Intermediate filament assembly and stability in vitro: effect and implications of the removal of head and tail domains of vimentin by the human immunodeficiency virus type 1 protease

HIV Infection: Shaping the Complex, Dynamic, and Interconnected Network of the Cytoskeleton

HIV-1 has evolved a plethora of strategies to overcome the cytoskeletal barrier (i.e., actin and intermediate filaments (AFs and IFs) and microtubules (MTs)) to achieve the viral cycle. HIV-1 modifies cytoskeletal organization and dynamics by acting on associated adaptors and molecular motors to productively fuse, enter, and infect cells and then traffic to the cell surface, where virions assemble and are released to spread infection. The HIV-1 envelope (Env) initiates the cycle by binding to and signaling through its main cell surface receptors (CD4/CCR5/CXCR4) to shape the cytoskeleton for fusion pore formation, which permits viral core entry. Then, the HIV-1 capsid is transported to the nucleus associated with cytoskeleton tracks under the control of specific adaptors/molecular motors, as well as HIV-1 accessory proteins. Furthermore, HIV-1 drives the late stages of the viral cycle by regulating cytoskeleton dynamics to assure viral Pr55Gag expression and transport to the cell surface, where it assembles and buds to mature infectious virions. In this review, we therefore analyze how HIV-1 generates a cell-permissive state to infection by regulating the cytoskeleton and associated factors. Likewise, we discuss the relevance of this knowledge to understand HIV-1 infection and pathogenesis in patients and to develop therapeutic strategies to battle HIV-1.

Impact of N-Terminal Tags on De Novo Vimentin Intermediate Filament Assembly

Vimentin, a type III intermediate filament protein, is found in most cells along with microfilaments and microtubules. It has been shown that the head domain folds back to associate with the rod domain and this association is essential for filament assembly. The N-terminally tagged vimentin has been widely used to label the cytoskeleton in live cell imaging. Although there is previous evidence that EGFP tagged vimentin fails to form filaments but is able to integrate into a pre-existing network, no study has systematically investigated or established a molecular basis for this observation. To determine whether a tag would affect de novo filament assembly, we used vimentin fused at the N-terminus with two different sized tags, AcGFP (239 residues, 27 kDa) and 3 × FLAG (22 residues; 2.4 kDa) to assemble into filaments in two vimentin-deficient epithelial cells, MCF-7 and A431. We showed that regardless of tag size, N-terminally tagged vimentin aggregated into globules with a significant proportion co-aligning with β-catenin at cell–cell junctions. However, the tagged vimentin aggregates could form filaments upon adding untagged vimentin at a ratio of 1:1 or when introduced into cells containing pre-existing filaments. The resultant filament network containing a mixture of tagged and untagged vimentin was less stable compared to that formed by only untagged vimentin. The data suggest that placing a tag at the N-terminus may create steric hinderance in case of a large tag (AcGFP) or electrostatic repulsion in case of highly charged tag (3 × FLAG) perhaps inducing a conformational change, which deleteriously affects the association between head and rod domains. Taken together our results shows that a free N-terminus is essential for filament assembly as N-terminally tagged vimentin is not only incapable of forming filaments, but it also destabilises when integrated into a pre-existing network.

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.

Interaction in vitro of type III intermediate filament proteins with higher order structures of single-stranded DNA, particularly with G-quadruplex DNA

Cytoplasmic intermediate filament (cIF) proteins interact strongly with single-stranded (ss) DNAs and RNAs, particularly with G-rich sequences. To test the hypothesis that this interaction depends on special nucleotide sequences and, possibly, higher order structures of ssDNA, a random mixture of mouse genomic ssDNA fragments generated by a novel "whole ssDNA genome PCR" technique via RNA intermediates was subjected to three rounds of affinity binding to in vitro reconstituted vimentin IFs at physiological ionic strength with intermediate PCR amplification of the bound ssDNA segments. Nucleotide sequence and computer folding analysis of the vimentin-selected fragments revealed an enrichment in microsatellites, predominantly of the (GT)n type, telomere DNA, and C/T-rich sequences, most of which, however, were incapable of folding into stable stem-loop structures. Because G-rich sequences were underrepresented in the vimentin-bound fraction, it had to be assumed that such sequences require intramolecular folding or lateral assembly into multistrand structures to be able to stably interact with vimentin, but that this requirement was inadequately fulfilled under the conditions of the selection experiment. For that reason, the few vimentin-selected G-rich ssDNA fragments and a number of telomere models were analyzed for their capacity to form inter- and intramolecular Gquadruplexes (G4 DNAs) under optimized conditions and to interact as such with vimentin and its type III relatives, glial fibrillary acidic protein, and desmin. Band shift assays indeed demonstrated differential binding of the cIF proteins to parallel four-stranded G4 DNAs and, with lower affinity, to bimolecular G'2 and unimolecular G'4 DNA configurations, whereby the transition regions from four- to single-strandedness played an additional role in the binding reaction. In this respect, the binding activity of cIF proteins was comparable with that toward other noncanonical DNA structures, like ds/ss DNA forks, triplex DNA, four-way junction DNA and Z-DNA, which also involve configurational transitions in their interaction with the filament proteins. Association of the cIF proteins with the corresponding nonfolded G-rich ssDNAs was negligible. Considering the almost universal involvement of ssDNA regions and G-quadruplexes in nuclear processes, including DNA transcription and recombination as well as telomere maintenance and dynamics, it is plausible to presume that cIF proteins as complementary constituents of the nuclear matrix participate in the cell- and tissue-specific regulation of these processes.

Interaction in vitro of type III intermediate filament proteins with Z-DNA and B-Z-DNA junctions

The selection of DNA fragments containing simple d(GT)(n) and composite d(GT)(m). d(GA)(n) microsatellites during affinity binding of mouse genomic DNA to type III cytoplasmic intermediate filaments (cIFs) in vitro, and the detection of such repeats, often as parts of nuclear matrix attachment region (MAR)-like DNA, in SDS-stable DNA-vimentin crosslinkage products isolated from intact fibroblasts, prompted a detailed study of the interaction of type III cIF proteins with left-handed Z-DNA formed from d(GT)(17) and d(CG)(17) repeats under the topological tension of negatively supercoiled plasmids. Although d(GT)(n) tracts possess a distinctly lower Z-DNA-forming potential than d(CG)(n) tracts, the filament proteins produced a stronger electrophoretic mobility shift with a plasmid carrying a d(GT)(17) insert than with plasmids containing different d(CG)(n) inserts, consistent with the facts that the B-Z transition of d(GT)(n) repeats requires a higher negative superhelical density than that of d(CG)(n) repeats and the affinity of cIF proteins for plasmid DNA increases with its superhelical tension. That both types of dinucleotide repeat had indeed undergone B-Z transition was confirmed by S1 nuclease and chemical footprinting analysis of the plasmids, which also demonstrated efficient protection by cIF proteins from nucleolytic and chemical attack of the Z-DNA helices as such, as well as of the flanking B-Z junctions. The analysis also revealed sensibilization of nucleotides in the center of one of the two strands of a perfect d(CG)(17) insert toward S1 nuclease, indicating cIF protein-induced bending of the repeat. In all these assays, vimentin and glial fibrillary acidic protein (GFAP) showed comparable activities, versus desmin, which was almost inactive. In addition, vimentin and GFAP exhibited much higher affinities for the Z-DNA conformation of brominated, linear d(CG)(25) repeats than for the B-DNA configuration of the unmodified oligonucleotides. While double-stranded DNA was incapable of chasing the Z-DNA from its protein complexes, and Holliday junction and single-stranded (ss)DNA were distinguished by reasonable competitiveness, phosphatidylinositol (PI) and, particularly, phosphatidylinositol 4,5-diphosphate (PIP(2)) turned out to be extremely potent competitors. Because PIP(2) is an important member of the nuclear PI signal transduction cascade, it might exert a regulatory influence on the binding of cIF proteins to Z- and other DNA conformations. From this interaction of cIF proteins with Z- and bent DNA and their previously detected affinities for MAR-like, ss, triple helical, and four-way junction DNA, it may be concluded that the filament proteins play a general role in such nuclear matrix-associated processes as DNA replication, recombination, repair, and transcription.

Cancer-associated cleavage of cytokeratin 8/18 heterotypic complexes exposes a neoepitope in human adenocarcinomas

The intermediate filament network in simple glandular epithelial cells predominantly consists of heterotypic complexes of cytokeratin 8 (K8) and cytokeratin 18 (K18). In contrast to other cytokeratins, K8 and K18 are persistently expressed during malignant transformation, but changes in cell morphology are accompanied by alterations in the intermediate filament network. To study molecular changes, K8 and K18 were purified from surgically removed colon cancer and normal epithelia tissues. Western blotting and amino acid sequencing revealed the presence of abundant K8 and K18 fragments, truncated at the N terminus, from cancerous, but not normal, epithelial cells. The fragmentation pattern indicates proteolysis mediated by several enzymes, including trypsin-like enzymes. The cancer-associated forms of K8 and K18 are specifically recognized by the human antibody, COU-1, cloned from the B cells of a cancer patient. We demonstrate that COU-1 recognizes a unique conformational epitope presented only by a complex between K8 and K18. The epitope is revealed after proteolytic removal of the head domain of either K8 or K18. A large panel of recombinant K8 and K18 fragments, deleted N- or C-terminally, allowed for the localization of the COU-1 epitope to the N-terminal part of the rod domains. Using surface plasmon resonance, the affinity of COU-1 for this epitope was determined to be 10(9) x m(-1), i.e. more than 2 orders of magnitude higher than for intact heterotypic K8/K18 complexes. The cellular distribution of truncated K8/K18 heterotypic complexes in viable adenocarcinomas cells was probed using COU-1 showing small fibrillar structures distinct from those of intact K8/K18 complexes. Previously we demonstrated the binding and subsequent internalization of recombinant Fab COU-1 to live cancer cells. We have thus characterized a cancer neoepitope recognized by the humoral immune system. The results have biological as well as clinical implications.

Sites of nucleic acid binding in type I-IV intermediate filament subunit proteins

A combination of enzymatic and chemical ladder sequencing of photo-cross-linked protein-single-stranded oligodeoxyribonucleotide complexes and analysis by MALDI-TOF mass spectrometry was employed to identify the amino acid residues responsible for the stable binding of nucleic acids in several intermediate filament (IF) subunit proteins. The IF proteins studied included the type I and type II cytokeratins K8, K18, and K19; the type III proteins desmin, glial fibrillary acidic protein (GFAP), peripherin, and vimentin; and the type IV neurofilament triplet protein L (NF-L). The site of nucleic acid binding was localized to the non-alpha-helical, amino-terminal head domain of all of the IF proteins tested. GFAP, which has the shortest head domain of the proteins tested, cross-linked via only two amino acid residues. One of these residues was located within a conserved nonapeptide domain that has been shown to be required for filament formation. One or more cross-linked residues were found in a similar location in the other proteins studied. The major binding site for nucleic acids for most of the proteins appears to be localized within the middle of the head domain. The two exceptions to this generalization are GFAP, which lacks these residues, and NF-L, in which a large number of cross-linked residues were found scattered throughout the first half of the head domain. Control experiments were also done with two bacteriophage ssDNA-binding proteins, as well as actin and tubulin. The single sites of cross-linkage observed with the bacteriophage proteins, Phe(183) for the T4 gene 32 protein and Phe(73) for the M13 gene 5 protein, were in good agreement with literature data. Actin and tubulin could not be cross-linked to the oligonucleotide. Aside from the insight into the biological activity of IF proteins that these data provide, they also demonstrate that this analytical method can be employed to study a variety of protein-nucleic acid interactions.

Amino-terminal polypeptides of vimentin are responsible for the changes in nuclear architecture associated with human immunodeficiency virus type 1 protease activity in tissue culture cells

Electron microscopy of human skin fibroblasts syringe-loaded with human immunodeficiency virus type 1 protease (HIV-1 PR) revealed several effects on nuclear architecture. The most dramatic is a change from a spherical nuclear morphology to one with multiple lobes or deep invaginations. The nuclear matrix collapses or remains only as a peripheral rudiment, with individual elements thicker than in control cells. Chromatin organization and distribution is also perturbed. Attempts to identify a major nuclear protein whose cleavage by the protease might be responsible for these alterations were unsuccessful. Similar changes were observed in SW 13 T3 M [vimentin(+)] cells, whereas no changes were observed in SW 13 [vimentin(-)] cells after microinjection of protease. Treatment of SW 13 [vimentin(-)] cells, preinjected with vimentin to establish an intermediate filament network, with HIV-1 PR resulted in alterations in chromatin staining and distribution, but not in nuclear shape. These same changes were produced in SW 13 [vimentin(-)] cells after the injection of a mixture of vimentin peptides, produced by the cleavage of vimentin to completion by HIV-1 PR in vitro. Similar experiments with 16 purified peptides derived from wild-type or mutant vimentin proteins and five synthetic peptides demonstrated that exclusively N-terminal peptides were capable of altering chromatin distribution. Furthermore, two separate regions of the N-terminal head domain are primarily responsible for perturbing nuclear architecture. The ability of HIV-1 to affect nuclear organization via the liberation of vimentin peptides may play an important role in HIV-1-associated cytopathogenesis and carcinogenesis.

Properties of the nonhelical end domains of vimentin suggest a role in maintaining intermediate filament network structure

To investigate the functional role of the nonhelical domains of the intermediate filament (IF) protein vimentin, we carried out transient transfection of constructs encoding fusion proteins of these domains with enhanced green fluorescent protein (EGFP). Expression of these fusion proteins did not have any effect on the endogenous IF networks of transfected cells. However, the head domain-EGFP fusion protein localized almost exclusively to the nucleus. This localization could be disrupted in a reversible fashion by chilling cells. Furthermore, the head domain was capable of targeting to the nucleus a strictly cytoplasmic protein, pyruvate kinase. Thus, the vimentin head domain contains information that specifically directs proteins into the nucleus. In contrast, the nonhelical tail domain of vimentin, when expressed as a fusion protein with EGFP, was retained in the cytoplasm. Cytoplasmic retention of tail domain-containing fusion proteins appeared to be dependent on the integrity of the microtubule network. Our results are consistent with a proposal that the nonhelical end domains of vimentin are involved in maintaining an extended IF network by exerting oppositely directed forces along the filaments. The head domains exert a nuclear-directed force while the tail domains extend the IF network toward the cell periphery via a microtubule-dependent mechanism.

Identification of the amino acid residues of the amino terminus of vimentin responsible for DNA binding by enzymatic and chemical sequencing and analysis by MALDI-TOF

The amino acid residues responsible for stable binding of nucleic acids by the intermediate filament (IF) subunit protein vimentin were identified by a combination of enyzmatic and chemical ladder sequencing of photo-cross-linked vimentin-oligodeoxyribonucleotide complexes and analysis by MALDI-TOF mass spectrometry. Three tryptic peptides of vimentin (vim(28)(-)(35), vim(36)(-)(49), and vim(50)(-)(63)) were found to be cross-linked to oligo(dG.BrdU)(12). dG.3'-FITC. From a methodological standpoint, it was necessary to remove the bulk of the bound oligonucleotide by digestion with nuclease P1 to get reproducible spectra for most of the peptides studied. Additionally, removal of the phosphate group of the residually bound dUMP or modification of the amino terminus of the peptide-oligonucleotide complexes with dimethylaminoazobenzene isothiocyanate dramatically improved the quality of the MALDI-TOF spectra obtained, particularly for the vim(28)(-)(35) peptide. A single Tyr residue within each of these peptides (Tyr(29), Tyr(37), and Tyr(52)) was unequivocally demonstrated to be the unique site of cross-linking in each peptide. These three Tyr residues are contained within the two beta-ladder DNA-binding wings proposed for the middle of the vimentin non-alpha-helical head domain. The experimental approach described should be generally applicable to the study of protein-nucleic acid interactions and is currently being employed to characterize the DNA-binding sites of several other IF subunit proteins.

Cleavage of vimentin by different retroviral proteases

Proteases (PRs) of retroviruses cleave viral polyproteins into their mature structural proteins and replication enzymes. Besides this essential role in the replication cycle of retroviruses, PRs also cleave a variety of host cell proteins. We have analyzed the in vitro cleavage of mouse vimentin by proteases of human immunodeficiency virus type 1 (HIV-1) and type 2 (HIV-2), bovine leukemia virus (BLV), Mason-Pfizer monkey virus (M-PMV), myeloblastosis-associated virus (MAV), and two active-site mutants of MAV PR. Retroviral proteases display significant differences in specificity requirements. Here, we show a comparison of substrate specificities of several retroviral proteases on vimentin as a substrate. Vimentin was cleaved by all the proteases at different sites and with different rates. The results show that the physiologically important cellular protein vimentin can be degraded by different retroviral proteases.

Integrating the actin and vimentin cytoskeletons. adhesion-dependent formation of fimbrin-vimentin complexes in macrophages

Cells adhere to the substratum through specialized structures that are linked to the actin cytoskeleton. Recent studies report that adhesion also involves the intermediate filament (IF) and microtubule cytoskeletons, although their mechanisms of interaction are unknown. Here we report evidence for a novel adhesion-dependent interaction between components of the actin and IF cytoskeletons. In biochemical fractionation experiments, fimbrin and vimentin coprecipitate from detergent extracts of macrophages using vimentin- or fimbrin-specific antisera. Fluorescence microscopy confirms the biochemical association. Both proteins colocalized to podosomes in the earliest stages of cell adhesion and spreading. The complex is also found in filopodia and retraction fibers. After detergent extraction, fimbrin and vimentin staining of podosomes, filopodia, and retraction fibers are lost, confirming that the complex is localized to these structures. A 1:4 stoichiometry of fimbrin binding to vimentin and a low percentage (1%) of the extracted vimentin suggest that fimbrin interacts with a vimentin subunit. A fimbrin-binding site was identified in the NH(2)-terminal domain of vimentin and the vimentin binding site at residues 143-188 in the CH1 domain of fimbrin. Based on these observations, we propose that a fimbrin-vimentin complex may be involved in directing the assembly of the vimentin cytoskeleton at cell adhesion sites.

Protein-protein interactions between keratin polypeptides expressed in the yeast two-hybrid system

Keratin filaments are obligatory heteropolymers of type I and type II keratin polypeptides. Specific type I/type II pairs are coexpressed in vivo. In contrast, all type I/type II pairs assemble into filaments in vitro, but the different pairs have different stabilities as demonstrated by treatment with increasing concentrations of urea. We have used the yeast two-hybrid system to analyse type I/type II interactions in a cellular context. We measured interactions between two different keratin pairs and we confirm the findings that K6+K17 form very stable heterodimers whereas K8+K18 interactions were weaker. The deletion of head domains did not reduce the strength of type I/type II interactions. Rather, the affinities were increased and the differences between the two pairs were retained in headless mutants. These findings argue against a major role of the head domains in directing heterodimer interactions and in defining heterodimer stabilities.

Structure and assembly properties of the intermediate filament protein vimentin: the role of its head, rod and tail domains

We have investigated the functional role of the non-helical end domains of vimentin on its assembly properties using truncated Xenopus and human recombinant proteins. Removal of the amino-terminal "head" domain yielded a molecule that did not assemble into 10 nm filaments but remained in a soluble oligomeric particle form with a sedimentation coefficient considerably smaller than that of wild-type vimentin (Vim(wt)). In contrast, removal of the carboxy-terminal "tail" domain had no obvious effect on the sedimentation characteristics. In particular, sedimentation equilibrium analysis under low ionic strength conditions yielded oligomeric particle species of Mr 135,000 to 360,000, indistinguishable from those obtained with Vim(wt). When induced to form filaments from this state by rapid dilution into filament forming buffer, Vim(wt) and Vim(deltaT) protein generated similar viscosity profiles. However, as determined by scanning transmission electron microscopy, under these conditions Vim(deltaT) formed filaments of heterogeneous diameter, corresponding to various distinct mass-per-length (MPL) values: whereas Vim(wt) yielded MPL values peaking between 40 and 45 kDa/nm, Vim(deltaT) filaments produced histograms which could be fitted by three Gaussian curves peaking between 37 and 131 kDa/nm. In contrast, when dialyzed against, instead of being rapidly diluted into, filament forming buffer, Vim(deltaT) gave histograms with one major peak at about 54 kDa/nm. The MPL heterogeneity observed for Vim(deltaT) was already evident at the earliest stages of assembly. For example, ten seconds after initiation, "unit-length" filament segments (58 to 63 nm) were formed with both wt and deltaT proteins, but the diameters were considerably larger for Vim(deltaT) compared to Vim(wt) (20(+/- 3) nm versus 16(+/- 3)nm), indicating a distinct role of the carboxy-terminal tail domain in the width control during unit-length filament formation. Despite this difference both Vim(deltaT) and Vim(wt) filaments appeared to grow stepwise in a modular fashion from such unit-length filament segments. This suggests that assembly occurred by a principally similar mechanism involving the end-on-fusion or annealing of unit-length filaments.

Contributions of the structural domains of filensin in polymer formation and filament distribution

Filensin and phakinin constitute the subunits of a heteropolymeric, lens-specific intermediate filament (IF) system known as the beaded-chain filaments (BFs). Since the rod of filensin is four heptads shorter than the rods of all other IF proteins, we decided to examine the specific contribution of this protein in filament assembly. For these purposes, we constructed chimeric proteins in which regions of filensin were exchanged with the equivalent ones of vimentin, a self-polymerizing IF protein. Our in vitro studies show that the filensin rod domain does not allow homopolymeric filament elongation. However, the filensin rod is necessary for co-polymerization of filensin with phakinin and seems to counteract the inherent tendency of the latter protein to homopolymerize into large, laterally associated filament bundles. Apart from the rod domain, the presence of an authentic or substituted tail domain in filensin is also essential for co-assembly with the naturally tail-less phakinin and formation of extended filaments in vitro. Finally, transfection experiments in CHO and MCF-7 cells show that the rod domain of filensin plays an important role in de novo filament formation and distribution. The same type of analysis further suggests that the end-domains of filensin interact with cell-specific, assembly-modulating factors.

Peripherin assembles into homopolymers in SW13 cells

The properties of full-length and mutant peripherins were studied in intermediate filament-less SW13 cells to define regions of peripherin that are essential for initiation of filament assembly. A full-length rat peripherin gene transfected into SW13 cells resulted in filament formation, consistent with the close structural relationship of peripherin to other type III intermediate filament proteins that readily form homopolymers. Translation of full-length rat peripherin is initiated predominantly at the second of two inframe AUGs. Deletions within the amino terminus of wild-type peripherin abolished its ability to form filaments in SW13 cells. In contrast, deletion of the entire carboxyl-terminal tail of peripherin did not affect its ability to form filamentous arrays in transfected SW13 cells. These results indicate that, of the intermediate filament proteins that are expressed in mature neurons, only peripherin and alpha-internexin are capable of making homopolymer intermediate filaments. In addition, mutations of the carboxyl tail of peripherin generally do not interfere with filament network formation.

Function of type I and type II keratin head domains: their role in dimer, tetramer and filament formation

To examine the role of the keratin head region and its subdomains in filament assembly we constructed several deletion mutants of type I and type II keratins and analysed their in vitro IF forming capacity. The delta K8 (1–74) and delta K18 (1–56), mutants formed only soluble oligomers, predominantly tetramers with their heterotypic partners. K8 mutants that retained either the entire (delta K8 (1–64)) or nearly the entire (delta K8 (1–66)) H1 subdomain formed some short and irregular IF-like structures with K18. However, filaments never reached the normal length and more protofilamentous material was observed. Analysis of the soluble complexes in 2 M guanidine-HCl indicated that tetramer formation was impaired in the truncated molecules. The length of the deletion correlated with the degree of tetramer destabilization. These results suggest that the head domain--specifically the H1 subdomain of type II keratins-plays a direct role in IF assembly. Its functions include a stabilization of the tetramer molecule, suggesting a role in directing the alignment of dimers as well as in elongation. We also analysed whether both head domains are required or if either type I or type II head domains alone are sufficient for IF formation. Hybrid molecules carrying their partner keratins head domains (K18 (8 head) and K8 (18 head)) were combined with their wild-type partners and tested for IF-forming ability. Both combinations formed filaments distinct from normal IF. The effect of the ‘replaced’ head domains was not compensated when both hybrid molecules were combined. Taken together, the results indicate that complete removal of the head domains of either K8 or K18 arrested IF assembly at the state of soluble oligomers. Replacement of the head domains by head domains of the complementary partner partly compensated for the effect. However, regular IF formation could not take place when either the head domain was missing or it was replaced by the partner's keratin head.

Vimentin's tail interacts with actin-containing structures in vivo

The tail domain of the intermediate filament (IF) protein vimentin is unnecessary for IF assembly in vitro. To study the role of vimentin's tail in vivo, we constructed a plasmid that directs the synthesis of a ‘myc-tagged’ version of the Xenopus vimentin-1 tail domain in bacteria. This polypeptide, mycVimTail, was purified to near homogeneity and injected into cultured Xenopus A6 cells. In these cells the tail polypeptide co-localized with actin even in the presence of cytochalasin. Two myc-tagged control polypeptides argue for the specificity of this interaction. First, a similarly myc-tagged lamin tail domain localizes to the nucleus, indicating that the presence of the myc tag did not itself confer the ability to co-localize with actin (Hennekes and Nigg (1994) J. Cell Sci. 107, 1019–1029). Second, a myc-tagged polypeptide with a molecular mass and net charge at physiological pH (i.e. -4) similar to that of the mycVimTail polypeptide, failed to show any tendency to associate with actin-containing structures, indicating that the interaction between mycVimTail and actin-containing structures was not due to a simple ionic association. Franke (1987; Cell Biol. Int. Rep. 11, 831) noted a similarity in the primary sequence between the tail of the type I keratin DG81A and vimentin. To test whether the DG81A tail interacted with actin-containing structures, we constructed and purified myc-tagged DG81A tail polypeptides. Unexpectedly, these keratin tail polypeptides were largely insoluble under physiological conditions and formed aggregates at the site of injection. While this insolubility made it difficult to determine if they associated with actin-containing structures, it does provide direct evidence that the tails of vimentin and DG81A differ dramatically in their physical properties. Our data suggest that vimentin's tail domain has a highly extended structure, binds to actin-containing structures and may mediate the interaction between vimentin filaments and microfilaments involved in the control of vimentin filament organization (Hollenbeck et al. (1989) J. Cell Sci. 92, 621; Tint et al. (1991) J. Cell Sci. 98, 375).

Pepstatin A: polymerization of an oligopeptide

Pepstatin A, a pentapeptide with the molecular weight of 686, is a naturally occurring inhibitor of aspartyl proteases secreted by Streptomyces species. Above a critical concentration of 0.1 mM at low ionic strength and neutral pH, it can polymerize into filaments which may extend over several micrometers. After negative staining, these filaments show a helical substructure with characteristic diameters ranging from 6 to 12 nm. Selected images at higher magnification suggest the filaments are composed of two intertwined 6 nm strands. This is in agreement with the optical diffraction analysis which additionally established a periodic pitch of 25 nm for the helical intertwining. Rotary shadowing of the pepstatin A filaments clearly demonstrated the right-handedness of the helical twist. In physiological salt solution or at higher concentrations of pepstatin A, a variety of higher order structures were observed, including ribbons, sheets and cylinders with both regular and twisted or irregular geometries. Pepstatin A can interact with intermediate filament subunit proteins. These proteins possess a long, alpha-helical rod domain that forms coiled-coil dimers, which through both hydrophobic and ionic interactions form tetramers which, in turn, in the presence of physiological salt concentrations, polymerize into the 10 nm intermediate filaments. In the absence of salt, pepstatin A and intermediate filament proteins polymerize into long filaments with a rough surface and a diameter of 15-17 nm. This polymerization appears to be primarily driven by nonionic interactions between pepstatin A and polymerization-competent forms of intermediate filament proteins, resulting in a composite filament. Polymerization-incompetent proteolytic fragments of vimentin, lacking portions of the head and/or tail domain, failed to copolymerize with pepstatin A into long filaments under these conditions. These peptides, as well as bovine serum albumin, were found to stick to the surface of pepstatin A filaments, ribbons and sheets. Independent evidence for direct association of pepstatin A with intermediate filament subunit proteins was provided not only by electron microscopy but also by UV difference spectra. Pepstatin A loses its ability to inhibit the aspartyl protease of the human immunodeficiency virus type 1 following polymerization into the higher order structures described here. The amazing fact that pepstatin A can spontaneously self-associate to form very large polymers seems to be a more rare event for such small peptides. The other examples of synthetic or naturally occurring oligopeptides discussed in this review which are able to polymerize into higher order structures possess a common property, their hydrophobicity, often manifested by clusters of valine or isoleucine residues.(ABSTRACT TRUNCATED AT 400 WORDS)

Keratin and keratinization

A flood of new knowledge and discoveries in the basic science of keratins and keratinization has appeared in the past several years. This review summarizes this recent information with a focus on the epithelial keratin polypeptides, keratin intermediate filaments, keratohyaline granule proteins, cell envelope formation and cell envelope proteins, "soft" keratinization, true disorders of keratinization (i.e., epidermolysis bullosa simplex and epidermolytic hyperkeratosis), and disease and drug effects on keratinization.

A conserved region in the tail domain of vimentin is involved in its assembly into intermediate filaments

Although the head and rod domains of intermediate filament (IF) proteins are known to play significant roles in filament assembly, the role of the tail domain in this function is unclear and the available information supports contradictory conclusions. We examined this question by comparing transfection of the same cDNA constructs, encoding vimentins with modified tail domains, into cell lines that do and do not contain endogenous IF proteins. By this approach, we were able to distinguish between the ability of a mutant IF protein to initiate assembly de novo, from that of incorporating into existing filament networks. Vimentins with modifications at or near a highly conserved tripeptide, arg-asp-gly (RDG), of the tail domain incorporated into existing IF networks in vimentin-expressing (vim+) cells, but were assembly-incompetent in cells that did not express IF proteins (vim-). The failure of the RDG mutant vimentins to assemble into filament arrays in vim- cells was reversible by re-introducing a wild-type vimentin cDNA, whereupon both wild-type and mutant vimentins coassembled into one and the same IF network. We conclude that the function of the tail domain of type III IF proteins, and possibly of keratins K8 and K18, in IF assembly is distinct from those of other domains; a region encompassing the RDG tripeptide appears to be important in the assembly process.

In vitro assembly properties of vimentin mutagenized at the beta-site tail motif

The intermediate filament (IF) proteins vimentin, desmin and peripherin share a 9-residue sequence motif (beta-site) located near the end of their COOH-terminal tail domain. Peptide inhibition experiments have previously suggested that the beta-site is involved in interactions that limit the lateral growth of IFs and prevent inappropriate filament-filament associations. To investigate this question further, we have constructed and expressed, in Escherichia coli, hamster vimentin bearing different mutations in the beta-site. We show here that a single exchange of glycine 450 with a valine residue, or an internal deletion of amino acids 444–452, strongly interferes with the normal assembly of IFs under in vitro conditions. These mutants polymerize into irregular fibrils that have a strong tendency to anastomose and laterally aggregate under isotonic conditions. In contrast, a non-conservative substitution of arginine 448 for glutamic acid does not significantly interfere with filament structure and yields subunits that polymerize into long, smooth filaments that show a slight aberration in thickness. All mutant proteins are soluble in low salt and form oligomers similar to the ones formed by wild-type vimentin. On the basis of these findings and on related observations, we propose that the tail domain of type III IF proteins contains important structural elements involved in lateral protofilament-protofilament interactions.

The roles of the rod end and the tail in vimentin IF assembly and IF network formation

Using mutagenesis, we investigated the importance of two vimentin domains: (a) a highly conserved segment near the carboxy end of the alpha-helical rod, and (b) the tail, with which the rod end is known to interact. As judged by in vitro filament assembly and expression in transiently transfected cells lacking an endogenous vimentin network, the rod-tail interaction is not essential for 10 nm filament structure in vitro or for formation of fibrous arrays in culture. However, when mutated, amino acid residues within the rod and the tail segments can cause perturbations in IF assembly and in IF network formation. Finally, our studies show that the vimentin tail seems to play a role both in thermodynamically stabilizing IF structure in vitro and in establishing proper IF networks in vivo.

The adenovirus L3 23-kilodalton proteinase cleaves the amino-terminal head domain from cytokeratin 18 and disrupts the cytokeratin network of HeLa cells

Immunofluorescence studies revealed that adenovirus induces a reorganization of the cytokeratin system in lytically infected HeLa cells. At 24 h postinfection, the cytokeratin network began to disassemble into prominent spheroid globules. By 36 h postinfection, host cell lysis occurred, accompanied by the formation of perinuclear cytokeratin clumps and additional spheroid globules. Immunoblots detected 41- and 44-kDa fragments of cytokeratin 18 and reduced levels of cytokeratin 7 at 24 and 36 h postinfection. Cytokeratin proteins isolated from HeLa cells at 36 h postinfection were deficient in filament polymerization. The 41-kDa proteolytic cytokeratin 18-specific fragment was purified, and its amino-terminal sequence was determined to be GGIQNEKETM. These residues correspond to amino acids 74 through 83 of cytokeratin 18, identifying a cleavage site at the junction of the globular head domain and the alpha-helical rod domain. Moreover, this truncation event occurs at a consensus cleavage site for the adenovirus L3 23-kDa proteinase. The temperature-sensitive mutant H2-ts1, which contains a mutation in the proteinase, neither induced cleavage of cytokeratin 18 nor precipitated the formation of spheroid globules during lytic infection at the nonpermissive temperature. The active proteinase is therefore required for cleavage of cytokeratin 18 and morphological rearrangement of the cytokeratins. We suggest that disruptions in the cytokeratin system weaken the mechanical integrity of the cell, thus promoting host cell lysis and release of progeny virions.

Binding of nucleic acids to intermediate filaments of the vimentin type and their effects on filament formation and stability

Guanine-rich polynucleotides such as poly(dG), oligo(dG)12-18 or poly(rG) were shown to exert a strong inhibitory effect on vimentin filament assembly and also to cause disintegration of preformed filaments in vitro. Gold-labeled oligo(dG)25 was preferentially localized at the physical ends of the aggregation and disaggregation products and at sites along filaments with a basic periodicity of 22.7 nm. Similar effects were observed with heat-denatured eukaryotic nuclear DNA or total rRNA, although these nucleic acids could affect filament formation and structure only at ionic strengths lower than physiological. However, whenever filaments were formed or stayed intact, they appeared associated with the nucleic acids. These electron microscopic observations were corroborated by sucrose gradient analysis of complexes obtained from preformed vimentin filaments and radioactively labeled heteroduplexes. Among the duplexes of the DNA type, particularly poly(dG).poly(dC), and, of those of the RNA type, preferentially poly(rA).poly(rU), were carried by the filaments with high efficiency into the pellet fraction. Single-stranded 18S and 28S rRNA interacted only weakly with vimentin filaments. Nevertheless, in a mechanically undisturbed environment, vimentin filaments could be densely decorated with intact 40S and 60S ribosomal subunits as revealed by electron microscopy. These results indicate that, in contrast to single-stranded nucleic acids with their compact random coil configuration, double-stranded nucleic acids with their elongated and flexible shape have the capability to stably interact with the helically arranged, surface-exposed amino-terminal polypeptide chains of vimentin filaments. Such interactions might be of physiological relevance in regard to the transport and positioning of nucleic acids and nucleoprotein particles in the various compartments of eukaryotic cells. Conversely, nucleic acids might be capable of affecting the cytoplasmic organization of vimentin filament networks through their filament-destabilizing potentials.

The two size alleles of human keratin 1 are due to a deletion in the glycine-rich carboxyl-terminal V2 subdomain

Two size variants of the type II human keratin 1 protein chain, termed 1a and 1b, have been described previously. Using amplification of genomic DNA by the polymerase chain reaction and sequence analysis we show here that the difference between these two alleles is due to a deletion of 21 bp in sequences encoding the V2 subdomain. This deletion corresponds to an entire glycine loop of seven amino acids. Pedigree analysis showed that the alleles are inherited as normal Mendelian traits. No additional alleles were detected in a survey of 88 alleles from 44 unrelated individuals, and the allelic frequency of 1a and 1b was 0.61 and 0.39. To determine the molecular basis of inherited dermatoses it is preferable to perform genetic linkage studies utilizing candidate genes directly as polymorphic markers. The PCR-based keratin 1 alleles characterized here, together with previously described PCR-based size variants in the keratin 10 gene, provide useful markers for the keratin clusters on chromosome 12 and 17, respectively.

Characterization of the nucleic acid-binding activities of the isolated amino-terminal head domain of the intermediate filament protein vimentin reveals its close relationship to the DNA-binding regions of some prokaryotic single-stranded DNA-binding proteins

In order to demonstrate that the nucleic acid-binding activities of vimentin are dictated by its Arg-rich N-terminal head domain, this was cut off at position Lys96 with lysine-specific endoproteinase and analysed for its capacity to associate with a variety of synthetic and naturally occurring nucleic acids. The isolated polypeptide (vim NT) showed a preference for single-stranded (ss) polynucleotides, particularly for ssDNAs of high G-content. A comparison of the sequence and predicted secondary structure of vim NT with that of two prokaryotic ssDNA-binding proteins, G5P and G32P of bacteriophages fd and T4, respectively, revealed that the nucleic acid-binding region of all three polypeptides is almost entirely in the beta-conformation and characterized by a very similar distribution of aromatic amino acid residues. A partial sequence of vim NT can be folded into the same beta-loop structure as the DNA-binding wing of G5P of bacteriophage fd and related viruses. As in the case of G5P, nitration of the Tyr residues with tetranitromethane was blocked by single-stranded nucleic acids. This and spectroscopic data indicate intercalation of the Tyr aromatic ring systems between the bases of the nucleic acids and thus the contribution of a stacking component to the binding reaction. The binding was accompanied by significant changes in the ultraviolet absorption spectra of both vim NT and single-stranded nucleic acids. Upon mixing of vim NT with nucleic acids, massive precipitation of the reactants occurred, followed by the quick rearrangement of the aggregates with the formation of specific and soluble association products. Even at very high ionic strengths, at which no electrostatic reaction should be expected, a distinct fraction of vim NT incorporated naturally occurring ssRNAs and ssDNAs into fast sedimenting complexes, suggesting co-operative interaction of the polypeptide with the nucleic acids. In electron microscopy, the complexes obtained from 28 S rRNA appeared as networks of extended nucleic acid strands densely covered with vim NT, in contrast to the compact random coils of uncomplexed RNA. The networks produced from fd DNA were heterogeneous in appearance and their nucleoprotein strands in rare cases were very similar to the rod-like structures of G5P-fd DNA complexes.

Species-specific recognition patterns of monoclonal antibodies directed against vimentin

Two commercially available monoclonal antibodies raised against the intermediate filament protein vimentin were characterized concerning their species-specific reaction pattern on vertebrate cells. The antibody V9 exhibited extensive reactivity with vimentin of all mammalian species tested, but specifically did not detect vimentin in mouse cells and chicken fibroblasts. The antibody VIM 3B4 recognized vimentin in cells of chicken and most mammalian species, except for rodent species. Characterization of the binding site of VIM 3B4 on human vimentin by limited proteolysis and immunoblotting as well as by sequence comparison strongly suggested that the epitope is located in the coil 2 part of the vimentin rod domain. Site-directed mutagenesis of a mouse vimentin cDNA clone followed by in vivo expression showed that VIM 3B4 could detect rodent vimentin containing a single amino acid substitution (valine for leucine) at position 353 of the mouse vimentin sequence. Practical application for this finding was demonstrated by the unequivocal identification of a modified murine vimentin protein, distinct from the endogenous vimentin, in a cytoplasmic intermediate filament network in mouse skin fibroblasts transfected with a recombinant plasmid expression vector.

Potential role of the viral protease in human immunodeficiency virus type 1 associated pathogenesis

Infection with the human immunodeficiency virus type 1 (HIV-1) results in a variety of pathological changes culminating in the acquired immune deficiency syndrome (AIDS). While most of these changes can readily be accounted for either by direct effects of HIV-1 on the immune system or by indirect effects of secondary infectious agents as a result of faulty immune surveillance, the direct cause for a number of disease states, including some neuropathies, myopathies, nephropathy, thrombocytopenia, wasting syndromes and increased incidence of cancers (primarily lymphoma) has remained an enigma. We have recently shown that the HIV-1 protease, a viral encoded enzyme necessary for virus maturation and infectivity, can cleave a variety of host cell cytoskeletal proteins in vitro. Potential substrates for the HIV-1 protease are found in all of the cell types affected in these unexplained diseases. Recent proposals suggest that elements of the cytoskeleton may play an important role in the regulation of large scale genetic regulation. We propose that some of the degenerative changes associated with infection by HIV-1 are a direct consequence of cleavage of host cell cytoskeletal proteins, which in turn may be responsible for the increased incidence of cancer in HIV-1 infected individuals as a result of the perturbation of the regulation of gene expression by cytoskeletal components.

Identification of a nonapeptide motif in the vimentin head domain involved in intermediate filament assembly

The assembly of soluble vimentin subunits into intermediate filaments (IFs) is dependent on information located in the amino-terminal domain. Using site-directed mutagenesis of a Xenopus laevis vimentin cDNA and an Escherichia coli production system to obtain pure mutated protein, we have identified, in the head domain, a nine amino acid motif (SSYRRIFGG), evolutionarily conserved from amphibia to man, which plays an important role in the orderly formation of IFs. Exchanges in the central di-arginine and in the two aromatic residues interfere with IF assembly of vimentin in vitro: on assembly under standard assembly conditions (160 mM-NaCl) most of the protein is included in dense aggregates, with a variable and minor proportion of IFs, whereas at lower ionic concentrations short and incomplete IF-like structures are formed. The deletion of the whole motif results in a protein that under standard assembly conditions (e.g. 160 mM-NaCl) predominantly and rapidly precipitates into large aggregates of non-IF material, whereas at lower ionic strength (e.g. 50 mM-NaCl) both IFs and dense aggregates are formed simultaneously. Our results show that the mutated protein can assume different forms at the same time and under the same conditions. This motif alone is insufficient for the formation of normal IFs as demonstrated by a mutant in which the motif has been brought closer to the alpha-helical rod domain by deletion of 55 internal amino acid residues. Corresponding observations have been made, by immunofluorescence microscopy, upon transfection of cultured epithelial cells lacking vimentin IFs. The importance of the head domain motif for the assembly and higher-order arrangement of IFs is discussed.

Intermediate filaments formed de novo from tail-less cytokeratins in the cytoplasm and in the nucleus

The roles of the different molecular domains of intermediate filament (IF) proteins in the assembly and higher order organization of IF structures have recently been studied by various groups but with partially controversial results. To examine the requirement of the aminoterminal (head) and the carboxyterminal (tail) domain of cytokeratins (CKs) for de novo IF formation in the living cell, we have constructed cDNAs coding for intact as well as head- and/or tail-less human CKs 8 and 18 and the naturally tail-less human CK 19, all under the control of the human beta-actin promoter. After transient and stable transfections of mouse 3T3-L1 cells, which are devoid of any CKs, we have studied, with such constructs, the resulting gene products by gel electrophoresis and immunolocalization techniques. By light and electron microscopy we show that extended cytoplasmic IF meshworks are formed from pairs of the type II CK 8 with the type I CKs 18 or 19 as well as from pairs of tail-less CK 8 with tail-less CKs 18 or 19 in the transfected cells, proving that the absence of the tail domain in both types of CKs does not prevent the de novo formation of regular IFs. Most surprisingly, however, we have observed spectacular alterations in the nucleocytoplasmic distribution of the IFs formed from tail-less CKs. In many of the transfected cells, a large part, or all, of the detectable CKs was found to occur in extensive IF bundles in the nucleoplasm. Intranuclear accumulations of CK deposits, however mostly nonfibrillar, were also observed when the cells had been transfected with cDNAs encoding tail-less CKs also lacking their head domains, whereas CKs deleted only in the head domain were found exclusively in the cytoplasm. The specific domain requirements for the assembly of cytoplasmic IF bundles are discussed and possible mechanisms of intranuclear accumulation of IFs are proposed.

A mathematical model for protein-induced lipid vesicle leakage: interaction of the intermediate filament protein vimentin and its isolated N-terminus with phosphatidylinositol vesicles

A mathematical model was developed to analyze the leakage of phosphatidylinositol small unilamellar vesicles induced by the intermediate filament protein vimentin and its isolated N-terminal polypeptide. This model describes the kinetic and steady-state characteristics of this vesicle leakage as a direct action of protein on the lipid bilayer. Moreover, qualitative information at the molecular level can be deduced about protein-protein or protein-lipid interactions from the derived initial rate of vesicle leakage and the value of vesicle leakage at steady-state condition as a function of the protein concentration. Additionally, quantitative data on the inhibitory effect of various substances (here Ca2+ or Mg2+) can also be derived. This approach offers a possibility to compare interactions occurring within different protein-lipid systems by determining the characteristic parameters for the respective kinetic and steady-state conditions.

Cleavage of phosphorylase kinase and calcium-free calmodulin by HIV-1 protease

Phosphorylase kinase and calcium-free calmodulin are digested by human immunodeficiency virus-1 protease. In phosphorylase kinase, the alpha subunit is preferentially hydrolyzed at arg748-val749. The beta subunit is cleaved only slowly at leu678-pro679, and calmodulin, the integral delta subunit of phosphorylase kinase, is not cleaved at all. However, free calmodulin in the calcium-depleted form showed to be a good substrate for the protease. Here the cleavage occurs at phe65-pro66 and met71-met72. This fast hydrolysis of free calmodulin can be blocked by micromolar concentrations of Ca2+ or millimolar concentrations of Mg2+.

A potential role for the COOH-terminal domain in the lateral packing of type III intermediate filaments

To identify sites of self-association in type III intermediate filament (IF) proteins, we have taken an "anti-idiotypic antibody" approach. A mAb (anti-Ct), recognizing a similar feature near the end of the rod domain of vimentin, desmin, and peripherin (epsilon site or epsilon epitope), was characterized. Anti-idiotypic antibodies, generated by immunizing rabbits with purified anti-Ct, recognize a site (presumably "complementary" to the epsilon epitope) common among vimentin, desmin, and peripherin (beta site or beta epitope). The beta epitope is represented in a synthetic peptide (PII) modeled after the 30 COOH-terminal residues of peripherin, as seen by comparative immunoblotting assays. Consistent with the idea of an association between the epsilon and the beta site, PII binds in vitro to intact IF proteins and fragments containing the epsilon epitope, but not to IF proteins that do not react with anti-Ct. Microinjection experiments conducted in vivo and filament reconstitution assays carried out in vitro further demonstrate that "uncoupling" of this site-specific association (by competition with PII or anti-Ct) interferes with normal IF architecture, resulting in the formation of filaments and filament bundles with diameters much greater than that of the normal IFs. These thick fibers are very similar to the ones observed previously when a derivative of desmin missing 27 COOH-terminal residues was assembled in vitro (Kaufmann, E., K. Weber, and N. Geisler. 1985. J. Mol. Biol. 185:733-742). As a molecular explanation, we propose here that the epsilon and the beta sites of type III IF proteins are "complementary" and associate during filament assembly. As a result of this association, we further postulate the formation of a surface-exposed "loop" or "hairpin" structure that may sterically prevent inappropriate filament-filament aggregation and regulate filament thickness.

Oxidation of thiol in the vimentin cytoskeleton

Sublethal doses of H2O2, which induces oxidative stress, cause substantial alteration to the vimentin cytoskeleton in various cell types. We have used a thiol-blot assay to assess thiol status in individual proteins from cell extracts. Vimentin thiol is oxidized in preference to other cytoskeleton proteins. Immunoblot analysis also demonstrated a loss of reactivity to an anti-vimentin monoclonal antibody under non-reducing conditions, possibly due to thiol-group oxidation. During induced oxidative stress a number of proteins become associated with the cytoskeleton extracts.

Polymerizing properties of pepstatin A

Pepstatin A, a pentapeptide aspartyl protease inhibitor, can spontaneously polymerize into filaments having a helical substructure and, after negative staining, characteristic diameters ranging from 6 to 12 nm. Optical diffraction analysis demonstrated that these filaments consist of a 6-nm-wide strand helically wound with a periodic pitch of 25 nm. Selected images suggest that these filaments may actually be composed of two, intertwined 6-nm-wide strands, an hypothesis not at variance with the diffraction data. These filaments may extend over several micrometers. At low ionic strength and neutral pH, the critical concentration for pepstatin A filament assembly is 0.1 mM. At higher pepstatin A concentrations or in physiological salt solutions, a variety of higher order structures were observed, including ribbons, sheets, and cylinders with both regular and twisted or irregular geometries. Pepstatin A polymerized into these higher order structures loses its ability to inhibit the aspartyl protease of the human immunodeficiency virus type 1. These results have implications not only for model studies on the polymerization of small peptides into higher order structures, but also for the practical development of soluble protease inhibitors.