP reconstruction of the srnail subunit, in agreement with the interpretations of cryoEM reconstructions (Agrawal et al., 1996; Frank et al., 1995a; 1995b; Lata el al., 1996). The Figure A-5: 3D reconstruction of E. coli 30s and SOS ribosomal subunits. Surface representation of the merged E. coli ribosomal subunits using the same colour schemes and viewing directions as in Figs. A-2 and A-4. The merging was guided by the overall sirniiarities of these reconstructions to those in Frank et al. (1995a, I995b), as weiI as on the positions of an interface protein (S20L26) as shown in Fig. A-3. The rRNA appears to form a dense central core, reaching the surface in a manner similar to that defined in

Kühlbrandt and Unwin (1982) by contrast variation using aurothioglucose and glucose embedding.

Figure A-6: 3D reconstruction of E. coli 30s and 50s ribosomal subunits, shown as a functional entity. Montage of merjed reconstructions from E. coli ribosomal subunits showin; the putative positions of tRNA's. mRNA and nascent polypeptide chain. Panels (a-d) are views of the complex related by 90° rotations about the vertical axis of the page. Panels (e. 0 are both top views. differing only that (0 is a cutaway view of the ribosomal reconstruction depicted as a solid surface. Panel (fl illustrates the interior cavity in which the tRNA's are contained. The colour schsms is as iollows: blue mesh or surface - 50s subunit, 150 eV loss: green mesh or surface - 30s subunit.

150 eV loss: red surface - 50s subunit. Net P: orange surface - 30s subunit. Net P: purple ribbon - mRNA; orange ribbon - nacient peptide; magenta tRNA - aminoacyl tWA: yellow RNA - peptidyl tRNA brown tRNA - exit tRNA. 26 1 hole in the back of the reconstructions of the large subunit can potentially be the outlet for the nascent polypeptide (Frank et al., 1995a; 1995b).

Finally, the 7'. lanuginonrs (eukaryotic) small ribosomal subunit reconstructions

(Fig. A-7) (Beniac et al., 1997b; Harauz and Flannigan. 1992) are also meaningful. It has been proposed that the extra non-core rRNA which has been identified by phylogenetic comparison of small ribosomal subunit rRNA secondary structure is deposited in the bill and basal lobe regions of eukaryotic small nbosomal subunits (Brimacombe et al.. 1990). nie prokaryotic (E. coli) and eukaryotic (T. lanuginosus) three-dimensional reconstructions clearly show sirnilarities and differences in the head and base domains

(Fig. A-8). In the head, there is a prominent bill on the eukaryotic small nbosomal subunit. In the base, the eukaryotic small subunit has a prominent basal lobe. whereas the prokaryotic small subunit does not. These features have been noted widely in the

Iiterature (Harauz and Flannigan, 1 992; Kyle and Harauz, 1993; Verschoor et al.. 198 9). and their exact compositions (whether protein or rRNA) have oniy been speculated upon.

It hm been proposed that the extra non-core rRNA which has been identified by phylogenetic comparison of small ribosomal subunit rRNA secondary structure is deposited in the bill and basal lobe regions of eukaryotic small nbosomal subunits

(Brimacombe et al., 1990). This investigation has provided direct evidence supporting this hypothesis.

It is aiready remarkable that the reconstructions of small ribosomal subunits from the two species cm be related in this matter. The two analyses were done independently - the results from one were not used as a mode1 to influence (or bias) the other. It is Figure A-7: 3D reconstmction of T. Innuginosus 40s ribosomal subunit. Surface representation of 3D reconstructions from 150 eV loss and NetP images of T. lanuginosus small nbosomal subunits. Blue is used to represent the reconstructions from

150 eV loss. Red is used to represent the the reconstnictions from NetP images. Panels

(a-1) use the same orientations. and use of solid and wire surfaces as described in Fig. A-

2. Features indicated on the small subunit are : b (beak), and bl (basal lobes).

Figure A-8: Cornparison of the stmctures of the E. coli 30s (prokaryotic) and T.

Zunuginosus 40s (eukaryotic) small ribosomal subunits. 3D reconstmctions from

NetP images from (a) E. coli , and T. lanuginosus (b)smal1 ribosomal subunits. In (c) the NetP reconstructions are superimposed, and in (d) the E. coli (gold), and T. lanuginosus (blue) 150 eV loss reconstn.tctions are superimposed. The four views in each row show the small subunit reconstructions rotating about their long ais at 90 O intervals. There are similarities in the head and base regions, as well as the presence of a channel through the neck. in both reconstructions. The extra NetP mass in the beak and basal lobes of the eukaryotic subunit. clearly depicted in (c), has previously been suggesied to be where "extra" phylogeneticaily conserved rRNA is Iocaiised

(Brimacombe et al., 1990).

266

difficult to gauge with the present data the importance of more subtle dissimilarities in

other regions of the two reconstructions. Future work involving advanced statistical analyses of such reconstructions (Liu et al., 1995; Liu and Frank. 1995) would provide

a quantification of significance of observed features and differences. A.5: CONCLUDING REMARKS

This investigation produced direct 3D information regarding ribosomal rRNA

localisation in the nbosomal subunits of E. coli and T. lanuginosus. The present E. coli results are in favourable agreement with reconstructions derived by cryoTEM.

Cornparison of the E. coli and T. lanuginosus small ribosomd subunit 3D reconstructions

have begun to shed light on the whereabouts of the "extra rRNA" present in eukaryotic

ribosomes.

With respect to ribosomes per se, many unanswered questions remain in this field.

from the structures of eukaryotic (Srivastava et al., 1995) or thermophilic (Harauz. 1992:

Harauz et al., 1992) or unusual rRNAs (Sogin et al.. 1989) to the issue of eocytes and photocytes (Lake et al., 1986). Ribosomology has only just begun a new stnictural phase. and ES1 can contribute via direct RNA mapping as a complement to cryoEM. and rRNA sequence analysis of these complexes.

In closing, this appendix has demonstrated that the IQAD process is a effective method by which to convert 2D electron image data into a 3D reconstnicted volume. The congruence shown here between the three separate ribosome data sets, and the current literature illustrates this point. Since IQAD is a very new technique (Farrow and

Ottensmeyer, 1992, 1993), this study provides support for IQAD as an accurate method of 3D structure determination. IMAGE EVALUATION TEST TARGET (QA-3)

APPLIED -& IMAGE. lnc -= 1653 East Main Street -. , --, Rochester. NY 14609 USA ------Phone: 71 6/48XUCJû ---- Fax: 7161288-5989

O 1993. Wied Image. Inc.. All RqhB Reserved