Cell Tissue Res (1988) 252:249-262

andTissue Research 9 Springer-Verlag 1988

In vitro polymorphism and phase transitions of the neurofilamentous network isolated from the giant axon of the squid (Lofigo peMei L.) J. Metuzals ~, H. Pant 2, H. Gainer 2, P.A.M. Eagles a, N.S. White 3, and S. Houghton 4 Department of Pathology, Faculty of Health Sciences, University of Ottawa, Ottawa, Ontario, Canada; z Laboratory of Neurochemistry, NIH, NINCDS, Bethesda, Maryland, USA; 3 Department of Biophysics, King's College, University of London, England; 4 Marine Biological Laboratory, Woods Hole, Massachusetts, USA

Summary. Using electron microscopy (EM), optical diffraction and image reconstruction techniques, we have demonstrated polymorphism of neurofilamentous network (NFN) in vitro based on phase transitions of the protein assemblies. The specific polymorphic appearances depended upon a number of factors, such as K +, Mg z+, Ca 2+ ions, as well as the charge and hydration state of the molecules. Furthermore, modifications initiated by the state of phosphorylation of the sidearm proteins played an important role, especially in determining the sidearm disposition of the NFN. The CaZ+-activated protease removed the sidearms. Other enzymes activated by Ca 2 + may initiate new association patterns of the peptide remnants and the intercoiling of two smooth neurofilaments (NFs) into paired helical filament-like (PHF-like) strands. Prolonged storage of the isolated NFs in Rubinson-Baker solution resulted in autocrosslinking and intercoiling of modified N F N components. The in vitro polymorphism and phase transitions of squid N F N induced under controlled conditions have been compared to modifications of cytoskeleton observed by EM in frontal lobe biopsies of Alzheimer patients. We conclude that similar processes, as induced in vitro, do occur in neurons of Alzheimer patients. Key words: Neurofilaments - Phase transition - Paired helical filaments - Paracrystals - Alzheimer's disease - Squid (Loligo pealei L.) Human

Since the early work of Cajal (Cajal 1952), a considerable amount of evidence has accumulated supporting the conclusion that the neurofilamentous network (NFN) and the intermediate filaments, in general, are dynamic and plastic structures. The N F N undergoes constant structural and chemical modifications and rearrangements during the lifetime of the neuron (Willard 1983; Shaw and Weber 1982; Pachter and Liem 1984). The assembly patterns of the N F N components are in a constant state of reversible modification between ordered and less ordered architecture (Metuzals et al. 1981 ; Metuzals et al. 1988a, b). Results obtained in several laboratories have provided evidence that the Send offprint requests to: Prof. J. Metuzals, Department of Pathology, Faculty of Health Sciences, University of Ottawa, 451 Smyth Road, Ottawa, Ontario, Canada K1H 8M5

order within the N F N is determined by its crossbridges and sidearms (Metuzals et al. 1982a; Metuzals et al. 1983c; Geisler et al. 1984; Weber et al. 1983; Metuzals and Eagles 1986). The coiled-coil a-helices of the middle rod domain of the NF protein seem unable to form 10-nm filaments by themselves. The non-a-helical terminal domains of the N F proteins, corresponding to the sidearms of the NFN, seem to be involved in the transition of the protofilaments into the full network structure (Weber et al. 1983). These observations have led us to the following paradigm. During the massive alterations of the cytoskeleton that occur in Alzheimer's disease (AD), the state and disposition of the sidearms of the N F N play an important role (Metuzals 1986a). They determine the character of the bundling and the helical and solubility properties of the filaments in the neurofibrillary tangles. This paradigm is also supported by the fact that the 200 K N F protein has a hybrid structure (Geisler et al. 1984). It consists of a constant, predominantly a-helical domain and a variable domain that protrudes from the core of the NF. In addition, the sidearm proteins may control the assembly state of the NF proper. To test this paradigm we have treated, in vitro, the N F N isolated from squid giant axon, with factors that are known to modify the tail portion of the 200 K NF protein and the high molecular weight (mol. wt.) protein NF-I copurifying with the NFN. Such well-known factors are Ca 2 § vated proteolysis and phosphorylation/dephosphorylation (Gallant et al. 1986; Pant et al. 1986). It can be expected, therefore, that the modified N F N components will reassemble in new architectural patterns composed of modified and crosslinked components (Metuzals et al. 1986). In a series of dialysis experiments we have investigated reassembly potentialities of the normal and modified components of the N F N (Metuzals and Clapin 1981; Metuzals etal. 1983 b). The information obtained from these experiments has a threefold significance. Firstly, it reveals the scope and possible mechanisms of rearrangement of the NFN. Secondly, the paracrystalline structures obtained by the dialysis are very suitable in studies of the supramolecular architecture of the N F N and its components. Thirdly, the data obtained may mirror certain aspects of the pathogenic mechanisms of the neuron cytoskeleton rearrangements occurring in AD. Therefore, we compared data from preparations of the squid giant axon with those obtained from frontal lobe biopsies of Alzheimer patients. There are strik-

250

251 ing similarities observed in both types of preparations (see the accompanying paper by Metuzals et al. 1988 a). The molecular mechanisms leading to the assembly of paired helical filaments (PHFs) in human neurons are apparently highly complex, involving changes in the bloodbrain barrier, posttranslational modifications of N F N and other neuronal proteins, followed by subsequent enzymatic or spontaneous crosslinking reactions and even synthesis of new proteins. It will, therefore, be exceedingly difficult to devise experimental systems in animals or in cultured neurons that reproduce some or all the features of the Alzheimer neurofibrillary modifications. Nonetheless, efforts to develop such models, based on meager information about the nature of the PHFs themselves, are of the greatest importance if the role of altered cytoskeleton in age-related human neuronal modifications is to be understood. The present paper represents a step towards this goal. Materials and methods

All experiments and observations were made on the giant nerve fiber of squid (LoIigo pealei L.) obtained at the Marine Biological Laboratory (Woods Hole, Massachusetts) during the summers of 1981-1986. The animals used had spent no more than a few hours in laboratory tanks containing running seawater. The dorsal giant nerve fiber was dissected under running seawater on a dissection table illuminated from below through a glass window. The dissected axons were about 60 mm in length and 400-600 gm in diameter. Each end of the axon was ligated with thread prior to excision. The cleared axons were blotted to remove the external seawater, and the axoplasm was extruded from the intact living axon in the conventional manner so as to avoid contamination of the axoplasm by seawater.

Solutions Artifical seawater (ASW) for external use was composed of 4 2 3 m M NaC1, 9 m M KC1, 2 3 m M MgCI2, 2 5 m M MgSO4, 10 mM CaClz, and 5 mM Tris (pH 7.9-8.1). The ASW used for solubilizing the fixatives contained 15 mM NaCHO3 (pH adjusted to 7.2-7.4) instead of Tris buffer. The physiological buffers used in extraction experiments were designed from published data to simulate the solution conditions within the intact squid giant axon. Physiological extraction buffer contained 346 mM amino acids (including taurine and betaine), 165 mM isethionic acid, 423 mM cations, 173 mM anions, 16.7 mM carbohydrates, 1 mM ATP and 0.5 mM GTP (Morris and Lasek 1982). To eliminate free Ca / + ions and proteolysis 1 mM ethylene glycol bis (/?aminoethylether)-N,N,N',N'-tetraacetic acid (EGTA) and 0.1 mM phenylmethyl-sulfonylfluoride (PMSF) were added. A simplified R - - B solution used in most extraction ex-

periments had the following composition: 300 mM potassium methane sulfonate, 150 mM taurine, 100 mM potassium glutamate, 12.9 mM MgC12, 5 mM ATP, 3 mM CaClz, 10 mM EDTA, 10 mM 3-(4-morpholino)-propanesulfonic acid (MOPS), pH 7.2 (Rubinson and Baker 1979).

Preparation of NF-enriched eytoskeleton (Pant et al. 1986) Extruded axoplasm was quickly transferred into Eppendorf tubes on ice containing either 100 mM KC1, 10 mM MgC12, 1 mM EGTA, and 20 mM HEPES, pH 7.0 (standard phosphorylation medium) or 400 mM NaF, 1% Triton X-100, and 20 mM HEPES, pH 7.0, solution for preparation of NF-enriched cytoskeleton (PREP buffer). The extruded axoplasm was collected in PREP buffer, homogenized, and vortexed for 5 min to make a uniform suspension. This was centrifuged at 20000 rpm for 10 min with a Beckman Airfuge. The supernatant was discarded, and the remaining pellet was washed three times with 100 mM HEPES, pH 7.0, to remove Triton-100 and NaF. The final pellet was resuspended in standard phosphorylating buffer.

Crystallization (cf. Caspar et al. 1969; Metuzals etal. 1982b; Metuzals et al. 1983 a). The extruded axoplasmic rods were extracted for up to 12 h at room temperature in R - B solution. N F N rods (10 to 18) were homogenized in 300 lal of 0.5 M KCI, 0.5 M NazHPO4, pH 7.0, in a chilled glass-teflon homogenizer. The homogenate was dialyzed overnight against 1 1 of 0.5 M Tris, pH 8.0, at 4 ~ C. The dialysis was continued against 2 1 of 0.12 M (NH4) 2 SO4, 0.01 M sodium acetate, pH 5.4, for 10 h at 4 ~ C. The crystallization produced was pelleted by centrifugation at 10000 x g for 10 min and was fixed and embedded according to standard procedures of thin-sectioning EM. Other samples of the product were placed on Formvar-coated grids and stained negatively with unbuffered 1% uranyl acetate solution for 1-3 min.

Microcrystal analysis Optical diffraction was carried out with a modified Rank Pullen Diffractometer (1 mW HeNe laser of 632.8 nm wavelength). Positive transparencies of intermediate contrast were digitised on a Joyce Loebl Scandig 3 microdensitometer under software control from a PDP 11/44 computer. Data presentation was via a digital image display system (Supervisor 214, Gresham Lion PPL Ltd) and an overlay cursor controlled manually with a bit-pad (Bitpad 2, Summagraphics Ltd). Regions of clearly preserved structure (usually 512 by 256 or 512 by 512 image points)

Fig. I. A spread of NFN isolated from axoplasm, treated as described in Results. NFs (arrowhead) oriented vertically are associated by crossbridges (arrow) in a network. Bar=0.1 lam. x 60000 Fig. 2. A spread of NFN isolated from axoplasm, untreated. Extensive areas of a web-like fine filamentous matrix (asterisk); ill-defined NF-like strands of varying diameter (arrows) are continuous with matrix filaments (arrowhead). Bar=0.1 lain. x 64000 Fig. 3. NF demonstrating intercoiled protofilament substructure (H). Bar=50 nm. x 200000 Fig. 4. Circular and crescent-shaped components of NFN isolated from axoplasm, and treated as described in Results. Bar = 0.1 ~tm. x 30000

252

253 were extracted from the recorded densities for digital analysis. Image processing Two-dimensional, discrete, Fourier Transforms (approximately equivalent to a calculated diffraction pattern) were computed for each image region using a standard program: radix-2, Cooley-Tukey F F T algorithm (Saxton 1978). For the analysis of all the paracrystal structures, an intensity display proportional to the Fourier Transform amplitudes was photographed from the video screen of the image-display system as were the original image regions before processing. For the microcrystal transforms, the positions of reciprocal lattice points were determined as detailed above. Amplitudes at these points were recorded for each image region and put into the same scale with respect to the maximum (see Results). Regions of the Fourier Transform outside of a box (corresponding typically to a size of 0.02 nm-1) around each observed reciprocal lattice point were reset to zero and a two-dimensional Inverse Fourier Transform (reconstructing the equivalent image) calculated for each region. In this way the microcrystal images were spatially filtered, resulting in the removal of non-periodic features, including noise, and partially averaged. The real space geometry of the repeating unit cell could be easily determined from these reconstructions using the image display system and cursor, and a spatial average of this unit cell over the whole image was calculated directly from each reconstruction. These spatially filtered and averaged images were photographed, as before, from the video screen. Only a small region of each microcrystal reconstruction was stored after averaging and this region was magnified (by a factor of 4) to the full video screen resolution for photography, so that the unit cell could be clearly seen. The additional points generated by this magnification were obtained from bilinear interpolation (Saxton 1978) between points of the original. The reconstruction of the microcrystal projection shown in the Results, therefore, does not suffer from the sampling or "cheque-board" appearance of a small image after simple magnification by increasing the pixel size. Phosphorylation/dephosphorylation Experiments were carried out according to the published protocols (Pant et al. 1986). Dephosphorylation according to the G/P protocol NF preparations from seven axons were suspended in a solution containing TBS buffer ( 5 0 m M Tris, 100mM NaC1, pH 8), 1 mM ZnSO4 and protease inhibitors (1 m M PMSF, 60 Ixq/ml each of leupeptin, pepstatin A and Bestatin). One hundred ~tl of E. coli alkaline phosphatase (Sigma III R) was added to 300 ~tl of the above suspension.

They were incubated at 37 ~ C for 5 h. The suspension was diluted 1:10 with TBS to yield a final concentration of 100 m M NaF, 15 mM NaC1, 40 mM Tris, 5 mM EDTA, pH 7.7. It was kept on ice for 1 h and centrifuged to obtain a NF pellet. The pellet was washed and centrifuged in 100 mM HEPES, pH 7.5. The pellet was placed on Formvar-coated grids and negatively stained with 1% uranyl acetate. Treatment with Ca 2+ ions (Gallant et al. 1986) Six axons ( ~ 5 lxl each) were suspended in 600 txl of buffer P (450 mM NaF, 1 0 m M HEPES, 1 m M EGTA, 1% Triton X-100, pH 7.5). They were centrifuged at 12000 g for 2 min. The resulting pellet was washed in buffer B (150 m M KC1, 100mM HEPES, 1 0 m M MgCI2, 1 mM EGTA, pH 7.0). This procedure of centrifuging and washing was carried out twice. The pellet was incubated for 30 min in 200 ~tl of buffer B containing 10 m M Ca 2§ Samples were placed on Formvar-coated grids and stained with unbuffered 1% uranyl acetate for 1-30 min. Gel electrophores& Polyacrylamide gel electrophoresis in the presence of dodecyl sodium sulfate (SDS) was carried out in the discontinuous system of Laemmli (Laemmli 1971). Cylindrical gels were used with a 3% stacking gel and either 5% or 7.5% separating gels. Gels were stained with Coomassie brilliant blue R and densitometric scans were made at 575 nm. Samples were prepared for electrophoresis by incubation at 100~ for 5 min in sample application solution prepared at 2 X strength and giving a final concentration of 0.01 M sodium phosphate (pH 7.0), 1% SDS, 0.01 M dithiothreitol, 10% glycerol and 0.01% bromophenol blue. Molecular weight standards were obtained from Bio-Rad. Electron microscopy Standard procedures were used to prepare samples for transmission EM of thin sections. The extruded preparations of axoplasm were fixed for 1-2 h in 1% paraformaldehyde and 1% glutaraldehyde dissolved in 0.5 M sodium cacodylate buffer (pH 7.4). After washing in three changes of 0.5 M sodium cacodylate buffer for 5 min, the axoplasmic rods were postfixed for 15 min in 1% osmium tetroxide dissolved in 0.5 M sodium cacodylate buffer (pH 7.4). The axoplasmic rods were washed briefly in 0.5 M cacodylate buffer and dehydrated through a graded series of aqueous ethanol solutions to 100%, 1-5 min each step. The axons were embedded in Vestopal W (M. Jaeger, Geneva). All stages of fixation, rinsing and dehydration were carried out at room temperature. Ultrathin sections were cut on an LKB Ultratome (LKB Instruments Inc., Rockville, Maryland) with glass knives. The sections were mounted on copper grids coated with Formvar and carbon and then double stained with satu-

Fig. 5. Unwinding of NFs (U) and associations of protofilaments in circular assemblies (A). Bar = 0.1 tim. x 100000 Fig. 6. Continuous network of tubes (S) and rope-like strands (R) of NFN protein paracrystals. Bar = 0.5 ~tm. x 48000 Fig. 7. Tube of NFN protein paracrystal. Bar=0.1 Ixm. x 150000

254

255 rated uranyl acetate in ethanol followed by Reynolds' lead citrate (Reynolds 1963). The sections were examined in Elmiskop 101, Philips E M 420, and Zeiss 10 C A electron microscopes, equipped with double-condenser illumination, 30-gm thin-foil objective aperture d i a p h r a g m and a cold stage. F o r m v a r - c o a t e d grids were negatively stained with 1% unbuffered uranyl acetate for 3 min and washed with distilled, deionized water. Samples o f extracted a x o p l a s m were exposed to solutions containing calcium or magnesium as noted below (see Results). Results The N F N o f the squid giant axon, isolated by selective dissolving o f other axonal proteins, was spread on F o r m var-coated grids and treated for 30 sec with a d r o p of solution containing 0.5 M KC1, 0.05 M imidazole, p H 7.0. Extensive areas o f the grids were covered more or less evenly with the N F N (Fig. 1). N F s ~ 1 0 n m in diameter are oriented vertically in this figure. They are associated in a netw o r k assembly by periodically spaced, fine crossbridges measuring ~ 5 nm in diameter and spaced ~ 40 n m apart. Individual N F s detached from the network retain sidearms oriented 60 ~ with respect to the N F axis. Apparently, the brief washing o f the extracted axoplasmic ghosts with 0.5 M KCI solution loosened the network architecture by dissociating the N F N into its subunits. W i t h o u t such treatment it is much more difficult to obtain evenly thin spreads from the extracted axoplasm cylinder. The N F N spreads, without any KCl-treatment, reveal extensive areas o f web-like netw o r k or matrix formations consisting o f filaments less than 5 nm in diameter (Fig. 2). In such areas o f the N F N , individual N F s a p p e a r as strands o f variable diameter a n d in direct contact or even continuity with the fine matrix filaments. Individual N F s isolated by means o f column chrom a t o g r a p h y from axoplasm homogenates are illustrated in Fig. 3. Negatively stained filaments show a substructure o f four intercoiled protofilaments or unit filaments. Circular and crescent-shaped structures were observed in N F N spreads after treatment for 30 sec on the grid with a solution containing 0.5 M KC1, 5 m M MgC1/, 0.05 M imidazole, p H 7.0. These structures range from 30 to 100 nm in diame-

ter and a p p e a r to represent an integral p a r t o f a modified N F N (Fig. 4). Treatment o f the N F N spread on the grid with 0.5 KC1 solution for 60 sec resulted in an unwinding o f most o f the individual N F s into their c o m p o n e n t protofilaments, which a p p e a r to be associated in circular assemblies (Fig. 5).

Paracrystals of NFN proteins A series o f recrystallization experiments o f N F N proteins was patterned after the procedure used for the p r e p a r a t i o n o f t r o p o m y o s i n paracrystals (Caspar et al. 1969). The extracted axoplasmic ghosts consisting o f N F N proteins were homogenized and the h o m o g e n a t e was dialyzed overnight against solutions as described in Materials and methods. Analysis o f the extracted axoplasm by S D S - P A G E showed that it consisted of 80% N F N proteins, 14% tubulin, a n d a small a m o u n t o f actin. The pellet obtained following the crystallization procedure h a d a similar composition. Light microscopy o f the retentate showed a network of highly birefringent coiled strands and numerous small birefringent crystals. Electron micrographs o f thin sections o f e m b e d d e d pellets o f the crystallization p r o d u c t showed sheets o f intercoiled filaments o f different diameter. The sheets were curved into tubes and rolled up to form cylindrical scrolls. E M o f negatively stained crystallization p r o d u c t reveals a continuous network o f collapsed tubes (diameter 300 n m 1 gm), rope-like strands (diameter 10-50 nm) a n d crystalline sheets or microcrystals (Figs. 6-8). The tubes consisted o f ~ 2 - n m wide unit-filaments intercoiled to compose ,-~10-nm wide N F - l i k e strands, which were associated in a dense paracrystalline lattice. The N F - l i k e strands are oriented at n a r r o w angles against the transverse axis o f the tube producing patterns o f overlapping striations (Fig. 7). Such patterns m a y result from superposition o f helically ordered filaments o f the upper and lower wall o f the collapsed tube. There is a d a r k line coinciding with the central axis o f the tube. Numerous, randomized, individual 10-nm N F s were also observed between the tubes on the grid. The crystalline sheets or microcrystals a p p e a r in negatively stained p r e p a r a t i o n s as alternate strands o f strong stain exclusion and stain penetration (Fig. 8). In addition, the excluding b a n d is seen to vary along its length forming

Fig. 8. N F N protein microcrystals. N F (N) continuous with patches of microcrystal-like assemblies. The boxed areas (1, 2, 3) represent regions of each microcrystal that were extracted from the digitized image for subsequent digital processing. Region 2 is shown in more detail in Fig. 9. Bar = 50 nm. x 209600 Fig. 9. A digital image, representing the optical density of region 2 in Fig. 8. The appearance of this unprocessed image suggests a near-rectangular unit cell ( ~ 5 nm 2) that is aligned with the microcrystal axes A, B. The white quadrangle represents a region equivalent in size to the reconstructed image in Fig. 11. Bar = 50 nm. • 293000 Fig. 10. The two-dimensional Fourier transform of the microcrystal region of Fig. 9. The intensity of this pattern is proportional to the amplitude of the computed transform. The reciprocal lattice (a * and b *= 0.15 and 0.16 nm x, respectively, O *= 87 deg) is indicated, and this can be compared directly to the image in Fig. 9 Fig. 11. A digital image reconstructured from the reciprocal lattice peaks in the Fourier transform in Fig. 10. Stain excluding bands (A) and (B) indicated in Fig. 2 can now be clearly seen and it is apparent that the lattice produced by these features is not the true repeating unit of the microcrystal. The unit cell (a=6.8 nm, b=6.2 nm and 0 = 9 3 deg for this region) is indicated. Two types of stain exclusion profile can be distinguished for the morphological units of this reconstruction, one being more elongated and thinner (at the corners of the indicated unit cell) than the other (in the center of the cell), x 1756000 Fig. 12. Phase transition of microfilament assembly (asterisk) into tropomyosin-like lattice array (L) of microfilaments. Hirano body in thin section of frontal lobe biopsy of Alzheimer patient. Bar=0.1 gm. x 75000

256

257 a line of domains positioned with a slight stagger between neighbouring strands. Regions of lesser stain exclusion are just visible, connecting domains in adjacent strands. From this description, a near-rectangular lattice can be envisaged that connects these domains, although the individual morphological units cannot be seen clearly in the unprocessed images. The corresponding two-dimensional Fourier Transform is shown in Fig. I0, and this can be interpreted in terms of the negative stain profile of Fig. 9. The major peaks in the transform arise from (i)the parallel lines of major stain exclusion (A in Fig. 9) and (ii) the spacing of morphological units along (A) that are connected to neighbouring strands by parallel lines of lesser stain exclusion (B). As would be expected, the more substantial bands described in (A) give rise to a larger peak than those from (B). For the image region of Fig. 9 the spacing of these features is approximately 4.7 nm for A and 5.0 nm for B. It is convenient to consider these structural directions as the axis of the microcrystal. Closer examination of the diffraction pattern shows, however, that the major peaks are a subset of the peaks defining the true reciprocal lattice. The arrangement of the morphological units in the microcrystal is such that the reciprocal lattice (indicated on the transform in Fig. 10) has dimensions: a * = 0 A 5 nm -1, b * = 0 . 1 6 n m 1 and, for the microcrystal region of Fig. 2, O * = 87 deg. The corresponding real space lattice (a = 6.8 nm, b = 6.2 nm and O = 93 deg) that relates equivalent morphological features is shown on the spatially filtered and averaged reconstruction of Fig. 11. The stain-excluding domains are now seen to have two different appearances in projection. The lattice is face-centred, which means that a stain excluding domain of one type is in contact with four surrounding units of the opposite kind. These contacts run along the axes of the microcrystal, and Fig. 11 shows that these two contacts have substantially different stain profiles in projection. From diffraction patterns of regions in different microcrystals, at least two reciprocal lattices have been found and these have the same dimensions but complementary angles (O = 93 or 87 deg). The geometry of these two lattices is identical if one of them is reflected about either axis. This relationship would be expected if the two types of patterns were derived from microcrystals in two different orientations on the EM grid. If such an orientation difference is taken into account the remaining features of the Figs. 13--20. NFN proteins treated for 30 min with 10 mM

patterns (and reconstructed images derived from them) derive from the respective stain profiles, in projection, of the microcrystal in the two orientations. An interesting feature of this comparison, which is not dependent on the above interpretation of orientation, is that the contacts between stain-excluding units have different geometries in the two projected structures. It has to be emphasized that large areas of the grid were covered with structural arrays that revealed certain aspects of the microcrystal architecture but not as highly ordered as illustrated in Fig. 8. There were continuous transitional arrays among intact NFs, unravelled NFs, patches and sheets of network arrays with some resemblance to the features revealed in the microcrystals. Thus, the whole range of phase transitions from disassembled NFs to the different stages of reassembly into microcrystals and tubes can be identified. That a similar phase transition process may also occur in vivo was demonstrated by the Hirano body structure in Alzheimer brains consisting of scrolls of tropomyosin-like lattice sheets (Metuzals et al. 1983 d). The filamentous components of these sheets reveal all the stages of architecture ranging from a random mesh to a regular crystal structure (Fig. 12). In the neurofibrillary tangles of neurons of frontal lobe biopsies from Alzheimer patients, we have seen continuous transitional structures between random mesh and highly ordered paracrystalline PHF structure (see the accompanying paper by Metuzals et al. 1988a). Phosphorylation/dephosphorylation

In N F fractions prepared from squid giant axon and incubated for 30 rain either with 5 mM ATP or alkaline phosphatase, variable conformations of N F and their assemblies exist (cf. Metuzals 1986b). An important observation was made in dephosphorylation experiments using highly purified NF fractions prepared from the squid giant axon according to protocol GP (see Materials and methods). The highly dephosphorylated samples of the pure N F N preparations, negatively stained, display extensive formations of the strands and tubes, as illustrated in Figs. 6, 7. The strands and tubes could not be observed in controls in which the phosphatase was omitted. Treatment with Ca 2 + ions

Exposure of isolated N F N to 5 mM Ca z+ for 1 min under conditions that activate an endogenous Ca 2 § pro-

C a 2+

Fig. 13. Different degrees of intertwining of smooth NFs. Bar = 0.1 I~m. x 62 500 Fig. 14. Crescent- and leaf-shaped aggregates arranged in loose strands and nets. Bar =0.1 ~tm. x 57900 Fig. 15. Smooth NF disassembled into protofilaments (arrowhead); protofilament lattice (arrows). Bar---50 nm.• 204000 Fig. 16. NFs with localized swellings. Bar=0.1 ~tm. x 127500 Fig. 17. Globular disintegration of a NF. Bar= 50 nm. x 169200 Fig. 18. Continuity of needle-shaped NF protein aggregates with a NF. Bar= 50 nm. x 169200 Fig. 19. Globular disintegration of a NF (left); needle- and crescent-shaped aggregates arranged in loose strands and nets. Bar= 50 nm. x 127 500

Fig. 20. Crescent-shaped aggregates continuous with NFs. Bar= 50 nm.• 169200

258

259 tease, gave rise to network formations having 25-50 nm globular densities distributed along the longitudinal filaments, see Fig. 4b in Clapin and Metuzals 1984. After treatment of the N F N isolated from squid giant axon with 10 mM C a 2 + for 30 min, the high mol. wt. NF1 protein was broken down into lower mol. wt. components. The 200 K protein was modified into 110 K and 100 K proteins. The gels showed that the 60 K NF proteins had not been attacked by the protease (Gallant et al. 1986). The structures observed in the negatively stained preparations can be ordered in the following stages: In stage 1 individual NFs of smooth outline and approximately 10 nm or smaller in diameter are irregularly intertwined (Fig. 13). In stage 2 individual NFs appear to be frayed into needlelike protofilaments (Figs. 14, 15). In stage 3, the needle-like protofilaments seem to be reassembled into crescent and leaf-shaped aggregates (Figs. 16-20), several of which appeared to be continuous with NFs.

N F N homogenate after 48 h in R -- B buffer Extruded axoplasm was extracted in R - B buffer, changed several times, then homogenized and left in the buffer for 48 h. These preparations reveal strands varying in diameter from 10 to 30 nm and larger, when negatively stained with 1% uranyl acetate solution (Figs. 21, 22). The 10-nm wide strands display irregularly spaced constrictions and appear to be intercoiled into double helical structures reminiscent of PHFs (Schlaepfer 1978). The double helical structures consist of intercoiled protofilaments ~ 5 nm wide. Ladderlike substructures can be identified regularly in the double helical strands. The sidearm projections of the double helical strands are of varying shape and size (Figs. 21, 22). Extensive segregation of the 10-nm wide NFs into their 2 nm wide protofilaments can be observed extending over long distances of myelinated axons in Alzheimer brains (Fig. 23). In these axons, the protofilaments have either a parallel orientation or they are entangled. Such a segregated state of the N F N also displays the initial steps of reassembly into PHF-like structures by intercoiling of the protofilaments into irregular PHF-like assemblies and into the typical paracrystalline PHFs (Fig. 23). In other axons, the amount of the PHF-like strands distributed among modified N F N is increased as compared with the amount illustrated in Fig. 23. In these axons PHF-like strands ranging in diameter from 12 to 20 nm and displaying irregularly spaced constrictions are still continuous with NFs (see Metuzals et al. 1988a, b). Intercoiled protofilaments can be identified also in the PHFs of Alzheimer brains. Furthermore, these filaments display "barbed-wire" morphology and ladder-like substructure similar to the structures observed in squid preparations.

Discussion Reversible phase transitions between ordered and less ordered cytoskeletal protein assemblies represent a common process in the organization of the cytoplasmic matrix (Metuzals et al. 1988b). We assume that hysteresis occurs in such a process during the lifetime of the organism, thus, increasing irreversibly the ordered and insoluble phase of the proteins (Frazer et al. 1970). The formation of the neurofibrillary tangles has been analyzed from such a conceptual viewpoint (Metuzals et al. 1981). In the present paper, we tested experimentally the irreversible phase-transition paradigm as a primary pathogenic process using preparations obtained from the giant axon of the squid, Loligo pealei. The results were compared with observations from brain biopsies of Alzheimer patients. The great advantages to using squid axon preparations are: (1) the ease in obtaining non-denatured, relatively pure N F N through extraction of extruded axoplasm (Morris and Lasek 1982; Metuzals et al. 1983c), and (2)the possibility of correlating the data with functional observations of live preparations, ascertaining, for example, the motility and excitability. There is every reason to assume that the changes of the neuronal cytoskeleton occurring during Alzheimer disease (AD) are of such a fundamental nature that certain aspects of these changes can be induced experimentally in preparations from phylogenetically distant species, such as the squid. We suggest that irreversible formation of paracrystalline multi-protein assemblies through phase transition may be a widespread pathogenic event. The production of ordered arrangements of the filamentous protein components has been the goal of many studies involving proteins of muscle cytoskeleton. So far, attempts to produce clearly defined aggregates of NFs, or their constituent subunits, have been disappointing. The wider application of methods developed to polymerize muscle-related proteins into regular arrays is not new, but until now has produced no useful results from NFs. The production of two paracrystalline structures from the application of a modified tropomyosin protocol has enabled, for the first time, the analysis of negatively stained aggregates of N F proteins by well-developed techniques of optical diffraction and digital image processing (DeRosier and Moore 1970). All of the analyses presented here refer to projections of the paracrystalline structures onto a plane normal to the axis of the electron beam and therefore do not distinguish between features at different depths in the objects. This is important when assessing the present results, since the presence of more than one layer of structure subunits would lead to superposition effects that may make it impossible to derive the arrangement of subunits within the structure from a single tilt position of the specimen. In addition, the number of different kinds of protein subunits within

Figs. 21, 22. Double helical arrays of in vitro aged squid NFs, displaying intercoiled protofilaments (N); ladder-like substructure (arrows); periodic constriction of the two NFs (arrowheads); and disarranged matrix components (asterisk). Fig. 21 : x 252000. Fig. 22: x t 98000 Fig. 23. Formation of PHF-like strands (S) through irreversible segregation of the NFN into protofilaments and remodelling: protofilaments in parallel arrangement (arrowhead); protofilament tangles (asterisk); PHF (P). Myelinated axon. Thin section of frontal lobe biopsy of Alzheimer patient. Scale bar = 50 nm; x 88480 Fig. 24. PHFs of a myelinated axon reminiscent of the double helical NF structure in Figs. 21, 22. Note the barbed-wire morphology in the PHFs and the double helical structures of squid preparations. Protofilaments (F) and ladder-like substructure in PHFs (arrows). Thin section of frontal lobe biopsy of Alzheimer patient, x 164000

260 the paracrystal is difficult to determine unless the threedimensional stain distribution of each type of unit is understood (Aebi et al. 1982). The tropomyosin protocol, used for the generation of "tactoids" and extended " n e t " paracrystals (Caspar et al. 1969; Cohen et al. 1971), seems ideally suited to attempt the polymerization of NF proteins into paracrystals. Both of these proteins are filamentous molecules composed of coiled a-helices. Because of this structural similarity, a number of useful comparisons can be made between tropomyosin paracrystals and tubes. Nets of pure tropomyosin are composed of near rectangular or square unit cells. By varying pH around the iso-electric point a so-called heteromorphic net can arise. This net consists of parallel thick strands defining the axis with thinner, less ordered, cross connections. This is essentially the kind of substructure found within the tubes, although the latter are far less regular than tropomyosin nets. This substructure undoubtedly reflects the coiled-coil nature of the component molecules. A comparison can be made between the apparent microcrystal structure and that of actin polymerized in the presence of cations or transition metals. These paracrystals can be obtained in a variety of forms including sheets and flattened tubes (Dickens and DosRemedios 1978). It is apparent that a number of similarities exist between the structures formed by actin, associated in the presence of cations and microcrystals. However, apart from the superficial similarities in overall size and packing of units within the paracrystals, there does not appear to be a close correlation between the fine structures of microcrystals and those of actin assemblies (Aebi et al. 1980). The evidence for the involvement of the N F N proteins in the structure of tubes and microcrystals is convincing. The analysis of the tube preparations by SDS-PAGE indicates that the tubes are composed of NF proteins. The tube paracrystals have been identified also in highly purified preparations of N F N proteins. The microcrystal-like sheets have been observed in samples of the high mol. wt. NF1 protein purified by chromatography (Metuzals and Clapin 1981; Metuzals etal. 1983a). In grids covered with tubes and microcrystals, continuous transitional arrays among NFs, unravelled NFs and patches and sheets of microcrystal-like assemblies existed (Fig. 8). Presently we are entertaining the possibility that the microcrystals represent assemblies of the high mol. wt. NF-I protein, which may be the main constituent of the fine filamentous matrix of the N F N (Fig. 2). It is significant that another major contender for a NF paracrystal, the so called "wooly-bears" and "cross-linked sheaf" structures (Chou 1984) formed in chronic/r nodipropionitrile (IDPN) neuropathies, have similar open lattice structures to those of tropomyosin nets. These paracrystals have been studied by the same procedures used in the work detailed above (Adam and White, unpublished results). I D P N is believed to produce its effects by interfering with the associations between NFs and the cytoskeletal network (Papasozomenos et al. 1981 ; Sayre et al. 1985). It seems reasonable that cross connections seen in tubes are directly related to crossbridging structures that associate NFs with the main components of squid cytoskeleton. The tube structure, as a whole, is clearly not the same as that of a two-dimensional paracrystal, and some additional factors may be required to account fully for its three-

dimensional scroll-type features. A relationship exists between the open, extended nets of tropomyosin, which are disordered partly because of the large amount of associated water (Cohen et al. 1971), and the tactoids produced by lead and barium cations (Caspar et al. 1969). Caspar et al. (1969) showed that nets can collapse to form disordered fibrous aggregates which display certain similarities to the 50-nm rope-like strands also seen in the squid paracrystal preparations (Fig. 6). The tropomyosin nets may collapse completely to form tactoids (Cohen et al. 1971). It seems reasonable to assume that hydration and surface charge effects are important factors in these transitions. They may well represent the major influence on the type of paracrystal, and its associated degree of order, produced from N F N preparations. This interpretation is supported by the observation that tube paracrystals were observed in highly dephosphorylated samples of pure N F N preparations, treated according to protocol G/P (see Materials and methods) but not in controls in which the phosphatase was omitted. Removal of phosphate groups causes changes in the charged and hydration state of sidearm proteins (Eagles et al. 1986; Sayre et al. 1985) that may favour autocrosslinking and incorporation of these proteins into the NF-like strands (Metuzals 1986b). It is interesting to note in this connection, that brief treatment with 0.5 M KC1 solution, as illustrated in Figs. 1 and 5, loosens the network architecture of the N F N and unravels the N F N strands, thereby providing opportunity for new assembly patterns as indicated by the circular profiles in Fig. 5. The circular and crescentshaped formations initiated by brief treatment with 5 mM Mg 2+ may result from reassembly of disassembled N F N components (Fig. 4) (Clapin and Metuzals 1984). Thin sections of fixed and embedded crystallization products indicated that the real structure of the tubes is a scroll formed by enrolled paracrystalline sheets. It is noteworthy that the basic structure of Hirano bodies is a twisted lattice sheet (Metuzals et al. 1983 b). Thus, comparisons of NF aggregates and those produced from tropomyosin, a similar coiled-coil molecule, tend to support the idea that the tube structures and the microcrystals may be formed by a phase transition from normal NFN, via one or more intermediate aggregates. It is likely that the tubes themselves do not represent the final stable aggregation of NFs. Further modification of the tropomyosin protocol may yield more regular structures providing far more useful information on the N F N subunits and their assemblies under normal conditions and in AD. The study of altered cross-connections of the NF cytoskeleton, possibly mediated through hydration and surface charge effects, is clearly important. This work should eventually lead to elucidation of optimum conditions for the production of extensive regular NF paracrystals yielding the first high resolution studies of the NF subunit morphology. In addition to its direct effects, Ca 2 § activates a specific protease in the axoplasm of the squid giant fiber. The proteolytic activity of this protease begins with the high mol. wt. N F N proteins, initiating their cleavage. The proteolytic products reassembled into new assembly patterns, such as the amorphous, crescent and leaf-shaped aggregates (Figs. 13-20), perhaps through activation of an assembly enzyme (Schlaepfer 1983; Traub 1985; Metuzals etal. 1986). The proteolytic removal of the sidearms from the

261 N F s m a y initiate twisting o f the individual N F s a r o u n d each other (Fig. 13). O u r results s u p p o r t the conclusion that the modifications o f the 200 K N F N sidearm protein and o f the high mol. wt. NF-1 protein by p h o s p h o r y l a t i o n / d e phosphorylation, Ca2+-activated protease and other factors, initiating modification and reassembly o f N F N proteins m a y be o f importance in the pathogenetic processes of A D . The possibility does exist that in neurons o f Alzheimer patients scroll paracrystals and microcrystals c o m p o s e d o f N F N proteins do exist but have been overlooked until now. The same m a y be true concerning the crescent-shaped aggregates observed in N F N preparations treated with Ca 2 + ions. W e intend to search for such stages o f N F N protein modifications in Alzheimer neurons; the process of their f o r m a t i o n is partially k n o w n from test-tube experiments. These efforts are o f significance in elucidating the causal factors o f the pathogenic changes occurring during A D . The structure o f the double helical filaments observed in the aged N F preparations (Figs. 21, 22) indicates that the N F N c o m p o n e n t s have the tendency to intercoil and cross-associate. The m o r p h o l o g y o f the double helical filaments observed in the preparations obtained from squid is in m a n y respects similar to those observed in thin sections o f brains o f patients suffering from Alzheimer's disease (Figs. 21-24). U n d e r certain conditions the a-helical domains o f N F proteins m a y become deformed and produce fl structure. Evidence for this suggestion comes from the fact that cross-fl structure has been observed in m a m m a l i a n N F s (Wais-Steider et al. 1983). In addition parallel-fl structure can be p r o d u c e d in Myxicola N F s by deforming the a-helices under various conditions (P.A.M. Eagles, unpublished results). W i t h these observations in mind it is most interesting that X-ray diffraction from P H F s indicates cross-fl conformations (Kirschner et al. 1986). The in vitro experiments described in this p a p e r demonstrate the p o l y m o r p h i s m and the phase transitions inherent in the organization o f the N F N . The comparison o f the d a t a obtained from squid p r e p a r a t i o n s with those obtained from brains o f patients suffering from Alzheimer's disease, support the conclusion that irreversible p o l y m o r p h i s m and phase transition processes o f the N F N have a m a j o r role in the pathogenic mechanisms of A D . A n o t h e r example o f irreversible phase transition during aging and pathological modifications of the cytoskeleton is represented by the form a t i o n o f H i r a n o bodies from r a n d o m microfilament assemblies (Fig. 12). This transition o f the microfilaments into the crystalline state reminiscent o f t r o p o m y o s i n crystals seems to be related to the chemical composition o f the Hirano bodies as represented by the presence o f actin crosslinking proteins (Galloway et al. 1987). Thus, the analogous physico-chemical processes - segregation, autocrosslinking a n d phase transitions - m a y be involved in the formation o f the H i r a n o bodies as well as o f the P H F assemblies (Metuzals et al. 1983d; Metuzals et al. 1988a, see accompanying paper).

Acknowledgements. This investigation was supported by grants MA-1247 and MA-8605 from the Medical Research Council of Canada to J. Metuzals. We wish to thank P. Chow-Chong, Pamela Sawyer and Louise Friebel for technical assistance. Dr. D.F. Clapin has participated in part of this investigation. P.A.M. Eagles thanks the MRC (Great Britain), the British Foundation for Age Research and the Wellcome Trust for generous support.

References Aebi U, Smith PR, Isenburg G, Pollard DP (1980) Structure of crystalline actin sheets. Nature 288:296-298 Aebi U, Fowler WE, Smith PR (1982) Three-dimensional structure of proteins determined by electron microscopy. Ultramicroscopy 8 : 191 206 Cajal SR (ed) (1952) Histologie dy Syst6me Nerveux de l'Homme et des vert6br6s. Tome premier. Consejo Superior de Investigaciones Cientificas. Institute Ramon y Cajal, Madrid Caspar DLD, Cohen C, Longley W (1969) Tropomyosin: Crystal structure, polymorphism and molecular interactions. J Mol Biol 41 : 87-107 Chou SM (1984) Chronic IDPN poisoning of monkey as a model for ALS : A pathogenetic consideration based on perikaryal pathology. Abstracts of Neurotoxicology Conference. Neurotoxicology (Little Rock) V. 5 (2) : 303 Clapin DF, Metuzals J (1984) Polymorphism in neurofilament protein assemblies induced by magnesium (II), calcium (II), and tannate ions. Proc Ann EMSA Meet 42: 304-305 Cohen C, Caspar DLD, Parry DAD, Lucas RM (1971) Tropomyosin crystal dynamics. Cold Spring Harbor Symp Quant Biol 36:205-216 DeRosier DJ, Moore PB (1970) Reconstruction of three-dimensional images from electron micrographs of structures with helical symmetry. J Mol Biol 52:355-369 Dickens M J, DosRemedios CG (1978) Actin microcrystals and tubes formed in the presence of gadolinium ions. Nature 276:731-733 Eagles PAM, Hopkins JM, Rao R, Chin TK (1986) Neurofilament sidearms, proteolysis and phosphorylation. In: Metuzals J (ed) Electron Microscopy and Alzheimer's Disease. San Francisco Press, Inc., San Francisco, pp 42-45 Fraser H, Smith W, Gray EW (1970) Ultrastructural morphology of cytoplasmic inclusions within neurons of ageing mice. J Neurol Sci 11 : 123-127 Gallant PE, Pant HC, Pruss RM, Gainer H (1986) Calcium-activated proteolysis of neurofilament proteins in the squid giant neuron. J Neurochem 46:1573-1581 Galloway PG, Perry G, Gambetti P (1987) Hirano body filaments contain actin and actin-associated proteins. J Neuropathol Exp Neurol 46:185 199 Geisler N, Fischer S, Vandekerckhove J, Plessmann U, Weber K (1984) Hybrid character of a large neurofilament protein (NFM): intermediate filament type sequence followed by a long and acidic carboxy-terminal extension. EMBO J 3 : 2701-2706 Kirschner DA, Abraham C, Selkoe DJ (1986) X-ray diffraction from intraneuronal paired helical filaments and extraneuronal amyloid fibers in Alzheimer disease indicates cross-fl conformation. Proc Natl Acad Sci USA 83 : 503 507 Laemmli UK (1970) Cleavage of structural proteins during the assembly of the head of the bacteriophage T-4. Nature 227 : 680-685 Metuzals J (ed) (1986 a) Electron Microscopy and Alzheimer's Disease. San Francisco Press, Inc., San Francisco Metuzals J (1986b) Neurofilaments and paired helical filaments. In: Metuzals J (ed) Electron Microscopy and Alzheimer's Disease. San Francisco Press, Inc., San Francisco, pp 34-37 Metuzals J, Clapin DF (1981) Modes of crosslinking in neurofilament protein isolated from squid giant axon: electron microscopic evidence for paracrystalline arrays. Biol Bull 161:308 Metuzals J, Eagles PAM (1986) Introduction. In: Metuzals J (ed) Electron Microscopy and Alzheimer's Disease. San Francisco, Inc., San Francisco, pp 1-5 Metuzals J, Montpetit V, Clapin DF (1981) Organization of the neurofilamentous network. Cell Tissue Res 214:455-482 Metuzals J, Clapin DF, Chapman GD (1982a) Axial and radial filamentous components of the neurofilamentous network. Cell Tissue Res 223:507-518

262

Metuzals J, Clapin DF, Fennelly GJ, Eagles PAM (1982b) Paracrystalline arrays of neurofilament protein. Biol Bull 163 : 387 Metuzals J, Clapin DF, Tasaki I (1983 a) The axolemma-ectoplasm complex of squid giant axon. In: Chang DC, Tasaki I, Adelman WJ Jr, Leuchtag H R (eds) Structure and Function in Excitable Cells. Plenum Press, New York, pp 53-73 Metuzals J, Clapin DF, Eagles PAM, Fennelly GJ (1983 b) Ordered assemblies of neurofilament proteins from squid giant axon. Biol Bull 165:530 Metuzals J, Hodge A J, Lasek R J, Kaiserman-Abramof IR (1983 c) Neurofilamentous network and filamentous matrix preserved and isolated by different techniques from squid giant axon. Cell Tissue Res 228:415-432 Metuzals J, Montpetit V, Clapin DF, Nelson RF (1983d) Filamentous arrays reminiscent of tropomyosin crystals in Hirano bodies of Alzheimer's disease. Proc Ann EMSA Meet 41 : 532-533 Metuzals J, Pant H, Gainer H, Houghton S (1986) In vitro modifications of the neurofilamentous network isolated from the squid giant axon. In: Metuzals J (ed) Electron Microscopy and Alzheimer's Disease. San Francisco Press, Inc., San Francisco, pp 38-41 Metuzals J, Robitaille Y, Houghton S, Gauthier S, Kang CY, Leblanc R (1988a) Neuronal transformations in Alzheimer's disease. Cell Tissue Res 252:239-248 Metuzals J, Robitaille Y, Houghton S, Gauthier S, Leblanc R (1988 b) Transformation of the neurofilamentous network components into PHF-Iike strands and PHF paracrystals. In : Perry GP (ed) Alteration of the Neuronal Cytoskeleton in Aging and Dementia. Plenum Press, New York (in press) Morris JR, Lasek RJ (1982) Stable polymers of the axonal cytoskeleton : the axoplasmic ghost. J Cell Biol 92:192-198 Pachter JS, Liem R K H (1984) The differential appearance of neurofilament triplet polypeptides in the developing rat optic nerve. Dev Biol 103:200-210 Pant HC, Gallant PE, Gainer H (1986) Characterization of a cyclic nucleotide- and calcium-independent neurofilament protein kinase activity in axoplasm from the squid giant axon. J Biol Chem 261 : 2968-2977 Papasozomenos S, Autilio-Gambetti L, Gambetti P (1981) Reor-

ganization of axoplasmic organelles following fl,fl'-iminodipropionitrile administration. J Cell Biol 91:866-871 Reynolds E (1963) The use of lead citrate at high pH as an electronopaque stain in electron microscopy. J Cell Biol 17:208 212 Rubinson KA, Baker PF (1979) The flow properties of axoplasm in a defined chemical environment: influence of anions and calcium. Proc R Soc Lond (Biol) 205:323-345 Saxton WO (1978) Computer Techniques for Image Processing in Electron Microscopy. In: Marton L, Marton C (eds) Advances in Electronics and Electron Physics. Suppl. 10. Academic Press, New York Sayre LM, Autilio-Gambetti L, Gambetti P (1985) Pathogenesis of experimental giant neurofilamentous axonopathies: A unified hypothesis based on chemical modification of neurofilaments. Brain Res Rev 10:6%83 Schlaepfer WW (1978) Deformation of isolated neurofilaments and the pathogenesis of neurofibrillary pathology. J Neuropathol Exp Neurol 38 : 244-254 Schlaepfer WW (1983) Neurofilaments of mammalian peripheral nerve. In : Marotta CA (ed) Neurofilaments. University of Minnesota Press, Minneapolis, pp 57-85 Shaw G, Weber K (1982) Differential expression of neurofilament triplet proteins in brain development. Nature 298:277-279 Traub P (1985) In: Intermediate Filaments. Springer, Berlin Heidelberg New York Wais-Steider C, Eagles PAM, Gilbert DS, Hopkins J (1983) Structural similarities and differences amongst neurofilaments. J Mol Biol 165:393-400 Weber K, Shaw G, Osborn M, Debus E, Geisler N (1983) Neurofilaments, a subclass of intermediate flaments : Structure and expression. Cold Spring Harbor Symp Quant Biol 48. Molecular Neurobiology 717-729 Willard M (1983) Neurofilaments and axonal transport. In: Marotta CA (ed) Neurofllaments. University of Minnesota Press, Minneapolis, pp 86-116

Accepted October 28, 1987