Assembly and architecture of invertebrate cytoplasmic intermediate filaments reconcile features of vertebrate cytoplasmic and nuclear lamin-type intermediate filaments1

Article No. mb981995

J. Mol. Biol. (1998) 282, 601±617

Assembly and Architecture of Invertebrate Cytoplasmic Intermediate Filaments Reconcile Features of Vertebrate Cytoplasmic and Nuclear Lamin-type Intermediate Filaments Norbert Geisler1*, JuÈrgen SchuÈnemann1, Klaus Weber1, Markus HaÈner2 and Ueli Aebi2 1

Department of Biochemistry Max-Planck Institute for Biophysical Chemistry D-37018 GoÈttingen, Germany 2

Maurice E. MuÈller Institute Biozentrum, University of Basel, CH-4056 Basel Switzerland

The two major intermediate ®lament (IF) proteins from the esophagus epithelium of the snail Helix pomatia and the two major IF proteins from muscle tissue of the nematode Ascaris suum were investigated under a variety of assembly conditions. The lowest-order complexes from each of the four protostomic invertebrate (p-INV) IF proteins are parallel, unstaggered dimers involving two-stranded a-helical coiled coil formation of their 350 amino acid residue central rod domain (i.e. long-rod). In the electron microscope these are readily recognized by their distinct 56 nm long rod with two globular domains (i.e. representing the non-helical carboxy-terminal tail domain of the p-INV IF proteins) attached at one end, closely resembling vertebrate lamin dimers. The next-higher-order oligomers are tetramers, which are easily recognized by their two pairs of globular tail domains attached at either end of a 72 nm long central rod portion. According to their size and shape, these tetramers are built from two dimers associated laterally in an antiparallel, approximately halfstaggered fashion via the amino-terminal halves of their rod domains. This is similar to the NN-type tetramers found as the most abundant oligomer species in all types of vertebrate cytoplasmic IF proteins, which contain a 310 amino acid residue central rod domain (i.e. short-rod). As a ®rst step toward ®lament formation, the p-INV IF tetramers anneal longitudinally into proto®laments by antiparallel CC-type association of the carboxy-terminal halves of their dimer rods. The next step involves radial growth, occurring initially through lateral association of two fourchain proto®laments into octameric sub®brils, which then further associate into mature, full-width ®laments. Head-to-tail polymers of dimers and paracrystalline ®bers commonly observed with vertebrate lamins were only rarely seen with p-INV IF proteins. The globular domains residing at the carboxy-terminal end of p-INV IF dimers were studding the surface of the ®laments at regular, 24.5 nm intervals, thereby giving them a ``beaded'' appearance with an axial periodicity of about 24.5 nm, which is 3 nm longer than the corresponding 21.5 nm repeat pattern exhibited by short-rod vertebrate IFs. # 1998 Academic Press

*Corresponding author

Keywords: intermediate ®laments (IFs); protostomic invertebrate IFs; nuclear lamins; ®lament assembly; structural biology

Abbreviations used: p-INV, protostomic invertebrate; IF, intermediate ®lament; MPL, mass-per-length; STEM, scanning transmission electron microscope/microscopy; CC, antiparallel, approximately half-staggered association of two IF dimer rods via their C-terminal halves; NN, antiparallel, approximately half-staggered association of two IF dimer rods via their N-terminal halves. E-mail address of the corresponding author: [email protected] 0022±2836/98/380601±17 $30.00/0

# 1998 Academic Press

602

Introduction Intermediate ®laments (IFs) represent an obligatory cytoskeletal moiety of animal cells from simple metazoa to vertebrates. IFs are expressed in a tissue-speci®c manner, and a large number of different types of IF proteins exist. All IF proteins have have a distinct three-domain molecular architecture in common: a central a-helical rod domain, either 310 or 350 amino acid residues long, forms a parallel, unstaggered two-stranded a-helical coiled coil rod domain that is ¯anked by highly variable end domains, i.e. an N-terminal head domain and a C-terminal tail domain. The central a-helical rod domain forms the basic structural framework of IFs through assembly into a number of distinct, hierarchically ordered oligomeric sub-complexes. Structurally, the non-helical head and tail domains are highly diverse in different IF proteins; they are either buried within or exposed to the ®lament surface (Lees et al., 1988). Functionally, the two end domains play multiple roles: whereas at least part of the N-terminal head domain is required for ®lament assembly to occur (Traub & Vorgias, 1983; Kaufmann et al., 1985; Hatzfeld & Burba, 1994; BeuttenmuÈller et al., 1994), the C-terminal tail domain, while not essential for ®lament assembly, controls ®lament caliber (Kouklis et al., 1991; Bader et al., 1986; Hatzfeld & Weber, 1992; Heins et al., 1993; Heins and Aebi, 1994; Herrmann et al., 1996; Herrmann & Aebi, 1998). The central rod domain is not a continuous a-helical coiled coil but is divided into four distinct helical segments by three non-helical linkers: starting from the N-terminal end, these four helical segments are called 1A, 1B, 2A and 2B. Helices 1A plus 1B, and helices 2A plus 2B add up to approximately one half of the rod domain. Although the central rod is the IF protein domain with the highest sequence conservation, two different types of rods have evolved: (1) the short type, 310 amino acid residue long rod (i.e. short-rod) common to the vertebrate cytoplasmic IF proteins, and (2) the long type, 350-amino acid residue long rod (i.e. long-rod) common to the protostomic invertebrate (p-INV) cytoplasmic IF proteins and the nuclear lamins (Fisher et al., 1986; McKeon et al., 1986; Parry et al., 1986). The exact evolutionary position of the change from long-rod to short-rod cytoplasmic IF proteins is unknown. Current results put this event onto the deuterostomic metazoan branch prior to the divergence of the chordates, which include the vertebrates (Riemer et al., 1992; Bovenschulte et al., 1995). The long-rod arises through a 42 residue insertion in helix 1B of the short-rod. These 42 residues are in-frame with the remainder of helix 1B and reveal the seven residue long heptad-repeat pattern of hydrophobic residues diagnostic for a-helical coiled coils. Given that this 42 residue insertion assumes an a-helical coiled coil conformation, it lengthens helix 1B by 6.3 nm. Hence, compared to short-rod IFs, this 6.3 nm extension of helix 1B might interfere with

Intermediate Filament Structure and Assembly

the mode of association of at least some of the distinct oligomeric sub-complexes into long-rod IFs. Current knowledge on the polymerization behaviour of long-rod IF proteins is restricted to vertebrate nuclear lamins. Electron microscope investigations revealed that the soluble building blocks are predominantly parallel, unstaggered a-helical coiled coil dimers with their two C-terminal end domains appearing as a pair of globular heads at one end of an approximately 52 nm long rod. Although different lamins do not behave uniformly, the general assembly scheme proceeds by head-to-tail association of dimers into long dimeric proto-®laments (Aebi et al., 1986; Heitlinger et al., 1991; Sasse et al., 1997; Stuurmann et al., 1998). Given appropriate conditions, these proto®laments next associate laterally in an antiparallel, approximately half-staggered fashion to yield ®laments of variable thickness and, ultimately, thick paracrystalline ®bers. Whereas no stable 10 nm ®laments are yielded at steady state in vitro, it is conceivable that lamin ®laments with a uniform diameter are formed in vivo. Therefore, assembly of long-rod lamins follows a different hierarchic route from that used by short-rod vertebrate cytoplasmic IF proteins. Dimers of the latter type ®rst associate laterally in an antiparallel, approximately half-staggered fashion into tetramers, octamers and, eventually, unit-length ®laments, before these anneal longitudinally into mature 10 nm ®laments (Herrmann et al., 1996; Herrmann & Aebi, 1998). Epithelial cells from the esophagus of the snail Helix pomatia and muscle tissue from the nematode Ascaris suum contain cytoplasmic ®laments whose molecular building blocks are IF-type proteins based on their primary sequence (Weber et al., 1988, 1989; Bartnik et al., 1985, 1986; Dodemont et al., 1990). In both species, these ®laments appear to be assembled from two IF-type proteins: in Helix, alternative RNA processing of transcripts from a single gene yields one variant with a lamin B-like C-terminal tail domain, and a second, nearly tailless variant being otherwise identical with the ®rst except for the very last C-terminal residue, which is different from the corresponding residue in the tail-bearing variant. The two Ascaris proteins are encoded by different genes with an overall identity of 55% and with only a small difference in length residing in the N-terminal head domain. The four protostomic invertebrate proteins exhibit strong sequence similaritiy to vertebrate nuclear lamin B, which extends to the C-terminal tail domains. However, they do not harbor the short and conserved sequence stretches ¯anking the central rod domain that are known to in¯uence the head-to-tail assembly of lamin dimers through their phosphorylation state (Peter et al., 1991; for a recent review, see Stuurmann et al., 1998). In addition, as cytoplasmic proteins, the p-INV IF proteins, like their vertebrate homologs, do not contain a nuclear localization signal, which in lamins is located in the C-terminal tail domain.

Intermediate Filament Structure and Assembly

Preliminary in vitro assembly experiments have indicated that all four p-INV IF proteins can polymerize into apparently normal-looking ®laments (Weber et al., 1988, 1989; Bartnik et al., 1985, 1986). We have now investigated the assembly and architecture of these invertebrate cytoplasmic IFs in detail. Our data provide evidence that these invertebrate cytoplasmic IF proteins unite structural features and assembly behaviour of vertebrate cytoplasmic and nuclear lamin-type IF proteins.

Results Overview The four intermediate ®lament proteins used in this study are closely related to each other and to all other members of the IF protein family (see Figure 1A and B for their schematic representation and their analysis by SDS-PAGE). In spite of their structural similarity, the Helix and Ascaris IF proteins exhibit unique polymerization characteristics different from most or even all other known types of IF proteins. The ®rst effective parameter regulating the polymerization of p-INV IFs is the ionic strength: high ionic strength favors oligomerization of the p-INV IF proteins, whereas low ionic strength promotes polymerization into sub®lamentous polymers and mature, full-width ®laments. This is the case for the IF proteins from Helix and Ascaris . The effect of the pH value as a second parameter is different for both types of proteins: the assembly of the two Ascaris proteins proceeds from pH 7.5 (oligomers) over pH 8.0 (intermediate assembly stages) to pH 7.0 (mature ®laments), while Helix oligomers are formed at pH 8.5 and

603 mature ®laments assemble at pH 8.0. Accordingly, although many more buffer conditions were tested initially, ®ve were eventually used in this study. These were: (1) Helix oligomer buffer (20 mM TrisHCl (pH 8.5), 0.25 M NaCl) for oligomers; (2) Helix ®lament buffer (1 mM Tris-HCl (pH 8.0), no additional salt) for ®laments; (3) Ascaris oligomer buffer (10 mM Tris-HCl (pH 7.5), 170 mM NaCl) for oligomers; (4) Ascaris high pH ®lament buffer (10 mM Tris-HCl (pH 8.0), no additional salt) for intermediate polymerization stages; and (5) Ascaris ®lament buffer (10 mM Tris-HCl (pH 7.0), no additional salt) for mature ®laments (see Materials and Methods for a more complete description of the various buffers). An overview of the polymerization behaviour is presented in Figure 2 with an equimolar mixture of the two Ascaris proteins: upon equilibration in Ascaris oligomer buffer, most of the protein is found as dimers and tetramers (Figure 2A). A change to Ascaris high pH ®lament buffer induces incomplete polymerization (Figure 2B): the ®lamentous products appear to be made of loosely woven proto®laments and sub®brils, with a signi®cant fraction of the protein remaining as unpolymerized oligomers. In Ascaris ®lament buffer, polymerization leads to mature full-width ®laments (Figure 2C). At ®rst sight, these ®laments exhibit at least some of the characteristic features shared by all IFs: these include, for instance, the typical axial beading pattern (Henderson et al., 1982; Milam & Erikson, 1982). The two Helix proteins exhibit a similar assembly behaviour, governed by similar rules: high ionic strength and high pH conditions produce small oligomers, mostly dimers and tetramers, whereas low ionic

Figure 1. Organization and PAGE of IF proteins from Helix and Ascaris. Note that for reasons of convenience, the protein names used here are not the original names given to the proteins when ®rst puri®ed and characterized (Weber et al., 1988, 1989; Bartnik et al., 1985, 1986; Dodemont et al., 1990; for details, see Materials and Methods). A, The primary structural organization of Helix-high and Helix-low proteins (Hel-high, Hel-low) and Ascaris-high and Ascaris-low proteins (Asc-high, Asc-low) is displayed in a schematic view. An example of a vertebrate cytoplasmic short-rod IF protein is included for comparison. The central part represents the rod domain with its distinct helical segments (i.e. 1A, 1B and 2). The position of the 42 residue insert in helix 1A is indicated. Head and Tail designate the N and C-terminal end domains. Domain borders are marked by numbers referring to the primary sequences of the corresponding p-INV IF proteins (Weber et al., 1988, 1989; Dodemont et al., 1990; and EMBL/Genbank accession numbers Q 17047 and Q 17049). B, SDS-PAGE of the Ascaris and Helix proteins used in this study run on a 7.5% (w/v) polyacrylamide gel (Laemmli, 1970). A molecular mass scale is included on the right-hand side of the gel.

604

Intermediate Filament Structure and Assembly

strength and a lower pH value promote polymerization into mature full-width ®laments (see below). At their first level of structural organization, Helix and Ascaris polypeptides form lamin-like dimers Under all conditions tested, dimers were never found to be a distinct species. Instead, they always occur together with signi®cant amounts of tetramers and octamers. In high ionic strength buffers, Helix-high protein (see Materials and Methods) and the two Ascaris proteins yield dimers that appear morphologically reminiscent of the wellknown vertebrate lamin dimers (Figure 3A to C: Aebi et al., 1986; Heitlinger et al., 1991; Sasse et al., 1997). Accordingly, they are made of an approximately 55 nm long rod (Table 1), which is often kinked in the middle. This rod is ¯anked at one end by two globular domains representing the C-terminal tail domains of the p-INV IF proteins. As documented in Figure 3C, the dimers from all three tail-bearing proteins look very similar, including those from the mixed Ascaris proteins. Helix-low dimers appear different, since they lack the two globular tail domains, in line with the primary sequence. If heterodimerization between the two Helix proteins occurs, then the heterodimers should carry one tail only. To date, it has not been possible to identify these positively because dimers built from the high protein alone also frequently exhibit only one tail. This latter observation may simply mean that the two tail domains have coalesced into one upon specimen preparation. In some instances, the p-INV IF dimers yield, in addition, one or two smaller globules at the opposite end of the rod, most likely representing the N-terminal head domains (Figure 3B and C; arrowheads). Such smaller globules are found also with some of the tailless Helix-low dimers (Figure 3C; arrowheads). Sometimes, and especially with the Ascaris dimers, the head globules appear almost as big as those representing the tail domains. It is conceivable that, for instance, salt or glycerol may cause ``decoration'' (Fowler & Aebi, 1983) of the small N-terminal globules so that they appear enlarged (Figure 3C; arrowheads). Figure 2. Polymerization of mixtures of Ascaris high and low proteins. The polymerization of an equimolar mixture of the two Ascaris proteins is shown at various stages of progression. A, Oligomer conditions: the solution consists predominantly of dimers and tetramers with some higher oligomers (Ascaris oligomer buffer). B, Under suboptimal polymerization conditions (Ascaris high pH ®lament buffer), a variety of products is depicted ranging from dimers, tetramers and octamers to proto®laments and proto®brils. For a more complete description, see the text. C, Mature ®laments polymerized under optimal conditions (Ascaris ®lament buffer, at 37 C). The magni®cation is identical in all micrographs, and the scale bar represents 250 nm.

Helix and Ascaris dimers form antiparallel, approximately half-staggered tetramers As documented in Figure 4B and C, the high ionic strength conditions of Helix oligomer and Ascaris oligomer buffer promote, in addition to dimers, the formation of tetramers. These are readily recognizable by their two pairs of globular domains residing at either end of an approximately 75 nm long rod (see Table 1). This structural constellation indicates that these distinct oligomers are built from two dimers, which are associated laterally in an antiparallel, approximately half-staggered fashion so that the N-terminal halves of their

605

Intermediate Filament Structure and Assembly

Figure 3. Dimers obtained from Helix and Ascaris proteins. A, Schematic drawing of dimers after different amounts of bending. The N and C-terminal end domains are indicated by N and C, respectively. B, Representative, higher magni®cation micrograph of a dimer obtained from Helix-high protein: the coiled coil rod domain is ¯anked at its C-terminal end (at the right-hand side of all molecules) by two globular tail domains. In some instances, the smaller N-terminal head domains are depicted (arrowheads in B and C). C, Gallery of selected dimers built from the four individual proteins and from a stoichiometric mixture of the two Ascaris proteins (high and low; see Figure 1B) as indicated. The rods are seen after different amounts of kinking. Dimers from Helix-low protein clearly differ from the other species by the absence of their C-terminal tail domain (see the representation of the two proteins in Figure 1A) but are occasionally ¯anked at one end by smaller globules representing the N-terminal head domain (arrowheads). For length data of dimers, see the histograms in Figure 4D. The scale bars in B and C represent 100 nm.

55 nm long rod domains interact with each other. As a consequence of this half-staggered association, the central third of the tetramer rod appears distinctly thicker than the outer two-thirds, where the C-terminal halves of the dimeric rods extend freely towards the globular tail domains. As with unassociated dimers (see Figure 3), the two dimer rods within the tetramer are often kinked in the middle. Sometimes, and especially in the case of the Helixhigh protein, the C-terminal halves of the two dimer rods within the tetramer are bent inward by more than 90 , thereby giving the tetramer the shape of a Z (see the bottom molecule in the Helixhigh gallery in Figure 4C). Occasionally, (e.g. see the molecule in Figure 4B indicated by a black and a white arrowhead), the N-terminal head domains

are clearly visible as small separate globules ¯anking the central overlapping part of the 75 nm long tetramer rod. The three types of homotypic tetramers from the tail-carrying INV IF proteins (i.e. Helix-high, Ascaris-low and Ascaris-high), as well as the heterotypic tetramers (i.e. Ascaris-mix), are of basically very similar construction (Figure 4C) with the exception that the Helix-high tetramers are frequently more strongly kinked as described above. The electron micrographs of the tetramers (Figure 4C) are fully consistent with the schematic representation of the tetramer shown in Figure 4A. The length histograms in Figure 4D display dimers and tetramers side-by-side because they were always found together under all conditions tested. Apparent differences in the average lengths

Table 1. Dimensions of INV IF oligomers and ®laments (in nm) Oligomers

Filaments

Dimers

Helix Ascaris

Tetramers

Helix Ascaris

Octamers Diameter

Ascaris Helix

Axial periodicity

Ascaris Helix Ascaris Chicken desmin

High High Low High High Low Mix Mix/high High (neg. stain) Low (neg. stain) Mix (neg.stain) High High subfibrils Mix subfibrils Mix full filaments (control)

57  5 55  4 56  3 75  4 70  5 73  6 72  6 129  5 10.0 10.0 5-10 24.4 24.2 24.1 24.0 21.5

Filament lengths and axial periodicities were determined using metal-shadowed specimens (Figures 2 to 6), and apparent ®lament diameters were measured from negatively stained samples (Figures 8 and 9).

606

Intermediate Filament Structure and Assembly

of the four proteins are rather small and support the view that all tetramers analyzed represent the same type of structure. Tetramers anneal longitudinally first into octamers and then into long protofilaments

Figure 4. Structural organization of tetramers from Helix-high, and from Ascaris-high and Ascaris-low proteins. A, Schematic drawing of p-INV IF tetramers. The N-terminal halves (i.e. helix 1B) of the rod overlap in the central region while the C-terminal halves (i.e. helix 2) extend towards the two ends of the molecules. Three molecules with different extents of kinking of their rods (i.e. from straight to Z-shaped) are displayed. B, A representative tetramer from Helix-high protein at high magni®cation. The micrograph reveals the two outer pairs of C-terminal tails and the two rods in between with the central overlapping (i.e. thicker) portion. N-terminal head domains (arrowheads) appear as smaller globules at the borders of the central overlapping portion of the two dimer rods. C, Gallery of selected tetramers from Helix-high, the individual Ascaris proteins, and the mixed Ascaris proteins, high plus low. The selected molecules are shown in various kinked forms, as shown schematically in A. Small globules representing the N-terminal head domain are marked by arrowheads. The Helix tetramers were obtained in Helix oligomer buffer, and the Ascaris tetra-

Octamers are found in great numbers under conditions where polymerization has proceeded towards ®laments but is not yet completed. For example, in the case of the mixed Ascaris proteins this happens upon dialysis against low ionic strength buffer at pH 8.0 (i.e. with Ascaris high pH ®lament buffer; see Materials and Methods) when some ®laments have already formed but a high fraction of oligomers including dimers and tetramers still remains. A micrograph taken under this condition is presented in Figure 2B. The great majority of the octamers displayed in Figure 5 are from Ascaris-mixed and Ascaris-high proteins. For unknown reasons, no signi®cant number of Ascaris-low octamers were seen under these conditions. A possible explanation may be the relatively strong dependence of the low protein on the presence of the high protein for polymerization (for details, see Figure 8). Similarly, relatively few octamers were formed from the Helix-high protein. Because they are missing a signi®cant tail domain, Helix-low octamers were dif®cult to identify and have therefore not been investigated. Octamers arise from the annealing of two tetramers by a lateral antiparallel association of their helix 2 coiled coils (see Figure 5A for a representation). This notion is supported by the characteristic distribution of the globular tail domains along the octamer: if T means a pair of globular tails, ± a single dimeric coiled coil segment from either helix 1 or helix 2, and ˆ a corresponding tetrameric coiled coil segment, then the tail distribution is described as T ± ˆ T ˆ T ˆ ± T , i.e. there are two terminal, and two central globular pairs of tail domains (see Figure 5A and C, for details, and marking of the globular tails by arrowheads). In most instances, the two outer and the two central pairs of globular tails are clearly displayed, but occasionally one of the two tails of a pair is hidden either under the other tail or under the rod. In other instances, the two tail domains of a pair are displayed as a close doublet on one side of the rod. The average overall length of the octamers measures 129 nm (Figure 5B and Table 1), which agrees remarkably well with the calculated length

mers in Ascaris oligomer buffer or Ascaris high pH ®lament buffer. The scale bars in B and C represent 100 nm. D, Length histograms of dimers and tetramers. Both types of oligomers are included because both occur simultaneously in the buffer solutions employed. Continuous lines, experimental values; broken lines, Gaussian curves ®tted to the experimental data. See Table 1 for the experimental length values. For examples of dimers, see Figure 3C.

Intermediate Filament Structure and Assembly

607

Figure 5. Tetramers anneal longitudinally, ®rst to octamers and further to proto®laments. A, Schematic representation of octamers: two tetramers anneal longitudinally to an octamer; a, the relative positions of the two tetramers are indicated; b, the two tetramers associate by aligning helices 2 in an antiparallel manner. The resulting characteristic distribution of tail domains (see the text for details) is indicated by arrowheads; c, octamers kinked to various degrees as is depicted in the gallery dislayed in C. B, Histogram of octamer lengths (octamers from Ascaris-mixed proteins, i.e. high plus low). C, Gallery of octamers from Ascaris-mixed proteins, Ascaris-high protein, and one octamer from Helix-high protein. The Ascaris octamers were found in Ascaris high pH ®lament buffer and in Ascaris oligomer buffer, whereas the Helix-high octamer was found in Helix oligomer buffer. D, Schematic view of the elongation of octamers to proto®laments by addition of further tetramers. Elongation is accomplished by antiparallel alignment of helices 2, the same mechanism as drawn in A above for octamer formation by staggered association of two tetramers. E, Micrographs displaying a 12-mer (octamer plus one tetramer; E, a) and a 16-mer (octamer plus two tetramers, or octamer plus octamer; E, b). The central pairs of C-terminal tail domains are indicated by arrowheads (see the text for a full explanation). Longer proto®laments are displayed in E, c. All micrographs are at the same magni®cation: the scale bars represent 100 nm.

608

Intermediate Filament Structure and Assembly

Figure 6 (legend opposite)

Intermediate Filament Structure and Assembly

from the model structure of the octamer: accordingly, its length is ®ve times the length of helix 1 or 2, i.e. 5  26 nm ˆ 130 nm. The octamer further elongates to proto®laments by addition of more tetramers at either end with the same mechanism as the condensation of two tetramers to an octamer, i.e. by antiparallel lateral CC-type association of helices 2. Figure 5E displays two micrographs, one with a 12-mer (a) and one with a 16-mer (b), produced by addition of one and two tetramers, respectively, to an octamer. Again, the distribution of tails and helices is as shown for the octamer, exept that these molecules display four and six central pairs of globular C-terminal domains, respectively, instead of two as for the octamer (marked by arrowheads). Examples of longer extensions to proto®laments are displayed in Figure 5E, c. Accordingly, these longer proto®laments, as the octamers, the 12-mer and the 16-mer, are four polypeptide chains wide. The elongation process from tetramers to proto®laments as described above is, however, not likely to proceed only by successive addition of tetramers but is likely to be additionally accomplished by condensation of longer oligomers. Because the tetramer is an apolar structure, higher-order structures built from it will also be bipolar (or apolar). This means that octamers, the larger oligomers and proto®laments, as well as mature full-width ®laments built from them, are also bipolar structures. Tetrameric protofilaments associate laterally to octameric protofibrils and eventually to full-width filaments As illustrated in Figure 5 and discussed above, two tetramers associate to octamers via a CC-type antiparallel association of their helices 2, and they further elongate to long tetrameric proto®laments by the same mechanism. Lateral association of two such proto®laments leads to the formation of proto®brils with eight polypeptides per cross-section (Figure 6A). Lateral association of several such octameric proto®brils in turn leads to the formation

609 of full-width ®laments as displayed in Figure 6C and D (more full-width ®laments may be gathered in the overview in Figure 2C). Some examples of this gradual lateral association of proto®laments into proto®brils and beyond are shown in Figure 6B: a reveals micrographs from a solution of Ascaris-mixed proteins under suboptimal polymerization conditions at pH 8.0 and at low ionic strength (high pH ®lament buffer). Under these conditions, two proto®laments loosely associate laterally into octameric proto®brils. The rod domains are depicted within the core of the sub®laments, and the C-terminal globular tails project radially with an axial repeat of about 25 nm (see Table 1). An example is also included (left, arrowheads) where the two proto®laments have separated to reveal the two-stranded substructure of the proto®bril. Examples of more tightly packed proto®brils are shown in Figure 6B, b. These were found predominantly under optimal polymerization conditions in 10 mM Tris-HCl buffer at pH 7.0 at 37 C (Ascaris ®lament buffer). With these examples it is dif®cult to be sure exactly how many proto®laments are contained within a proto®bril but the images are most compatible with two. Irrespective of the exact number of proto®laments, the proto®brils reveal their structure: the central rod domains constitute the core or shaft of the proto®brils, wheras the globular C-terminal end domains decorate the shaft of the ®bril at regular intervals, thereby yielding a periodic axial repeat pattern. The average repeat length is about 24.5 nm (Table 1). Thus the axial peridicity of these p-INV IFs exceeds that of vertebrate desmin IFs (i.e. 21.5 nm; see Figure 6D, b, and Table 1) by approximately 3 nm. Examples of full-width ®laments such as displayed in Figure 6D, a, a are evidently polymorphic with regard to their width, so there may exist several distinct width classes (see below and Figure 7). The thinner ®laments appear similar in width to chicken desmin ®laments as shown for comparison in Figure 6D, b. The thicker ®laments may have arisen through lateral association of two thinner ®laments, or by addition of one or more proto®laments or proto®brils to a thin ®lament

Figure 6. Lateral alignment of proto®laments to full-width ®laments. A, Schematic description of the lateral association of tetrameric proto®laments to octameric proto®brils; compare the scheme to the micrographs below. B, a, Two micrographs revealing polymerization products obtained under suboptimal polymerization conditions (i.e. Ascaris high pH ®lament buffer): the ®lamentous polymers appear loosely woven and frequently seem to be built from two proto®laments, a scenario particularly evident in the micrograph on the left, where the two proto®laments have separated (arrowheads). Hence, these polymerization products might represent proto®brils with eight polypeptides per cross-section. B, b, Sub®lamentous oligomers obtained under optimal polymerization conditions in Ascaris ®lament buffer: this width class represents either octameric proto®brils as above, or it reveals intermediates between proto®brils and full-width ®laments. The globular C-terminal tail domains are decorating the ®laments at regular intervals, thereby yielding a 24.5 nm axial repeat pattern (for the repeat lengths see Table 1). C, Schematic drawing illustrating lateral association of octameric proto®brils to full-width ®laments. The number of proto®laments or proto®brils per ®lament cross-section is arbitrary and will depend on the actual width class of the ®lament as shown below. D, a, Full-width ®laments obtained in Ascaris ®lament buffer at 37 C under optimal in vitro polymerization conditions. The ®laments appear in different width classes, the thinner ones could be of the same width class (i.e. 32 polypeptides per ®lament cross-section) as the desmin ®laments shown for comparison in D, b (see Figure 7 for mass-per-length (MPL) data). All micrographs are at the same magni®cation and the scale bar represents 250 nm.

610

Intermediate Filament Structure and Assembly

Figure 7. Mass-per-length (MPL) data of full-width ®laments from Ascaris-high plus low protein mixtures. A, Unstained, 0.1% (v/v) glutaraldehyde-®xed p-INV IFs used to collect MPL data recorded in a STEM by the annular dark-®eld mode. The positions and size of measured areas are indicated. The scale bar represents 250 nm. B, MPL histogram of full-width ®laments as displayed in A. The main peak, i.e. peak I, 38(11) kDa, may contain 32 polypeptides in cross-section and the MPL increments (20 kDa/nm) correspond roughly to half-width ®laments (i.e. those harboring 16 polypeptides per ®lament cross-section) of the major MPL class (i.e. the one corresponding to ®laments with 32 polypeptides per ®lament cross-section).

(see below and Figure 7 for mass per length determination of full-width ®laments). All ®laments exhibit a distinct beading with a 24± 25 nm axial repeat (see Table 1). Based on their mass-per-length (MPL), fully polymerized filaments are polymorphic The mass-per-length (MPL) data for mature ®laments assembled from mixed Ascaris proteins are presented in Figure 7. The thinner ®laments such as depicted on the left in Figure 6D, a after glycerol spraying and rotary metal shadowing, yield a MPL of 38 kDa/nm, corresponding to about 32 polypeptides per ®lament cross-section. The MPL increments to thicker ®laments correspond to 16 polypeptides or two octameric proto®brils, or half a 38 kDa/nm type ®lament. MPL varies between different ®laments and changes along a ®lament. Both homotypic and heterotypic INV IFs can be assembled The two Ascaris IF proteins polymerize into ®laments both as a mixture of the high and low polypeptides and as individual polypeptides (Figure 8A, d and e). However, when both polypeptides are present at approximately equimolar

amounts, heterotypic polymerization is strongly favored over homotypic polymerization. This notion is obtained from the electron microscopical appearance of the dialyzed protein solutions. The individual polypeptides, especially the Ascaris-low protein, yield only small amounts of ®laments, with the unpolymerized material remaining as small oligomers. Homotypic ®laments exhibit a clear beading pattern, with a 24 nm axial repeat (Figure 8A, d and e) similar to that of the mixed type ®laments (see Table 1). The blot overlays displayed in Figure 8B with the four biotinylated INV proteins complement this observation: the Ascaris-low protein strongly recognizes the Ascaris-high protein, but itself by comparison only very weakly. In contrast, the Ascaris-high protein recognizes both the low protein and itself in equal amounts (marked by arrowheads). This binding behaviour indicates a strong dependence of the Ascaris-low protein on the high protein. The situation is rather different for the Helix proteins, as they both recognize themselves and each other equally well (Figure 8B, indicated by dots). This is plausible, as their primary structures are identical exept for the presence of the C-terminal tail domain in the high-protein and the very last C-terminal residue of the low protein. In

Intermediate Filament Structure and Assembly

addition, the blots indicate that the Ascaris proteins, although they are long-rod proteins, are not recognized by the Helix proteins. Similarly, the vertebrate short-rod cytoplasmic protein desmin is not recognized by the Ascaris proteins.

611 Although the polymerization steps of the Helix proteins beyond dimers and tetramers have not yet been followed in great detail, it is conceivable that they behave similarly to the two Ascaris proteins (Figure 8A, a ± c). The homotypic polymerization

Figure 8. Filaments polymerized by single Helix and Ascaris proteins. A, Helix ®laments polymerized from high protein in Helix ®lament buffer are displayed after glycerol spraying/low-angle rotary metal shadowing in a, and negatively stained, un®xed specimens from single Helix-high and Helix-low proteins in b, high and c, low. Shadowed examples of ®laments formed by both single Ascaris proteins in Ascaris ®lament buffer at 37 C are displayed in d, low protein and e, high protein. Note the distinct 24.5-nm axial beading exhibited by the metal-shadowed specimens (a, d, and e; for ®lament widths and axial repeat lengths; see Table 1). The magni®cation is the same for all micrographs, and the scale bar represents 250 nm. B, Blot overlay tests of the mutual recognition of the p-INV IF protein pairs from Helix and Ascaris: the upper row displays the protein samples separated by SDS-PAGE and after protein staining as indicated above. The lower row presents the blots after incubation and staining with the biotinylated proteins as given below. First, both Helix polypeptides recognize themselves and each other equally well (left two rows, indicated by dots). Second, the Ascaris-low protein recognizes primarily the Ascaris-high protein strongly. In contrast, it recognizes itself at a signi®cantly attenuated level. The high protein, on the other hand, recognizes both the small protein and itself equally well (right two columns, indicated by arrowheads below). This characteristic behaviour is seen also when mixtures of the two Ascaris proteins are used; see the extreme right-hand side lanes, which are also marked by arrowheads at the side. In addition, it is shown that the Ascaris proteins are not detected by the Helix proteins and, moreover, that vertebrate desmin is not detected by the Ascaris proteins (all four columns, side lanes).

612 potentials of Helix-high and Helix-low proteins are similar and heterotypic polymerization has no obvious advantage. Filaments assembled from the Helix-high protein exhibit a beaded surface resulting from the exposure of the C-terminal globular tail domains at the surface of the ®laments with an axial repeat of 25 nm (Figure 8A, a and Table 1). In contrast, the ®laments formed by the Helix-low protein, which has a tail domain only a few residues long, appear smooth without any obvious beading (not shown). Negative staining turns out to be less effective in revealing the globular tail domains and the beading of ®laments assembled from the Helix proteins ( Figure 8A, b and c) and the Ascaris proteins (not shown), however, it yields a more realistic estimation of their width (see Figure 8A, a and b, for ®laments made of Helixhigh protein). Shape and morphology of INV IFs are closer to lamin IFs than to type III IFs Negatively stained ®laments from three different types of IFs polymerized under optimal conditions are displayed in Figure 9: (A) ®laments from recombinant Xenopus vimentin as a member of the short-rod vertebrate type III cytoplasmic proteins, (B) ®laments from a stoichiometric mixture of the two Ascaris proteins polymerized in Ascaris ®lament buffer at 37 C, and (C) ®laments from chicken B2 lamin as an example of an mammalian long-rod protein. The lamin ®laments (C) and the Ascaris ®laments (B) appear quite different from the vimentin ®laments (A): they are less dense, less uniform, more wavy, and more heterogeneous in width. Hence, the two long-rod Ascaris and lamin ®lament types appear more related to each other

Intermediate Filament Structure and Assembly

than the two cytoplasmic types, i.e. the Ascaris and the vimentin ®laments.

Discussion So far, all known schemes on IF architecture and assembly have been restricted to vertebrate proteins, either to the short-rod cytoplasmic or to the long-rod nuclear lamin-type IF proteins. Hence the polymerization characteristics and structural products of four protostomic invertebrate (p-INV) IF proteins have now been investigated in some detail. The four proteins used in this study occur as two pairs in the epithelium of the esophagus of the snail Helix pomatia and in the musle of the nematode Ascaris suum, respctively. Within both species, the two IF proteins appear to be expressed in approximately equal amounts. These p-INV IF proteins differ from the well-characterized shortrod (i.e. 310 residues long) cytoplasmic IF proteins by having a 355 residues long (i.e. a long-rod) central rod domain, very much like the vertebrate nuclear lamin-type IF proteins. Therefore, it is conceivable that the assembly characteristics and distinct oligomer architecture of p-INV IFs might be related either to vertebrate cytoplasmic short-rod IFs or to vertebrate long-rod nuclear lamin IFs, or alternatively, that they might reconcile features of both. A schematic comparison of a number of distinct structural and polymerization features among the three IF types is presented in Figure 10. As depicted in Figure 10, 1, under assembly conditions where small oligomers predominate, the abundance and molecular architecture of p-INV IF protein dimers appear very similar to there of vertebrate nuclear lamin dimers. In contrast, short-rod

Figure 9. Comparison of Ascaris IFs with those assembled from Xenopus vimentin and from Xenopus lamin B2 protein. Ascaris ®laments were polymerized from mixed high and low proteins in Ascaris ®lament buffer at 37 C, vimentin as given by Herrmann et al. (1996), and the lamin B2 as described by Heitlinger et al. (1991). Note that the Ascaris ®laments are closer in appearance to the lamin ®laments than to the vimentin ®laments. The specimens are un®xed, negatively stained, and all three micrographs were recorded at the same magni®cation. The scale bar represents 250 nm.

Intermediate Filament Structure and Assembly

Figure 10. Schematic comparison of the polymerization characteristics of p-INV IFs with vertebrate cytoplasmic IF proteins and nuclear lamins. The vertebrate cytoplasmic ®lament type refers to the type III IF proteins. An X indicates that a certain type of complex is not formed. For a more complete explanation of the different panels, see Discussion.

vertebrate cytoplasmic IF proteins favour tetramer formation under corresponding assembly conditions. NN-type tetramers consisting of an antiparallel, approximately half-staggered pair of dimers (Herrmann & Aebi, 1998) represent an obligatory building block both for vertebrate and invertebrate IF formation, which can be accumulated under appropriate in vitro assembly conditions (Figure 10, 2). In contrast, this distinct tetramer has never been observed in any signi®cant amount during nuclear lamin polymerization (Stuurman et al., 1998). In some instances, p-INV IF tetramers unveil their N-terminal head domains as distinct small globules at the borders of the central overlapping portion of the tetramer rod (Figure 4B and C; see also Figure 3 for head domains exhibited by dimers). However, it is not yet clear whether these globules just represent the N-terminal head domains or, in addition, parts of helix 1B of the central rod domain to which they are attached. The overall size and shape of the p-INV IF tetramer provides convincing evidence for the close to full-length antiparallel overlap of the two helix 1 coiled coils. However, their exact positioning will be deter-

613 mined only by chemical crosslinking, or by solving the atomic structure of such a tetramer. Instead of forming NN-type tetramers, under appropriate conditions nuclear lamin dimers anneal longitudinally by a short NC-type overlap of adjacent rod domains (Stuurman et al., 1998) to form long head-to-tail polymers (Figure 10, 3). Under no assembly conditions tried have such head-to-tail polymers of dimers ever been found with either vertebrate (Herrmann & Aebi, 1998) or invertebrate (this work) cytoplasmic IF proteins. Although p-INV IF proteins do not form long dimeric proto®laments, under appropriate conditions their NN-type tetramers do anneal longitudinally to form long tetrameric proto®laments, which eventually laterally associate to yield mature full-width ®laments (Figure 10, 4). Hence, longitudinal annealing of dimers (i.e. in the case of lamins) or tetramers (i.e. in the case of p-INV IF proteins) into long proto®laments followed by lateral association of these proto®laments into proto®brils and eventually full-width ®laments is evidently a hallmark shared among all long-rod IF proteins such as p-INV IF proteins and the nuclear lamins. This polymerization behaviour is distinctly different from that of vertebrate short-rod IF proteins, which ®rst form short but already full-width unit-length (65 nm long) ®laments that, in a second step, anneal longitudinally to eventually yield long mature IFs (Herrmann et al., 1996; Herrmann & Aebi, 1998). Mature, full-width ®laments from all three types of IF proteins appear to be built according to a similar molecular architecture (Figure 10, 5). It includes antiparallel alignment of distinct rod segments (i.e. helix 1 or helix 2) so that adjacent dimers come to lie approximately half-staggered to one another. Of the four types of different dimerdimer interactions identi®ed for short-rod ®laments (Stewart et al., 1989; Geisler et al., 1992; Steinert et al., 1993a,b; Heins & Aebi, 1994; Herrmann & Aebi, 1998), two of these have now been positively identi®ed in the long-rod p-INV IF ®laments: the NN-type and the CC-type, where helix 1 coiled coils or helix 2 coiled coils align in an antiparallel fashion. Whether the other two types of dimer-dimer interactions observed with shortrod ®laments (i.e. the antiparallel unstaggered lateral association, and the longitudinal annealing via a short head-to-tail overlap) occur with the p-INV IF proteins remains to be seen. The similarity of ®lament design principles is further supported by the number of polypeptides found per ®lament cross-section within the major width classes of invertebrate and vertebrate cytoplasmic IF types, which, while usually being polymorphic, is always an integer multiple of eight, for example, 24, 32 or 40 polypeptides per ®lament cross-section. Since at least in vitro the nuclear lamins do not form stable ®laments at steady state, they cannot be compared in this respect. Interestingly, contrary to vertebrate cytoplasmic IFs, where the discrete width increments

614 usually go by eight polypeptides (i.e. by octameric ``proto®brils''), in the case of p-INV IFs discrete increments of 16 polypeptides have been found, indicating that ``half-®laments'' (i.e. consisting of two octameric proto®brils or one ``sub®bril'' made of 16 polypeptides) represent a distinct, relatively stable species (Heins et al., 1993; Heins & Aebi, 1994; Herrmann & Aebi, 1998). Comparison of the general appearance of the three different types of ®laments (see Figure 9) indicates a closer relationship between the two long-rod types than between long and short-rod types. One clear difference is the length of the axial ``beading'' repeat (see Figure 10, 5), which is 24.5 nm for the two long-rod IF types but only 21.5 nm for the short-rod types (Henderson et al., 1982; Aebi et al., 1986; Heins et al., 1993). This 3 nm increment is most likely due to the 42 residue insertion in helix 1B of the long-rod proteins. Formation of thick, striated paracrystalline ®bers represents a hallmark of in vitro lamin assembly (Figure 10, 6). Evidently, neither vertebrate nor invertebrate cytoplasmic IF proteins form such paracrystalline ®bers in any signi®cant amounts under physiological or in vitro assembly conditions. Taken together, the assembly and molecular architecture of p-INV IFs share features with both vertebrate cytoplasmic IFs and the nuclear lamin assemblies. Both the snail and the nematode IF protein pairs are expressed in approximately equal amounts, thus raising the question of homo- versus heteropolymer formation. Since the two Helix proteins (i) are identical in sequence except for the missing C-terminal tail domain in the low protein and its very last C-terminal residue, (ii) recognize themselves and each other equally well on a blot overlay, and (iii) polymerize as single proteins and as mixtures without signi®cant differences in yield (see Figure 8), at least in vitro they represent facultative hetero- or homopolymers. In contrast, the Ascaris protein pair may behave differently: whereas according to the blot overlay experiment the low protein strongly binds to the high protein, it binds to itself only very weakly. In contrast, the high protein recognizes both the low protein and itself equally well. If this observation suggests a dependence of the low protein on the high protein, it would be consistent with the electron microscope-based observation that a mixture of the two proteins polymerizes much more ef®ciently than either of the two proteins by themselves. Nevertheless, the dependence of the low protein on the high protein cannot be complete, as it is still capable of forming dimers and tetramers (though in comparatively low yield), and occasionally even mature ®laments (Figure 8A, d). When assessing the in vivo signi®cance of the distinct oligomeric/polymeric subcomplexes observed under our in vitro assembly conditions, it should be kept in mind that the buffer conditions chosen to produce certain types of subcomplexes

Intermediate Filament Structure and Assembly

have not necessarily been ``physiological'' in terms of pH or ionic strength (e.g. the Helix oligomer buffer; see Materials and Methods). As might be gathered, dimers and tetramers from Helix protein, for example, would never accumulate as a major oligomeric species under physiological buffer conditions. Nevertheless, the proposed assembly scheme from dimers to tetramers, via octamers and proto®laments, to sub®laments and eventually mature ®laments (see Figure 10) is a continuous one, i.e. none of the observed oligomeric/polymeric subcomplexes is a ``dead-end'' species but rather productive intermediates. In this view, the observed subcomplexes, although sometimes accumulated under non-physiological buffer conditions, represent true assembly intermediates. Importantly, it should be noted that under no assembly conditions explored was any of the distinct subcomplexes observed produced exclusively. Rather, it accumulated as a major species together with the next higher or next lower oligomeric/polymeric species occurring along the ®lament assembly pathway. And last but not least, it should be emphasized that the elementary building block for all oligomeric/polymeric subcomplexes described is the dimer, i.e. built as an unstaggered parallel two-stranded a-helical coiled coil involving the central rod domain of two homo- or heterotypic p-INV-IF polypeptides. To establish the generality of the distinct oligomer formation and assembly behaviour of the two p-INV IF protein pairs investigated here, additional p-INV IF proteins have to be analyzed by similar approaches. Moreover, a more comprehensive structural analysis of the distinct oligomers (dimers, tetramers, octamers, etc.) identi®ed in the present study should be conducted, and the question of homo versus hetero-oligomer formation should be pursued more systematically. Last but not least, as has been investigated previously with vertebrate cytoplasmic IF proteins (Kaufmann et al., 1985; Heins et al., 1993; Herrmann et al., 1996) and nuclear lamins (Heitlinger et al., 1992; Sasse et al., 1997), the role of the head and tail domains of p-INV IF proteins in ®lament assembly, architecture, dynamics, and network formation should be dissected.

Materials and Methods Purification of p-INV IF proteins In the ®rst publications describing the sequence characterization, isolation of the four p-INV IF proteins from the animal tissues, and a preliminary characterization of their polymerization into ®laments (Weber et al., 1988, 1989; Bartnik et al., 1985, 1986; Dodemont et al., 1990), these polypeptides were given the names Helix A, Helix B, Ascaris A and Ascaris B proteins, with A being the polypeptides with the higher molecular masses in both species. For convenience, these polypeptide names have now been changed into Helix-high, Helix-low, Ascaris-high and Ascaris-low, according to their molecular masses.

615

Intermediate Filament Structure and Assembly

Helix pomatia were purchased from a snail farm. Whole guts including the esophaguses were removed and stored frozen at ÿ70 C. Puri®cation of the high and low proteins followed exactly the procedure described (Bartnik et al., 1985; Weber et al., 1988). Ascaris suum were collected in a slaughterhouse and  muscle was removed and stored at ÿ70 C. Homogenization, extraction and chromatography on DEAE-cellulose was performed as described (Bartnik et al., 1986; Weber et al., 1989). The resulting mixture of high and low IF proteins was freed from contaminating proteins by an additional gel-®ltration step on Sepharose 6B (Pharmacia Biotech, Uppsala, Sweden), before the ®nal separation of high and low proteins was achieved by ion-exchange chromatography on an anionic MonoS column (Pharmacia Biotech, Uppsala, Sweden) as described. All chromatography steps were performed using 8 M urea buffers supplemented with 1 mM 2-mercaptoethanol. Authentic chicken gizzard desmin, bacterially expressed Xenopus vimentin, and recombinant chicken lamin B2 were obtained as described (Geisler & Weber, 1980; Herrmann et al., 1996; Heitlinger et al., 1991). Protein renaturation and assembly procedures p-INV IF proteins at concentrations of 0.1 to 1.0 mg/ ml were dialyzed at room temperature or at 37 C using ¯oating nitrocellulose ®lters with 0.01 mm pore size (Millipore, Eschborn, FRG). For some experiments, dialysis was carried out in bags. Buffers with pH values ranging between 6.6 and 9.0 were used. Salt concentrations ranged between 1 mM and 500 mM NaCl (in 1 to 20 mM Tris-HCl buffers). Speci®c assembly buffers used were: for Helix proteins: Helix oligomer buffer was 20 mM Tris-HCl (pH 8.5), 250 mM NaCl, 1 mM 2-mercaptoethanol; Helix ®lament buffer was 1 mM Tris-HCl (pH 8.0), 1 mM 2-mercaptoethanol. For Ascaris proteins: Ascaris oligomer buffer was 10 mM Tris-HCl (pH 7.5), 170 mM NaCl, 1 mM 2-mercaptoethanol; Ascaris high pH ®lament buffer was 10 mM Tris-HCl (pH 8.0), 1 mM 2-mercaptoethanol; Ascaris ®lament buffer was 10 mM Tris-HCl (pH 7.0), 1 mM 2-mercaptoethanol. All assembly buffers were used at ambient temperature, except for the Ascaris ®lament buffer, which was used at 37 C. Chicken desmin, Xenopus vimentin, and chicken lamin B2 were polymerized as described by Geisler & Weber (1980), Herrmann et al. (1996) and Heitlinger et al. (1991), respectively. Blotting and overlay techniques Blotting was performed with puri®ed, unmodi®ed and biotinylated proteins. Biotinylation was achieved with the EZ-LinkTM Sulfo-NHS-Biotinylation kit from Pierce (Rockford, Il, USA) following the recommended procedure. The average relative content of biotin was approximately 7 mol/mol protein. The proteins were separated by SDS-PAGE on Laemmli gels (Laemmli, 1970), blotted onto nitrocellulose membranes and stained with Ponceau Red. The blots were next incubated with the biotinylated proteins at ambient temperature for one hour in 10 mM Tris-HCl buffer (pH 8.0) containing 0.05% (v/v) Tween 20 and 3% (w/v) non-fat dry milk powder, followed by washing according to the recommended procedures. The blots were processed using horse radish peroxidase labelled streptavidin (Amersham, Braunschweig, FRG), developed by using the ECL

RPM 2106 kit from Amersham (Braunschweig, FRG) and exposed to X-ray ®lms. Electron microscopy For negative staining, samples were applied to glowdischarged, carbon-coated copper grids, allowed to adsorb for 30 seconds, washed with water, and contrasted with 1% (w/v) uranyl acetate. For glycerol spraying/rotary metal shadowing, samples were mixed with glycerol to a ®nal concentration of 30%, sprayed onto a piece of freshly cleaved mica, and rotary shadowed at a low elevation angle (i.e. between 3 and 5 ) with platinum/carbon, essentially as described (Fowler & Aebi, 1993; HaÈner et al., 1997). Electron micrographs were recorded on either a Philips CM12 transmission electron microscope (TEM; Philips Electron Optics, Ltd, Eindhoven, The Netherlands) operated at 80 kV, or on a Hitachi H-8000 TEM (Hitachi, Ltd, Tokyo, Japan) operated at 100 kV. Electron micrographs were digitized with a Scitex LeafScan 45 ¯at-bed scanner at a step size of 2400 dpi. The digitized micrographs were processed using Adobe Photoshop software and printed with the original contrast. STEM mass measurements For the determination of the mass-per-length (MPL) of p-INV ®laments, a Vacuum Generators (East Grinstead, UK) HB-5 scanning transmission electron microscope (STEM) interfaced to a modular computer system (Tietz Video and Image Processing Systems GmbH, Gauting, Germany) was used (MuÈller et al., 1992). For this purpose, p-INV ®laments were ®xed with 0.1% (v/v) glutaraldehyde at 20 C and then adsorbed for 45 seconds to a thin carbon support ®lm, supported by a thicker, fenestrated carbon layer on a gold-coated copper grid. Without negative staining, the grid was washed four times on droplets of quartz double-distilled water before being freeze-dried in the microscope pre-treatment chamber at ÿ80 C overnight. For calibration, tobacco mosaic virus (TMV) was prepared in the same way and loaded into the microscope at the same time (Engel et al., 1985). The STEM was operated at 80 kV at a nominal magni®cation of 200,000. To minimize beaminduced mass loss, low-dose images were recorded at 200,000 with electron doses ranging between 250 and 500 eÿ/nm2. MPL values were determined according to Engel et al. (1985) using a dedicated software package, IMPSYS (MuÈller et al., 1992). Length measurements Of small oligomers were performed by computerized recording directly from tenfold projection enlarged electron micrograph negatives. The collected data were processed with the IMPSYS program (MuÈller et al., 1992), and displayed in histograms that were analyzed by multiple Gaussian curve ®tting. The length of IF ®laments taken for the determination of axial repeat lengths was measured with the Quantitative Picture Analysis apparatus MOP AM02 by Intron (Eching, Munich, FRG) using calibrated paper prints. At least ten different ®lament stretches were measured for each probe and then averaged.

616

Acknowledgements Dr Tom Roberts (Tallahassee) deserves special thanks for his generous help in providing Ascaris muscle tissue. Uwe Plessmann provided an initial preparation of Ascaris proteins, while Robert Haering and Robert Wyss offered most valuable help with the scanning of the micrographs and other aspects of preparing Figures. Dr Harald Herrmann (Heidelberg) is thanked for careful reading of the manuscript and suggestions for its improvement. The ®nancial support of the Swiss National Research Foundation, the M.E. Mueller Foundation of Switzerland and the Kanton Basel-Stadt (to M.H. and U.A.) is gratefully acknowledged.

References Aebi, U., Cohn, J., Buhle, L. & Gerace, L. (1986). The nuclear lamina is a meshwork of intermediate-type ®laments. Nature, 323, 560± 564. Bader, B. L., Magin, T. M., Hatzfeld, M. & Franke, W. W. (1986). Amino acid sequence and gene organization of cytokeratin no. 19, an exceptional tailless intermediate ®lament protein: change in protein conformation corresponds to presence of a landmark intron. EMBO J. 5, 1865± 1875. Bartnik, E., Osborn, M. & Weber, K. (1985). Intermediate ®laments in non-neuronal cells of invertebrates: isolation and biochemical characterisation of intermediate ®laments from the esophageal epithelium of the mollusc Helix pomatia. J. Cell. Biol. 101, 427± 440. Bartnik, E., Osborn, M. & Weber, K. (1986). Intermediate ®laments in muscle and epithelial cells of nematodes. J. Cell. Biol. 102, 2033± 2041. BeuttenmuÈller, M., Chen, M. Janetzko, A., KuÈhn, S. & Traub, P. (1994). Structural elements of the aminoterminal head domain of vimentin essential for intermediate ®lament formation in vivo and in vitro. Expt. Cell Res. 213, 128± 142. Bovenschulte, M., Riemer, D. & Weber, K. (1995). The structure of a cytoplasmic intermediate ®lament (IF) protein from the annelid Lumbricus terrestris emphasizes a distinctive feature of protostomic IF proteins. FEBS Letters, 360, 223± 226. Dodemont, H., Riemer, D. & Weber, K. (1990). Structure of an invertebrate gene encoding cytoplasmic intermediate ®lament (IF) proteins: implications for the origin and the diversi®cation of IF proteins. EMBO J. 9, 4083± 4094. Engel, A., Eichner, R. & Aebi, U. (1985). Polymorphism of reconstituted human epidermal keratin ®laments: determination of their mass-per-length and width by scanning transmission electron microscopy (STEM). J. Ultrastruct. Res. 90, 323± 335. Fisher, D. Z., Chaudhary, N. & Blobel, G. (1986). cDNA sequencing of nuclear lamins A and C reveals primary and secondary structural homology to intermediate ®lament proteins. Proc. Natl Acad. Sci. USA, 83, 6450± 6454. Fowler, W. & Aebi, U. (1983). Preparation of single molecules and supramolecular complexes for high resolution metal shadowing. J. Ultrastruct. Res. 83, 319± 334. Geisler, N. & Weber, K. (1980). Puri®cation of smooth muscle desmin and a protein-chemical comparison of desmins from chicken gizzard and hog stomach. Eur. J. Biochem. 111, 425± 433.

Intermediate Filament Structure and Assembly Geisler, N., Schuenemann, J. & Weber, K. (1992). Chemical crosslinking indicates a staggered and antiparallel proto®lament of desmin intermediate ®laments and characterizes one higher level complex between proto®laments. Eur. J. Biochem. 206, 841± 852. HaÈner, M., Bremer, A. & Aebi, U. (1997). Glycerol spraying/low angle rotary metal shadowing. In Cell Biology: A Laboratory Handbook (Celis, J. E., ed.), 2nd edit., vol. 3, pp. 292± 298, Academic Press, Inc., San Diego, CA. Hatzfeld, M. & Burba, M. (1994). Function of type I and type II keratin head domains: their role in dimer, tetramer and ®lament formation. J. Cell Sci. 107, 1959± 1972. Hatzfeld, M. & Weber, K. (1990). Tailless keratins assemble into regular intermediate ®laments in vitro. J. Cell Sci. 97, 317± 324. Heins, S. & Aebi, U. (1994). Making heads and tails of intermediate ®lament assembly, dynamics and networks. Curr. Opin. Cell Biol. 6, 25 ± 33. Heins, S., Wong, P. C., MuÈller, S., Goldie, K., Cleveland, D. & Aebi, U. (1993). The rod domain of NF-L determines neuro®lament architecture, whereas the end domains specify ®lament assembly and network formation. J. Cell Biol. 123, 1517± 1533. Heitlinger, E., Peter, M., HaÈner, M., Lustig, A., Aebi, U. & Nigg, E. A. (1991). Expression of chicken lamin B2 in Escherichia coli: characterization of its structure, assembly, and molecular interactions. J. Cell. Biol. 113, 485± 495. Heitlinger, E., Peter, M., Lustig, A., Nigg, E. A. & Aebi, U. (1992). The role of the head and tail domain in lamin structure and assembly: analysis of recombinant chicken lamin a and truncated B2 lamins. J. Struct. Biol. 108, 74 ± 91. Henderson, D., Geisler, N. & Weber, K. (1982). A periodic ultrastructure in intermediate ®laments. J. Mol. Biol. 155, 173± 176. Herrmann, H. & Aebi, U. (1998). Intermediate ®lament assembly: ®brillogenesis is driven by decisive dimer-dimer interactions. Curr. Opin. Struct. Biol. 8, 177± 185. Herrmann, H., HaÈner, M., Brettel, M., MuÈller, S. A., Goldie, K. N., Fedtke, B., Lustig, A., Franke, W. W. & Aebi, U. (1996). Structure and assembly properties of the intermediate ®lament protein vimentin: the role of its head, rod and tail domains. J. Mol. Biol. 264, 933± 953. Kaufmann, E., Weber, K. & Geisler, N. (1985). Intermediate ®lament forming ability of desmin derivatives lacking either amino-terminal 67 or the carboxy-terminal 27 residues. J. Mol. Biol. 185, 733± 742. Kouklis, P. D., Papamarcaki, T., Merdes, A. & Georgatos, S. D. (1991). A potential role for the COOH-terminal domain in the lateral packing of type III intermediate ®laments. J. Cell. Biol. 114, 773± 786. Laemmli, U. K. (1970). Cleavage of structural proteins during the assembly of the head of bacteriophage T4. Nature, 277, 680± 685. Lees, J. F., Schneidman, P. S., Skuntz, S. F., Carden, M. J. & Lazzarini, R. A. (1988). The structure and organization of human heavy neuro®lament subunit (NF-H) and the gene encoding it. EMBO J. 7, 1947± 1955. McKeon, F. D., Kirschner, M. W. & Caput, D. (1986). Homologies in both primary and secondary struc-

Intermediate Filament Structure and Assembly ture between nuclear envelope and intermediate ®lament proteins. Nature, 319, 463± 468. Milam, L. & Erickson, H. P. (1982). Visualization of a 21-nm axial periodicity in shadowed keratin ®laments and neuro®laments. J. Cell. Biol. 94, 592± 596. MuÈller, S. A., Goldie, K. N., BuÈrki, R., HaÈring, R. & Engel, A. (1992). Factors in¯uencing the precision of quantitative scanning electron microscopy. Ultramicroscopy, 46, 317± 334. Parry, D. A. D., Conway, J. F. & Steinert, P. M. (1986). Structural studies on lamin. Similarities and differences between lamin and intermediate-®lament proteins. Biochem. J. 238, 305± 308. Peter, M., Heitlinger, E., HaÈner, M., Aebi, U. & Nigg, E. A. (1991). Disassembly of in vitro formed lamin head-to-tail polymers by CDC2 kinase. EMBO J. 10, 1535± 1544. Riemer, D., Dodemont, H. & Weber, K. (1992). Analysis of the cDNA and gene encoding a cytoplasmic intermediate ®lament (IF) protein from the cephalochordate Branchiostoma lanceolatum; implications for the evolution of the IF protein family. Eur. J. Cell Biol. 58, 128± 135. Sasse, B., Lustig, A., Aebi, U. & Stuurman, N. (1997). In vitro assembly of drosophila lamin dm0: lamin polymerization properties are conserved. Eur. J. Biochem. 250, 30 ± 38. Steinert, P. M., Marekov, L. N. & Parry, D. A. D. (1993a). Diversity of intermediate ®lament structure: evidence that the alignment of coiled-coil molecules in vimentin is different from that in keratin intermediate ®laments. J. Biol. Chem. 268, 24916 ±24925.

617 Steinert, P. M., Marekov, L. N., Fraser, R. D. & Steinert, P. M. (1993b). Keratin intermediate ®lament structure. Crosslinking studies yield quantitative information on molecular dimensions and mechanisms of assembly. J. Mol. Biol. 230, 436±452. Stewart, M., Quinlan, R. A. & Moir, R. D. (1989). Molecular interactions in paracrystals of a fragment corresponding to the a-helical coiled coil rod portion of glial ®brillary acidic protein: evidence for an antiparallel packing of molecules and polymorphism related to intermediate ®lament structure. J. Cell Biol. 109, 225± 234. Stuurman, N., Heins, S. & Aebi, U. (1998). The nuclear lamins: their structure, assembly, and interactions. J. Struct. Biol. In the press. Traub, P. & Vorgias, C. E. (1983). Involvement of the N-terminal polypeptide of vimentin in the formation of intermediate ®laments. J. Cell. Sci. 63, 43 ± 67. Weber, K., Plessmann, U. & Ulrich, W. (1989). Cytoplasmic intermediate ®lament proteins of invertebrates are closer to nuclear lamins than are vertebrate intermediate ®lament proteins; sequence characterization of two muscle proteins of a nematode. EMBO J. 8, 3221± 3227. Weber, K., Plessmann, U., Dodemont, H. & KossmagkStephan, K. (1988). Amino acid sequences and homopolymer-forming ability of intermediate ®lament proteins from an invertebrate epithelium. EMBO J. 7, 2995± 3001.

Edited by W. Baumeister (Received 27 February 1998; received in revised form 26 May 1998; accepted 26 May 1998)