Aggregation of Neurofilaments in NF-L Transfected Neuronal Cells: Regeneration of the Filamentous Network by a Protein Kinase C Inhibitor

Abstract: Cytoplasmic inclusion bodies that are accumulations of neurofilaments are the pathological hallmark of many neurodegenerative diseases and have been produced in transgenic mice by overexpression of mouse (NF-L and NF-M; light and medium chains, respectively) and human (NF-M and NF-H; medium and heavy chains, respectively) neurofilament subunits. This report describes a neuronal culture model in which human NF-L was overexpressed to produce cytoplasmic accumulations of neurofilaments within cell bodies concomitant with the collapse of the endogenous neurofilament network. Electron microscopy showed that, within accumulations, neurofilaments retained a filamentous structure. The culture model thus provides a novel system in which the effect on neurofilament accumulations of manipulating protein phosphorylation can be studied. Treatment of cells containing neurofilament accumulations with bisindolylmaleimide, a specific protein kinase C inhibitor, resulted in regeneration of the filamentous network; this effect was not due to a change in the level of transfected NF-L expression. These findings lend support to the suggestion that an impairment in the regulation of protein phosphorylation may lead to the accumulation of neurofilaments seen in neurodegenerative disease.     

Effects of phosphorylation of the neurofilament L protein on filamentous structures.

Effects of phosphorylation of the neurofilament L protein (NF-L) on the reassembly system were studied by both sedimentation experiments and low-angle rotary shadowing. Bovine spinal cord NF-L was phosphorylated with 3-4 mol/mol protein by either the catalytic subunit of cAMP-dependent protein kinase or protein kinase C. Phosphorylated NF-L could not assemble into filaments. Phosphorylation by either cAMP-dependent protein kinase or protein kinase C inhibited the same step of the reassembly process. Phosphorylated NF-L remained as an 8-chain complex even in favorable conditions for reassembly. The extent of the effect of phosphorylation on the filamentous structure of NF-L was also investigated by using the catalytic subunit of cAMP-dependent protein kinase. The amount of unassembled NF-L increased linearly with increased phosphorylation in the sedimentation experiments. Structural observations indicated that 1 or 2 mol of phosphorylation is enough to inhibit reassembly and to induce disassembly, and the disassembly process was also observed. The filaments were shown to unravel with disassembly. Star-like clusters, which we reported as being the initial stage of reassembly, were also identified.     

Stable interaction between G-actin and neurofilament light subunit in dopaminergic neurons

Excessive accumulation of neurofilaments in the cell bodies and proximal axons of motor neurons is a major pathological hallmark of motor neuron diseases. In this communication we provide evidence that the neurofilament light subunit (68 kDa) and G-actin are capable of forming a stable interaction. Cytochalasin B, a cytoskeleton disrupting agent that interrupts actin-based microfilaments, caused aggregation of neurofilaments in cultured mesencephalic dopaminergic neurons, suggesting a possible interaction between neurofilaments and actin; which was tested further by using crosslinking reaction and affinity chromatography techniques. In the cross-linking experiment, G-actin interacted with individual neurofilament subunits and covalently cross-linked disuccinimidyl suberate, a homobifunctional cross-linking reagent. Furthermore, G-actin was extensively cross-linked to the light neurofilament subunit with this reagent. The other two neurofilament subunits showed no cross-linking to G-actin. Moreover, neurofilament subunits were retained on a G-actin coupled affinity column and were eluted from this column by increasing salt concentration. All three neurofilament subunits became bound to the G-actin affinity column. However, a portion of the 160 and 200 kDa neurofilament subunits did not bind to the column, and the remainder of these two subunits eluted prior to the 68 kDa subunit, suggesting that the light subunit exhibited the highest affinity for G-actin. Moreover, neurofilaments demonstrated little or no binding to F-actin coupled affinity columns. The phosphorylation of neurofilament proteins with protein kinase C reduced its cross-linking to G-actin. The results of these studies are interpreted to suggest that the interaction between neurofilaments and actin, regulated by neurofilament phosphorylation, may play a role in maintaining the structure and hence the function of dopaminergic neurons in culture.     

Molecular architecture of the neurofilament. II. Reassembly process of neurofilament L protein in vitro

Reassembly of the neurofilament (NF) in vitro was studied by low-angle rotary shadowing electron microscopy. Various intermediate stages of the reassembly were reconstructed from the smallest molecular mass subunit (NF-L) under controlled reassembly conditions. NF-L in 6 M-urea took the form of spherical particles with a diameter of about 12 nm. NF-L aggregated into rodlets of 70 to 80 nm long in a low-salt solution at alkaline pH. By reducing the pH of the dialyzing solution to 6.6, a pair of rods was formed by association side-by-side. Increasing the temperature of low-salt solutions from 4 degrees C to 35 degrees C did not produce intermediate-sized filaments. The addition of Mg2+ to the dialyzing solution resulted in the formation of short intermediate-sized filaments even at 4 degrees C. Further dialysis of the short intermediate-sized filaments against reassembly solution containing both NaCl and MgCl2 at 37 degrees C failed to elongate them into longer filaments, suggesting that annealing does not contribute to the elongation of neurofilaments. Different roles for Mg+ and NaCl in neurofilament reassembly were indicated. While Mg2+ strengthened the lateral association between 70 to 80 nm rods, NaCl appeared to promote the end-to-end association of filaments preferentially. Longer filaments were formed by increasing the NaCl concentration. By dialyzing NF-L against a buffer containing 50 mM-NaCl in the absence of Mg2+, unraveled filaments were formed. The many unraveled filaments were composed of four 8 nm wide filaments, which have been called the subfilament or the protofibril. Time-course experiments of the reassembly were performed in the absence of Mg2+, in which condition the rate of neurofilament reassembly appeared to be reduced. Star-like clusters, about four protofibrils joined together at one end, were suggested to be the initial stage of the intermediate-sized filament formation. The following two-step elongation mechanism of neurofilaments was deduced from these results. The pairs of rods were added to the ends of the protofibrils of neurofilaments, and after all four protofibrils were elongated they were then packed into neurofilaments. Distribution of larger molecular mass subunits, NF-M and NF-H, was studied. Addition of NF-M or NF-H to NF-L did not change the assembly properties of neurofilaments. Unraveled filaments reconstituted from NF-L plus either NF-M or NF-H indicated that NF-M and NF-H are incorporated evenly into each protofibril.     

Physicochemical characterization of the 68 000-dalton protein of bovine neurofilaments

The 68 000-dalton protein from bovine neurofilaments was purified by a combination of chromatography on DEAE-cellulose and on hydroxylapatite in buffers containing 8 M urea. Although the separation of this protein from the other proteins of the neurofilament appeared to be hampered by a mixed association of the several components, a nearly homogeneous product was obtained for study. Sedimentation equilibrium experiments in buffers containing 8 M urea showed the molecule to be a monomer with a molecular weight of 70 600 +/- 2000. Circular dichroic spectra taken under the same conditions gave no evidence of residual alpha-helix. Molecular sieve chromatography in 8 M urea on controlled-pore glass showed that the molecule eluted at an unexpectedly small volume. The small elution volume did not depend significantly on protein concentration and is unlikely to be the result of intermolecular association. Rather, the monomer probably has a conformation more rigid or extended than a classical random coil. When dialyzed into 0.01 M tris(hydroxymethyl)aminomethane/1 mM ethylene glycol bis(beta-aminoethyl ether)-N,N,N',N'-tetraacetic acid/0.1 mM dithioerythritol, pH 8.5, the protein does not assemble into filaments. Sedimentation velocity reveals that under these conditions it consists mainly of a 4.8S molecular species, containing few large particles; sedimentation equilibrium shows that it is composed of oligomers, the smallest present in significant concentration having a molecular weight approximately that of a trimer. Circular dichroism measurements lead to the interpretation that the molecule has refolded in this buffer into a structure that has approximately 55% alpha-helix. Assembly into filamentous particles resembling neurofilaments occurs when the protein is dialyzed against 0.1 M 2-(N-morpholino)ethane-sulfonic acid/0.1% beta-mercaptoethanol/1 mM ethylene glycol bis(beta-aminoethyl ether)-N,N,N',N'-tetraacetic acid/0.17 M NaCl, pH 6.5. We suggest that the oligomeric species present in 0.01 M tris(hydroxymethyl)aminomethane may frequently be present in solubilized preparations of intermediate filaments and may represent an intermediate in the assembly process.     

Intermediate Filament Disassembly in Cultured Dorsal Root Ganglion Neurons Is Associated with Amino-Terminal Head Domain Phosphorylation of Specific Subunits

Abstract: We previously reported that activation of protein kinase A in cultured rat dorsal root ganglion neurons, treated concomitantly with low concentrations of okadaic acid that selectively inhibit protein phosphatase-2A, enhanced the Triton X-100 solubility of neurofilament triplet proteins. We now show that peripherin and α-internexin follow the same fragmentation profile as the neurofilament subunits, consistent with the notion that all five cytoplasmic intermediate filament proteins in these neurons form an integrated filamentous network whose assembly can be modulated by protein kinase A. Similar to the situation previously observed for the light neurofilament subunit, there was a strong correlation between phosphorylation of the amino-terminal head domain of peripherin and filament fragmentation. In contrast, insignificant levels of ³²P were incorporated into α-internexin under conditions promoting disassembly, indicating that phosphorylation of this protein is not involved directly in filament fragmentation. The situation for the mid-sized neurofilament subunit (NFM) was not as clear-cut. Phosphopeptide mapping of NFM revealed many head and tail domain phosphorylation sites. However, changes in NFM head domain phosphorylation under conditions promoting filament disassembly were not as pronounced as for peripherin.     

RGD-葡激酶的凝胶过滤层析法复性及其纯化

构建的溶栓和抗栓双重功能的RGD-葡激酶突变体(RGD-Sak)在大肠杆菌中高表达,目的蛋白质以包涵体形式存在。为获得有活性的蛋白质,需要对包涵体进行变复性。利用凝胶层析方法对包涵体中RGD-Sak进行复性,并与稀释复性法进行比较,发现凝胶柱复性方法具有操作周期短、简便、成本低而高效等优点。复性后蛋白质用Q-Sepharose FF离子交换进一步纯化,纯度达95%,酪蛋白凝胶板活性测定表明两种复性法得到的蛋白质比活性相当。圆二色谱测定显示两种复性法得到的蛋白质的二级结构成份和谱形一致,说明在两种复性过程中完成了RGD-Sak分子的正确折叠。     

Phosphorylation of native and reassembled neurofilaments composed of NF-L, NF-M, and NF-H by the catalytic subunit of cAMP-dependent protein kinase.

Phosphorylation of neurofilament-L protein (NF-L) by the catalytic subunit of cAMP-dependent protein kinase (A-kinase) inhibits the reassembly of NF-L and disassembles filamentous NF-L. The effects of phosphorylation by A-kinase on native neurofilaments (NF) composed of three distinct subunits: NF-L, NF-M, and NF-H, however, have not yet been described. In this paper, we examined the effects of phosphorylation of NF proteins by A-kinase on both native and reassembled filaments containing all three NF subunits. In the native NF, A-kinase phosphorylated each NF subunit with stoichiometries of 4 mol/mol for NF-L, 6 mol/mol for NF-M, and 4 mol/mol for NF-H. The extent of NF-L phosphorylation in the native NF was nearly the same as that of purified NF-L. However, phosphorylation did not cause the native NFs to disassemble into oligomers, as was the case for purified NF-L. Instead, partial fragmentation was detected in sedimentation experiments and by electron microscopic observations. This is probably not due to the presence of the three NF subunits in NF or to differences in phosphorylation sites because reassembled NF containing all three NF subunits were disassembled into oligomeric forms by phosphorylation with A-kinase and the phosphorylation by A-kinase occurred at the head domain of NF-L whether NF were native or reassembled. Disassembling intermediates of reassembled NF containing all three NF subunits were somewhat different from disassembling intermediates of NF-L. Thinning and loosening of filaments was frequently observed preceding complete disassembly. From the fact that the thinning was also observed in the native filaments phosphorylated by A-kinase, it is reasonable to propose the native NF is fragmented through a process of thinning that is stimulated by phosphorylation in the head domain of the NF subunits.     

Preferential phosphorylation of the 150,000 molecular weight component of neurofilaments by a cyclic AMP-dependent, microtubule-associated protein kinase

Highly purified preparations of bovine brain and rabbit nerve root neurofilaments were found to be lacking in protein kinase activity when either histone FIIA or the neurofilaments themselves were used as acceptors. There was no augmentation of activity in the presence of cyclic AMP. Addition of microtubule proteins prepared by cycles of assembly and disassembly resulted in phosphorylation of histone, phosphorylation of tubulin and the microtubule-associated proteins, and phosphorylation of neurofilament subunits. The phosphorylation of neurofilaments was predominantly in the 150,000-dalton species and was completely cyclic AMP dependent.     

Chemical cross-linking analyses of ox neurofilaments.

Freshly isolated intact ox neurofilaments have been incubated with copper(II)-o-phenanthroline complex to induce thiol cross-linking between the two largest (apparent Mr 205 000 and 158 000) polypeptide components. Subsequent tryptic digestion shows that the thiol bonds formed between these polypeptides are distributed exclusively among 'rod-domain' fragments that remain associated with intact sedimentable filaments. These observations suggest that the polypeptide chains of the two largest neurofilament components are closely arranged within the backbone but are separate from one another in more peripheral regions. Soluble protofilaments derived from neurofilament disassembly at low ionic strength and high pH have also been cross-linked via thiol bonds in order to determine the polypeptide arrangement within these structures. All three neurofilament polypeptides cross-link more readily when in the form of protofilaments than when in the form of fully assembled filaments, and the pattern of cross-linked complexes formed is different. Analysis of one of these complexes shows that at least some of the protofilaments are composed of oligomers containing both the 72 000- and the 158 000-Mr neurofilament polypeptides arranged in close proximity.     

Properties of highly viscous gels formed by neurofilaments in vitro. A possible consequence of a specific inter-filament cross-bridging.

Neurofilaments freshly isolated from bovine spinal cord form a reversible gel in vitro, consisting of nearly parallel and interlinked filaments organized in bundles. This phenomenon is obtained above a critical neurofilament concentration and is highly sensitive to denaturation. No gelation occurs with neurofilaments reconstituted from urea-solubilized subunits. The velocity of the gelation kinetics, optimum at a slightly acidic pH, is inhibited by low and high ionic strength and activated by millimolar concentrations of Mg2+ and other bivalent cations. No protein other than the purified neurofilament preparation itself (80-95% neurofilament triplet) is necessary for the formation of a gel. However, purified cytoskeletal proteins from microtubules and neurofilaments influence the viscosity of the native preparation. These observations suggest a reticulation in vitro between neurofilaments, dependent upon a fragile conformation of the polymers and possibly mediated through the high-Mr neurofilament subunits (200 kDa and 150 kDa). The significance of these results is discussed with regard to the inter-neurofilament cross-bridging in situ involving the 200 kDa subunit described by Hirokawa, Glicksman & Willard [(1984) J. Cell Biol. 98, 1523-1536].     

Acrylamide Retards the Slow Axonal Transport of Neurofilaments in Rat Cultured Dorsal Root Ganglia Neurons and the Corresponding Mechanisms

Chronic acrylamide (ACR) exposure induces peripheral-central axonopathy in occupational workers and laboratory animals, but the underlying mechanisms remain unclear. In this study, we first investigated the effects of ACR on slow axonal transport of neurofilaments in cultured rat dorsal root ganglia (DRG) neurons through live-cell imaging approach. Then for the underlying mechanisms exploration, the protein level of neurofilament subunits, motor proteins kinesin and dynein, and dynamitin subunit of dynactin in DRG neurons were assessed by western blotting and the concentrations of ATP was detected using ATP Assay Kit. The results showed that ACR treatment results in a dose-dependent decrease of slow axonal transport of neurofilaments. Furthermore, ACR intoxication significantly increases the protein levels of the three neurofilament subunits (NF-L, NF-M, NF-H), kinesin, dynein, and dynamitin subunit of dynactin in DRG neurons. In addition, ATP level decreased significantly in ACR-treated DRG neurons. Our findings indicate that ACR exposure retards slow axonal transport of NF-M, and suggest that the increase of neurofilament cargoes, motor proteins, dynamitin of dynactin, and the inadequate ATP supply contribute to the ACR-induced retardation of slow axonal transport.     

Study of the 10-nm-filament fraction isolated during the standard microtubule preparation.

The cold non-depolymerizable fractions obtained during the standard procedure for the isolation of microtubules from ox brain stem-cerebral hemispheres and spinal cord have been studied. The cerebral-hemisphere preparation was composed of 10-nm filaments but also contained large amounts of membranes. The polypeptide content included tubulin, microtubule-associated proteins and minor proteins corresponding to the neurofilament triplet of proteins of mol.wt. 210 000, 160 000 and 70 000 respectively. The brain-stem preparation contained more 10-nm filaments than membranes. The polypeptide content consisted of the neurofilament triplet (35%), tubulin (30%) and minor proteins. In contrast, the spinal-cord preparation was mainly composed of 10-nm filaments, free of membranes and containing essentially the neurofilament protein triplet (64%). These filaments appeared very similar to the peripheral-nervous-system neurofilaments described by several authors. Since the best neurofilament from the central nervous system often contained less than 15% of the neurofilament protein triplet, our spinal-cord preparation is an improvement on the usual neurofilament preparation. This simple and rapid method gave large amounts of 10-nm filaments (100 mg per 100 g of spinal cord) characterized by the absence of membranous material, a low content of tubulin and the 50 000-mol.wt.-protein component, and a high content of neurofilament peptides. Thus, the presence of tubulin in 10-nm filament preparations seems to be related to the contaminant membranous material and not to be linked to the interaction in vitro of tubulin or microtubules with neurofilaments, as has been suggested previously.     

Neurofilament homeostasis and motoneurone degeneration

Neurofilament disorganisation is a hallmark of various neurodegenerative diseases. We review here current knowledge of neurofilament structure, gene expression and function. Neurofilament involvement in motoneurone neurological diseases is discussed in view of recent data from transgenic and spontaneous mouse mutants. In the mammalian neurone, the three neurofilament subunits are assembled into intermediate filaments as obligate heteropolymers. The subunits are expressed differentially during development and adult life according to the cell type and its physiological state. In addition to the well-established role of neurofilaments in the control of axonal calibre, there is increasing evidence that neurofilaments can interact with other cytoskeletal components and can modulate the axoplasmic flow. Although the extent to which neurofilament abnormalities contribute to the pathogenesis in human diseases remains unknown, emerging evidence suggests that disorganised neurofilaments can provoke degeneration and death of neurones. BioEssays 23:24-33, 2001.     

Ca2(+)-calmodulin-dependent protein kinase II phosphorylates various types of non-epithelial intermediate filament proteins

We have investigated the actions of Ca2(+)-calmodulin (CaM)-dependent protein kinase II on various types of non-epithelial intermediate filament proteins, vimentin, desmin, glial fibrillary acidic protein (GFAP) and neurofilament triplet proteins. Most of these filament proteins could serve as substrates. The effects of phosphorylation on the filamentous structure of vimentin were investigated in sedimentation experiments and by using electron microscopy. The amount of unassembled vimentin increased linearly with increased phosphorylation. However, the extent of the effect of phosphorylation on the potential to polymerize was also affected by the MgCl2 concentration, under conditions for reassembly. The actions of Ca2(+)-CaM-dependent protein kinase II on non-epithelial intermediate filaments under physiological conditions are given attention.     

Molecular properties of R-phycocyanin subunits from Polysiphonia urceolata in potassium phosphate buffer.

The two subunits of R-phycocyanin from Polysiphonia urceolata were isolated and renatured. The renatured subunits were characterized by electrophoresis, molecular weights and spectra. The blue-shifted spectra, fluorescence recovery and restoring of the energy transfer suggested correct refolding of the subunits. The molecular properties of the subunits in potassium phosphate buffer (KPB) were investigated in detail. The total fluorescence yields (Q(T)) of the beta subunit declined while the energy transfer efficiency (E(T)) in the beta subunit was promoted with the increase of KPB concentration. On the other hand, both Q(T) and E(T) were enhanced with the increasing of the subunit concentrations. Based on the structural information, the fluorescence quenching in high concentrations of KPB was ascribed to less rigid chromophores caused by the weakening of the hydrogen-bond interaction network, while the enhancement of the fluorescence and E(T) was due to the aggregation of the subunits in the ionic solvent. Aggregation was confirmed by cysteine-assisted promotion of renaturation yield and stability, as well as equilibrium unfolding tests. Optimal conditions were proposed for the refolding/unfolding studies, under which the subunits were mainly monomeric. Compared to that in C-PC, the blue-shifted spectrum of PCB in R-PC is suggested to bring larger energy transfer efficiency, probably due to the necessity of the light harvesting for P. urceolata living in deep water.     

Microtubule-associated proteins bind specifically to the 70-kDa neurofilament protein

Morphological and biochemical evidence have suggested that the components of the neuronal cytoskeleton, microtubules and neurofilaments (NF), interact with each other. Microtubule-associated proteins (MAPs) are plausible candidates for mediating some of these interactions and have been shown to bind to neurofilaments, as well as induce the formation of a viscous complex between neurofilaments and microtubules. By binding 32P-labeled MAPs to neurofilament proteins, which were transferred electrophoretically to nitrocellulose, we determined that, of the three neurofilament subunits, only the core NF70 subunit bound MAPs. The binding to electrophoretically transferred NF70 was specific, saturable, and reversible. Binding parameters were estimated by binding 32P-labeled MAPs to purified NF70 immobilized on nitrocellulose. Approximately 1 mol of MAPs bound per 45 +/- 15 mol of NF70 with an approximate Kd approximately 2.0 +/- 0.9 X 10(-7) M (n = 8). Reassembled filaments in suspension were used to confirm the specific binding. Tubulin and NF70 apparently bind to different sites on MAPs.     

The disassembly and reassembly of mutants of Escherichia coli heat-labile enterotoxin: replacement of proline 93 does not abolish the reassembly-competent and reassembly-incompetent states

The carrier moiety of heat-labile enterotoxin of Escherichia coli (EtxB) is formed by the noncovalent association of identical monomeric subunits, which assemble, in vivo and in vitro, into exceptionally stable pentameric complexes. In vitro, acid disassembly followed by neutralization results in reassembly yields of between 20% and 60% depending on the identity of the salts present during the acid denaturation process. Loss of reassembly competence has been attributed to isomerization of the native cis-proline residue at position 93. To characterize this phenomenon further, two mutants of EtxB at proline 93 (P93G and P93A) were generated and purified. The proline variants reveal only minor differences in their biophysical and biochemical properties relative to wild-type protein, but major changes were observed in the kinetics of pentamer disassembly and reassembly. Additionally, a loss of assembly competence was observed following longer term acid treatment, which was even more marked than that of the wild-type protein. We present evidence that the loss of assembly competence of these mutants is best explained by a cis/trans peptidyl isomerization of the unfolded mutant subunits in acid conditions; this limited reassembly competence and the biophysical properties of the native P93 mutant pentamers imply the retention of the native cis conformation in the nonproline peptide bond between residues 92 and 93 in the mutated proteins.     

Intermediate filaments exchange subunits along their length and elongate by end-to-end annealing

Actin filaments and microtubules lengthen and shorten by addition and loss of subunits at their ends, but it is not known whether this is also true for intermediate filaments. In fact, several studies suggest that in vivo, intermediate filaments may lengthen by end-to-end annealing and that addition and loss of subunits is not confined to the filament ends. To test these hypotheses, we investigated the assembly dynamics of neurofilament and vimentin intermediate filament proteins in cultured cells using cell fusion, photobleaching, and photoactivation strategies in combination with conventional and photoactivatable fluorescent fusion proteins. We show that neurofilaments and vimentin filaments lengthen by end-to-end annealing of assembled filaments. We also show that neurofilaments and vimentin filaments incorporate subunits along their length by intercalation into the filament wall with no preferential addition of subunits to the filament ends, a process which we term intercalary subunit exchange.     

Partial purification of neurofilament subunits from bovine brains and studies on neurofilament assembly

The 200,000-dalton neurofilament subunit (P200) and the 160,000-dalton (P160) and 78,000-dalton (P78) neurofilament subunits were partially purified from bovine brain. Intact neurofilaments were prepared by high- speed and sucrose-zone centrifugation. The crude neurofilament was solubilized in 8 M urea solution containing pyridine, formic acid, and 2-mercaptoethanol. The solubilized neurofilament was purified by carboxymethyl (CM) cellulose column and hydroxylapatite column chromatography. The P200 was purified as separate from P160 and P78, but the P160 and P78 subunits were copurified on CM cellulose, hydroxylapatite, Bio-Gel A150m, and Sephadex G-150 column chromatography. Electron microscopy of these purified neurofilament subunits revealed the P200 subunit as a globular structure, and the P160 and P78 subunits as a rod-shaped structure extending up to 120 nm with a 8- to 12-nm width. In the presence of 200 mM KCl, 15 mM MgCl2, and 1 mM ATP, the purified subunits assembled into long filaments. Under the assembly condition, P160 and P78 subunits elongated up to 500 nm, but the longer filament formation required the presence of P200 subunits. The filaments formed in vitro were of two types: long straight filaments and intertwined knobby-type filaments. From these results, we have suggested that P160 and P78 form the neurofilament backbone structure and P200 facilitates the assembly of the backbone units into longer filaments.