Sequence comparisons of intermediate filament chains: evidence of a unique functional/structural role for coiled-coil segment 1A and linker L1

A comprehensive analysis of the sequences of all types of intermediate filament chains has been undertaken with a particular emphasis on those of segment 1A and linker L1. This has been done to assess whether structural characteristics can be recognized in the sequences that would be consistent with the role of each region in the recently proposed "swinging head" hypothesis. The analyses show that linker L1 is the most flexible rod domain region, that it is the most elongated structure (on a per residue basis), and that it is the most variable region as regards sequence and length. Segment 1A has one of the two most highly conserved regions of sequence in the rod domain (the other being at the end of segment 2B), with seven particular residues conserved across all chain types. It also contains one of the very few potential interchain ionic interactions that could be conserved across all chain types. However, the aggregation of chains in segment 1A is specified less precisely overall by interchain ionic interactions than are the other coiled-coil segments. The apolar residue contents in positions a and d of the heptad substructure are the highest of any coiled-coil segment in the intermediate filament family. Segment 1A also displays an amino acid composition atypical of not only coiled-coil segments 1B and 2B, but indeed of two-stranded coiled coils in general. Nonetheless, molecular modeling based on the crystal structure of the monomeric 1A fragment from human vimentin shows that coiled-coil formation is plausible. The most extensive regions of apolar/aromatic residues lie at the C-terminal end of segment 2B in the helix termination motif and in segment 1A in and close to the helix initiation motif. The predicted stability of the individual alpha-helices in segment 1A is greater than in those comprising segments 1B and 2B, though potential intrachain ionic interactions are either lacking or are minimal in number. Analysis of the 1A sequence and those regions immediately N- and C-terminal to it has shown that the capping residues are near optimal close to the previously predicted ends, thus adding to the likely stability of the alpha-helical structure. However, a second terminating sequence is predicted in 1A (about 10 residues back from the C-terminus). This allows the possibility of some unwinding of the alpha-helical structure of 1A immediately adjacent to linker L1 when the head domains no longer stabilize the coiled-coil structure. All of these data are consistent with the concept of a flexible hinge at L1 and with the ability of the two alpha-helical coiled-coil strands to separate under appropriate conditions and partly unwind at their C-terminal ends to allow the head domains a greater degree of mobility, thus facilitating function.     

Hendecad repeat in segment 2A and linker L2 of intermediate filament chains implies the possibility of a right-handed coiled-coil structure

The conformation adopted by intermediate filament chains (IF) has been described in terms of a central rod domain with four, alpha-helical, left-handed coiled-coil segments (1A, 1B, 2A, and 2B) joined by linkers (L1, L12, and L2, respectively). The rod domain is terminated at its N- and C-terminal ends by "globular" head and tail domains, respectively. This analysis, initially undertaken about 20-25 years ago, was based on the recognition of an underlying heptad substructure in the sequence of the rod domain, the presence of which can be directly associated with an alpha-helical coiled-coil structure. In this work, a hendecad sequence motif that is closely related to the heptad repeat but which is nonetheless significantly different from it has been recognized in the primary structure of segments 2A and linker L2. This motif, which is 11 residues long and structurally equivalent to a true heptad plus another heptad with an inclusive stutter, is consistent with the chains adopting a continuous right-handed coiled-coil structure with a long-period pitch length. It is therefore predicted that segment 2 as a whole may have a coiled-coil conformation with both right-handed (2A+L2) and left-handed (2B) regions. The changeover in handedness would be expected to occur at the C-terminal end of linker L2 and N-terminal end of segment 2B.     

Bovine filensin possesses primary and secondary structure similarity to intermediate filament proteins

The cDNA coding for calf filensin, a membrane-associated protein of the lens fiber cells, has been cloned and sequenced. The predicted 755- amino acid-long open reading frame shows primary and secondary structure similarity to intermediate filament (IF) proteins. Filensin can be divided into an NH2-terminal domain (head) of 38 amino acids, a middle domain (rod) of 279 amino acids, and a COOH-terminal domain (tail) of 438 amino acids. The head domain contains a di- arginine/aromatic amino acid motif which is also found in the head domains of various intermediate filament proteins and includes a potential protein kinase A phosphorylation site. By multiple alignment to all known IF protein sequences, the filensin rod, which is the shortest among IF proteins, can be subdivided into three subdomains (coils 1a, 1b, and 2). A 29 amino acid truncation in the coil 2 region accounts for the smaller size of this domain. The filensin tail contains 6 1/2 tandem repeats which match analogous motifs of mammalian neurofilament M and H proteins. We suggest that filensin is a novel IF protein which does not conform to any of the previously described classes. Purified filensin fails to form regular filaments in vitro (Merdes, A., M. Brunkener, H. Horstmann, and S. D. Georgatos. 1991. J. Cell Biol. 115:397-410), probably due to the missing segment in the coil 2 region. Participation of filensin in a filamentous network in vivo may be facilitated by an assembly partner.     

Electron paramagnetic resonance analysis of the vimentin tail domain reveals points of order in a largely disordered region and conformational adaptation upon filament assembly

Very little data have been reported that describe the structure of the tail domain of any cytoplasmic intermediate filament (IF) protein. We report here the results of studies using site directed spin labeling and electron paramagnetic resonance (SDSL‐EPR) to explore the structure and dynamics of the tail domain of human vimentin in tetramers (protofilaments) and filaments. The data demonstrate that in contrast to the vimentin head and rod domains, the tail domains are not closely apposed in protofilaments. However, upon assembly into intact IFs, several sites, including positions 445, 446, 451, and 452, the conserved “beta‐site,” become closely apposed, indicating dynamic changes in tail domain structure that accompany filament elongation. No evidence is seen for coiled‐coil structure within the region studied, in either protofilaments or assembled filaments. EPR analysis also establishes that more than half of the tail domain is very flexible in both the assembly intermediate and the intact IF. However, by positioning the spin label at distinct sites, EPR is able to identify both the rod proximal region and sites flanking the beta‐site motif as rigid locations within the tail. The rod proximal region is well assembled at the tetramer stage with only slight changes occurring during filament elongation. In contrast, at the beta site, the polypeptide backbone transitions from flexible in the assembly intermediate to much more rigid in the intact IF. These data support a model in which the distal tail domain structure undergoes significant conformational change during filament elongation and final assembly.     

Residues in the 1A rod domain segment and the linker L2 are required for stabilizing the A11 molecular alignment mode in keratin intermediate filaments

Both analyses of x-ray diffraction patterns of well oriented specimens of trichocyte keratin intermediate filaments (IF) and in vitro cross-linking experiments on several types of IF have documented that there are three modes of alignment of pairs of antiparallel molecules in all IF: A11, A22 and A12, based on which parts of the major rod domain segments are overlapped. Here we have examined which residues may be important for stabilizing the A11 mode. Using the K5/K14 system, we have made point mutations of charged residues along the chains and examined the propensities of equimolar mixtures of wild type and mutant chains to reassemble using as criteria: the formation (or not) of IF in vitro or in vivo; and stabilities of one- and two-molecule assemblies. We identified that the conserved residue Arg10 of the 1A rod domain, and the conserved residues Glu4 and Glu6 of the linker L2, were essential for stability. Additionally, conserved residues Lys31 of 1A and Asp1 of 2A and non-conserved residues Asp/Asn9 of 1A, Asp/Asn3 of 2A, and Asp7 of L2 are important for stability. Notably, these groups of residues lie close to each other when two antiparallel molecules are aligned in the A11 mode, and are located toward the ends of the overlap region. Although other sets of residues might theoretically also contribute, we conclude that these residues in particular engage in favorable intermolecular ionic and/or H-bonding interactions and thereby may play a role in stabilizing the A11 mode of alignment in keratin IF.     

Self-assembly incompetence of synemin is related to the property of its head and rod domains

The mechanisms regulating the intermediate filament (IF) protein assembly are complex and not yet fully understood. All vertebrate cytoplasmic IF proteins have a central alpha-helical rod domain flanked by variable head and tail domains. The IF protein synemin cannot homopolymerize to form filament networks; it needs an appropriate copolymerization partner. To elucidate the roles of the vimentin head domain, the TAAL motif in the 2A region, and the TYRKLLEGEE motif in the 2B region of the rod domain in synemin filament formation, we have prepared a series of synemin constructs by site-directed mutagenesis and chimeric synemins having the vimentin head domain. The assembly properties of synemin constructs were assessed by the immunofluorescence of transient transfection into cultured SW13 cells without endogenous IFs. Our data showed that the formation of a filamentous network required at least the vimentin-like head domain and both the 2A and 2B regions of the rod domain.     

In vitro assembly properties of mutant and chimeric intermediate filament proteins: insight into the function of sequences in the rod and end domains of IF

The factors and mechanisms regulating assembly of intermediate filament (IF) proteins to produce filaments with their characteristic 10 nm diameter are not fully understood. All IF proteins contain a central rod domain flanked by variable head and tail domains. To elucidate the role that different domains of IF proteins play in filament assembly, we used negative staining and electron microscopy (EM) to study the in vitro assembly properties of purified bacterially expressed IF proteins, in which specific domains of the proteins were either mutated or swapped between a cytoplasmic (mouse neurofilament-light (NF-L) subunit) and nuclear intermediate filament protein (human lamin A). Our results indicate that filament formation is profoundly influenced by the composition of the assembly buffer. Wild type (wt) mouse NF-L formed 10 nm filaments in assembly buffer containing 175 mM NaCl, whereas a mutant deleted of 18 NH2-terminal amino acids failed to assemble under similar conditions. Instead, the mutant assembled efficiently in buffers containing CaCl2 > or = 6 mM forming filaments that were 10 times longer than those formed by wt NF-L, although their diameter was significantly smaller (6-7 nm). These results suggest that the 18 NH2-terminal sequence of NF-L might serve two functions, to inhibit filament elongation and to promote lateral association of NF-L subunits. We also demonstrate that lengthening of the NF-L rod domain, by inserting a 42 aa sequence unique to nuclear IF proteins, does not compromise filament assembly in any noticeable way. Our results suggests that the known inability of nuclear lamin proteins to assemble into 10 nm filaments in vitro cannot derive solely from their longer rod domain. Finally, we demonstrate that the head domain of lamin A can substitute for that of NF-L in filament assembly, whereas substitution of both the head and tail domains of lamins for those of NF-L compromises assembly. Therefore, the effect of lamin A "tail" domain alone, or the synergistic effect of lamin "head" and the "tail" domains together, interferes with assembly into 10-nm filaments.     

Properties of the desmin tail domain: studies using synthetic peptides and antipeptide antibodies

Intermediate filament (IF) proteins have a common structural motif consisting of an alpha-helical rod domain flanked by non-alpha-helical amino-terminal head and carboxy-terminal tail domains. Coiled-coil interaction between neighboring rod domains is though to generate the backbone of the 10-nm filament. There must also be other interactions between subunits to bring them into alignment and to effect elongation of the filament, but these are poorly understood. To examine the involvement of the tail domain in filament structure and stabilization, we have studied the interaction between a synthetic peptide corresponding to residues 442-450 of avian desmin, and authentic desmin protein. The potential importance of this region lies in its hydrophilic nature and its high degree of homology among the Type III IF proteins and cytokeratins 8 and 18. The peptide, D442-450, binds to a 27-residue region between lys-436 and leu-463, the carboxy terminus. The presence of the peptide during assembly causes the filaments to appear much more loosely packed than normal desmin IF. We have also generated polyclonal antibodies against this peptide and attempted to localize this portion of the tailpiece along desmin IFs by immunological procedures. By immunoblotting, we found that anti-D442- 450 antibodies recognize desmin and only those proteolytic fragments that contain the tailpiece. In contrast, the antibodies do not label any structure in adult gizzard smooth muscle and skeletal muscle myofibrils in immunofluorescence experiments during which conventional antidesmin antibodies do. At the ultrastructural level, anti-D442-450 antibodies label free desmin tetramers but not desmin IFs. These results show that, as part of an assembled IF, the epitope of anti-D442- 450 is inaccessible to the antibodies, and suggest that either the tailpiece of an IF protein may not be entirely peripheral to the filament backbone, or the interaction between end domains during assembly masks this particular region of the IF molecule.     

Coiled-coil trigger motifs in the 1B and 2B rod domain segments are required for the stability of keratin intermediate filaments

Many alpha-helical proteins that form two-chain coiled coils possess a 13-residue trigger motif that seems to be required for the stability of the coiled coil. However, as currently defined, the motif is absent from intermediate filament (IF) protein chains, which nevertheless form segmented two-chain coiled coils. In the present work, we have searched for and identified two regions in IF chains that are essential for the stability necessary for the formation of coiled-coil molecules and thus may function as trigger motifs. We made a series of point substitutions with the keratin 5/keratin 14 IF system. Combinations of the wild-type and mutant chains were assembled in vitro and in vivo, and the stabilities of two-chain (one-molecule) and two-molecule assemblies were examined with use of a urea disassembly assay. Our new data document that there is a region located between residues 100 and 113 of the 2B rod domain segment that is absolutely required for molecular stability and IF assembly. This potential trigger motif differs slightly from the consensus in having an Asp residue at position 4 (instead of a Glu) and a Thr residue at position 9 (instead of a charged residue), but there is an absolute requirement for a Glu residue at position 6. Because these 13 residues are highly conserved, it seems possible that this motif functions in all IF chains. Likewise, by testing keratin IF with substitutions in both chains, we identified a second potential trigger motif between residues 79 and 91 of the 1B rod domain segment, which may also be conserved in all IF chains. However, we were unable to find a trigger motif in the 1A rod domain segment. In addition, many other point substitutions had little detectable effect on IF assembly, except for the conserved Lys-23 residue of the 2B rod domain segment. Cross-linking and modeling studies revealed that Lys-23 may lie very close to Glu-106 when two molecules are aligned in the A(22) mode. Thus, the Glu-106 residue may have a dual role in IF structure: it may participate in trigger formation to afford special stability to the two-chain coiled-coil molecule, and it may participate in stabilization of the two-molecule hierarchical stage of IF structure.     

The last twenty residues in the head domain of mouse lamin A contain important structural elements for formation of head-to-tail polymers in vitro

Nuclear lamins are a type of intermediate filament (IF) proteins. They have a characteristic tripartite domain structure with a alpha-helical rod domain flanked by non-alpha-helical N-terminal head and C-terminal tail domains. While the head domain has been shown to be important for the formation of head-to-tail polymers that are critical assembly intermediates for lamin IFs, essential structural elements in this domain have remained obscure. As a first step to remedy this, a series of mouse lamin A mutants in which the head domain (30 amino acid residues) was deleted stepwise from the N-terminus at intervals of 10 residues were bacterially expressed. The assembly properties in vitro of the purified recombinant proteins were explored by electron microscopy. We observed that while a lamin A mutant lacking N-terminal 10 residues formed head-to-tail polymers, a mutant lacking N-terminal 20 residues or the whole head domain (30 residues) showed significantly decreased potency to form head-to-tail polymers. These results suggest that the last 20 residues (from Arg-11 to Gln-30) of the head domain of mouse lamin A contain essential structures for the formation of head-to-tail polymers. The last 20 residues of the head domain include several conserved residues between A- and B-type lamins and also the phosphorylation site for cdc2 kinase, which affects lamin IF organization in vivo and in vitro. Our results provide clues to the molecular mechanism by which the head domain plays a crucial role in lamin polymerization.     

Desmin-binding specificities of two desmin CNBr fragments that correspond to the headpiece domain and Helix 1B

D88 and D109, two cyanogen bromide fragments of desmin which essentially correspond to the amino terminal headpiece domain and Helix 1B, respectively, bind to intact desmin with different topological specificities. D88, the headpiece domain fragment, binds only to the headpiece of intact desmin. In contrast, D109, which encompasses Helix 1B and most of the linker L10 binds to desmin even when its headpiece is removed. Additionally, these fragments only bind desmin if they are present during filament assembly; they do not bind pre-assembled desmin IF or tetramers. These observations suggest that, while alpha-helical coiled-coil interaction between rod domains provides the major driving force behind IF protein dimer formation, homophilic binding of head domains of these proteins may provide an additional stabilizing force and/or specify axial registration in certain IF proteins.     

Plectin interacts with the rod domain of type III intermediate filament proteins desmin and vimentin

Plectin is a versatile cytolinker protein critically involved in the organization of the cytoskeletal filamentous system. The muscle-specific intermediate filament (IF) protein desmin, which progressively replaces vimentin during differentiation of myoblasts, is one of the important binding partners of plectin in mature muscle. Defects of either plectin or desmin cause muscular dystrophies. By cell transfection studies, yeast two-hybrid, overlay and pull-down assays for binding analysis, we have characterized the functionally important sequences for the interaction of plectin with desmin and vimentin. The association of plectin with both desmin and vimentin predominantly depended on its fifth plakin repeat domain and downstream linker region. Conversely, the interaction of desmin and vimentin with plectin required sequences contained within the segments 1A-2A of their central coiled-coil rod domain. This study furthers our knowledge of the interaction between plectin and IF proteins important for maintenance of cytoarchitecture in skeletal muscle. Moreover, binding of plectin to the conserved rod domain of IF proteins could well explain its broad interaction with most types of IFs.     

Assembly properties of two CNBr fragments of avian desmin that correspond to the headpiece domain and helix 1B

To study how different domains of the muscle-specific intermediate filament protein, desmin, contribute to its polymerization, two of its CNBr fragments were examined as to their oligomeric structure under assembly conditions. One of these, D88, covers residues 1-88 and represents almost the entire headpiece; the other, D109, covers residues 145-254, and includes the entire Helix 1B and part of linker L12 of the intact molecule. Chemical cross-linking followed by SDS-PAGE, and analytical gel filtration, revealed that in 10 mM Tris-HCl, pH 8.5, conditions that favor tetramerization of intact desmin D88 formed only dimers. D109, on the other hand, formed primarily a dimeric species but low levels of trimeric and tetrameric species were also detectable. These data are consistent with the proposal that, during assembly of intact protein molecules into IF, the headpiece and Helix 1 contribute to dimerization of two polypeptides into a parallel, in-register coiled-coil. However, additional interactions, including headpiece-to-rod binding and hydrophobic interaction along the entire rod domain, are required to stabilize the tetramers and full-size IF.     

Intermediate filament structure: 2. Molecular interactions in the filament

Molecules of intermediate filament (IF) proteins contain a central rod domain in which the two constituent chains have a predominantly α-helical conformation and are coiled around one another to form segments of two-strand rope. Possible interactions between the two long segments, termed 1B and 2 were investigated by a technique successfully employed in studies of the modes of association of collagen molecules by Miller and coworkers. Prominent maxima were found in all of the six possible modes of association between the rod domain segments in individual IF proteins and certain maxima were found to be common to all IF. The surface lattice of the IF from α-keratin has been determined and possible bonding arrangements between the rod-domain segments are catalogued. A systematic search was carried out for combinations of interaction maxima which were consistent with the dimensions of the surface lattice. By the further application of stereochemical constraints, models for the topological arrangement of the rod-domain segments on the surface lattice were derived and these are illustrated and discussed.     

The alpha-helical rod domain of human lamins A and C contains a chromatin binding site.

We examined regions of human lamins A and C involved in binding to surfaces of mitotic chromosomes. An Escherichia coli expression system was used to produce full-length lamin A and lamin C, and truncated lamins retaining the central alpha-helical rod domain (residues 34-388) but lacking various amounts of the amino-terminal 'head' and carboxy-terminal 'tail' domains. We found that lamin A, lamin C and lamin fragments lacking the head domain and tail sequences distal to residue 431 efficiently assembled into paracrystals and strongly associated with mitotic chromosomes. Furthermore, the lamin rod domain also associated with chromosomes, although efficient chromosome coating required the pH 5-6 conditions needed to assemble the rod into higher order structures. Biochemical assays showed that chromosomes substantially reduced the critical concentration for assembly of lamin polypeptides into pelletable structures. Association of the lamin rod with chromosomes was abolished by pretrypsinization of chromosomes, and was not seen for vimentin (which possesses a similar rod domain). These data demonstrate that the alpha-helical rod of lamins A and C contains a specific chromosome binding site. Hence, the central rod domain of intermediate filament proteins can be involved in interactions with other cellular structures as well as in filament assembly.     

Recent insight into intermediate filament structure

Intermediate filaments (IFs) are key players in multiple cellular processes throughout human tissues. Their biochemical and structural properties are important for understanding filament assembly mechanisms, for interactions between IFs and binding partners, and for developing pharmacological agents that target IFs. IF proteins share a conserved coiled-coil central-rod domain flanked by variable N-terminal ‘head’ and C-terminal ‘tail’ domains. There have been several recent advances in our understanding of IF structure from the study of keratins, glial fibrillary acidic protein, and lamin. These include discoveries of (i) a knob–pocket tetramer assembly mechanism in coil 1B; (ii) a lamin-specific coil 1B insert providing a one-half superhelix turn; (iii) helical, yet flexible, linkers within the rod domain; and (iv) the identification of coil 2B residues required for mature filament assembly. Furthermore, the head and tail domains of some IFs contain low-complexity aromatic-rich kinked segments, and structures of IFs with binding partners show electrostatic surfaces are a major contributor to complex formation. These new data advance the connection between IF structure, pathologic mutations, and clinical diseases in humans.     

Keratin intermediate filament chains in the European common wall lizard (Podarcis muralis) and a potential keratin filament crosslinker

On the basis of sequence homology with mammalian α-keratins, and on the criteria that the coiled-coil segments and central linker in the rod domain of these molecules must have conserved lengths if they are to assemble into viable intermediate filaments, a total of 28 Type I and Type II keratin intermediate filament chains (KIF) have been identified from the genome of the European common wall lizard (Podarcis muralis). Using the same criteria this number may be compared to 33 found here in the green anole lizard (Anole carolinensis) and 25 in the tuatara (Sphenodon punctatus). The Type I and Type II KIF genes in the wall lizard fall in clusters on chromosomes 13 and 2 respectively. Although some differences occur in the terminal domains in the KIF chains of the two lizards and tuatara, the similarities between key indicator residues – cysteine, glycine and proline – are significant. The terminal domains of the KIF chains in the wall lizard also contain sequence repeats commonly based on glycine and large apolar residues and would permit the fine tuning of physical properties when incorporated within the intermediate filaments. The H1 domain in the Type II chain is conserved across the lizards, tuatara and mammals, and has been related to its role in assembly at the 2–4 molecule level. A KIF-like chain (K80) with an extensive tail domain comprised of multiple tandem repeats has been identified as having a potential filament-crosslinking role.     

Characterization of structural changes in vimentin bearing an epidermolysis bullosa simplex-like mutation using site-directed spin labeling and electron paramagnetic resonance

Mutations in intermediate filament protein genes are responsible for a number of inherited genetic diseases including skin blistering diseases, corneal opacities, and neurological degenerations. Mutation of the arginine (Arg) residue of the highly conserved LNDR motif has been shown to be causative in inherited disorders in at least four different intermediate filament (IF) proteins found in skin, cornea, and the central nervous system. Thus this residue appears to be broadly important to IF assembly and/or function. While the genetic basis for these diseases has been clearly defined, the inability to determine crystal structure for IFs has precluded a determination of how these mutations affect assembly/structure/function of IFs. To investigate the impact of mutation at this site in IFs, we have mutated the LNDR to LNDS in vimentin, a Type III intermediate filament protein, and have examined the impact of this change on assembly using electron paramagnetic resonance. Compared with wild type vimentin, the mutant shows normal formation of the coiled coil dimer, with a slight reduction in the stability of the dimer in rod domain 1. Probing the dimer-dimer interactions shows the formation of normal dimer centered on residue 191 but a failure of dimerization at residue 348 in rod domain 2. These data point toward a specific stage of assembly at which a common disease-causing mutation in IF proteins interrupts assembly.     

Intermediate filament structure: 1. Analysis of IF protein sequence data

Amino acid sequence data for intermediate filament proteins have been analysed with a view to identifying structurally invariant segments and determining their likely secondary structure. The sequences in these segments have also been analysed for periodic distributions of particular types of residue. The results support the classification of intermediate filament proteins into three main groups and also reinforce the concept of a molecular structure with a central domain of coiled-coil segments, together with essentially non-helical N-terminal and C-terminal domains of variable size and composition. Regions exhibiting the greatest homology between the three types of IF chain are identified and significant variation in charged residue disposition along the length of individual chains is noted. The conservation in all IF protein chains of specific sites of coiled-coil rope interruption are discussed in terms of the probable molecular structure. Stabilizing ionic interactions between coiled-coil chain segments have been investigated quantitatively as a function of the relative chain stagger. In all cases and calculations favour ropes in which the constituent chains are in-register and parallel rather than antiparallel.     

The rod domain of NF-L determines neurofilament architecture, whereas the end domains specify filament assembly and network formation

Neurofilaments, assembled from NF-L, NF-M, and NF-H subunits, are the most abundant structural elements in myelinated axons. Although all three subunits contain a central, alpha-helical rod domain thought to mediate filament assembly, only NF-L self-assembles into 10-nm filaments in vitro. To explore the roles of the central rod, the NH2- terminal head and the COOH-terminal tail domain in filament assembly, full-length, headless, tailless, and rod only fragments of mouse NF-L were expressed in bacteria, purified, and their structure and assembly properties examined by conventional and scanning transmission electron microscopy (TEM and STEM). These experiments revealed that in vitro assembly of NF-L into bona fide 10-nm filaments requires both end domains: whereas the NH2-terminal head domain promotes lateral association of protofilaments into protofibrils and ultimately 10-nm filaments, the COOH-terminal tail domain controls lateral assembly of protofilaments so that it terminates at the 10-nm filament level. Hence, the two end domains of NF-L have antagonistic effects on the lateral association of protofilaments into higher-order structures, with the effect of the COOH-terminal tail domain being dominant over that of the NH2-terminal head domain. Consideration of the 21-nm axial beading commonly observed with 10-nm filaments, the approximate 21-nm axial periodicity measured on paracrystals, and recent cross-linking data combine to support a molecular model for intermediate filament architecture in which the 44-46-nm long dimer rods overlap by 1-3-nm head-to-tail, whereas laterally they align antiparallel both unstaggered and approximately half-staggered.