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

All intermediate filament (IF) proteins share a common structural organization consisting of a non-α-helical N-terminal head domain. a central α-helical rod domain and a C-tcrminal tail domain (Geisler and Weber. 1982; Geisler et al.. 1982: Hanukoglu and Fuchs, 1983: Steinert et al.. 1983. 1985: Weber and Geisler. 1985). Based on sequence homologies in the rod domain, six different types of IF protein have been classified. The keratins comprising the type I IF proteins (smaller acidic keratins) and the type II IF proteins (larger neutral to basic keratins) are the largest group of IF proteins (Fuchs et al., 1987: Hanukoglu and Fuchs. 1983; Schiller el al.. 1982: Tseng et al.. 1982). The in vitro assembly of IF seems to involve several association steps that depend on different interactions between molecular domains. Whereas some IF proteins can form homopolymers, keratins are obligatory heteropolymers containing equimolar amounts of type I plus type II keratins (Hatzfeld and Franke, 1985). The heteropolymcric nature is imposed at the level of the double-stranded coiled coil (Hatzfeld and Weber. 1990a: Coulombe and Fuchs. 1990; Steinert. 1990). Dimer assembly in vitro depends solely on the a-helical rod domain. In the next step heterodimers interact with each other to firm antiparallel tetramers that are stable under certain experimental conditions such as 4 M urea or 2 M guanidine-HC! (Gu-HCl ) (Geisler and Weber, 1982: Geisler ct al.. 1985: Quinlan el al.. 1984, 1986: Eichner et al.. 1986: Parry et al.. 1985). Tetramers are thought to be built from a staggered arrangement of the two dimer molecules. although unstaggered tetramers have also been described (Geisler el al.. I 992: Fraser ct al.. 1985: Potschka et al., 1990: Stewart ct al.. 1989: Steinert. 1991 a.b: Steinert el al.. 1993).

Stable letramers of desmin. vimcntin and keratins (Kaufmann et al., 1985: Hatzfeld et al., 1987: Sauk et al.. 1984) have been prepared using proteolytically prepared rod domains. indicating that head and tail domains are not essential for tetramer fonnation. Assembly of filaments from tetramers, however. requires some contribution of the end domains. The C-terminal tail domain does not seem to be essential for !F formation in vitro (Kaufmann et al.. 1985: Shoeman et al..1990; Hatzfeld and Weber. 1990b; Bader et al., 1991; Wilson et al., 1992; Eckelt et al., 1992) although it has a stabilizing effect on filaments (Hatzfeld and Weber, 1990b; Wilson et al., 1992). Moreover, it may interact with IF-associated proteins and therefore play a role in vivo in the context of the cell or tissue.

In contrast, the head domain plays an important role in IF assembly. Most insights into the function of the head domain stem from studies of type Ill IF proteins. Proteolytically processed desmin (Kaufmann et al.. 1985) and vimentin (Traub and Vorgias, 1983) are unable to assemble into IF. More recently, the function of the head domains has been analysed using deletion as well as point mutations. In agreement with the in vitro studies, Raats et al. (1990) found that headless desmin could be incorporated into a vimentin network in vivo but did not assemble into IF in cells that did not express any type III IF proteins. Moreover, a conserved nonapeptide motif (SSYRRIFGG) essential for assembly was identified close to the amino-terminal end of type III IF proteins (Raats et al., 1990; Hatzfeld et al.. 1992; Herrmann et al., 1992). While the head domain of vimentin binds to specific regions of the rod domain, thereby influencing filament formation, the precise acceptor site(s) and the level at which this interaction occurs have not vet been identified (Hofmann and Herrmann, 1992; Traub et ál.. 1992).

In the case of keratins the situation is more complex due to the heteropolymeric nature of these filaments. The head domains of keratins have been subdivided into three parts: the end domains (El); the variable regions (VI), which differ strongly in size and sequence even within one subclass; and a region of sequence homology (HI) close to the rod domain (Steinert and Roop, 1988; Chipev et al., 1992). The sequence conservation of H 1 is very strong in type II but not in type I keratins.

Studies on the function of the keratin head domains in filament assembly have led to seemingly conflicting results. Head-truncated K 14 was shown to be incorporated into a preexisting keratin IF network without perturbation (Albers and Fuchs, 1989). Moreover, it assembled into IF in vitro when combined with wild-type K5 (Coulombe et al., 1990; Wilson et al., 1992). Similarly Lu and Lane (1990) found that a K7 mutant containing only 20 amino acids of its head plus a 12 amino acid tag sequence assembled into filaments when coexpressed with KI8, but not with KI9. Cotransfection of head-deleted K8 and head-deleted KI8 or KI9 resulted in the formation of a disperse non-fibrillar pattern, while cotransfection of combinations of one headless plus one intact keratin resulted in the formation of fibrils or cytoplasmic granules (either AK8 plus KI8 or K8 plus AKI8; Bader et al., 1991). These experiments seem to suggest that the head domain of one keratin type is sufficient for IF assembly. Surprisingly, head-truncated K5 only assembled into IF with very low efficiency when combined with wild-type K14 or head-truncated KI4. In contrast, a truncated K5 molecule lacking only -50% of its head domain was able to form IF with intact K14 in vitro (Wilson et al., 1992).

One possible explanation for these differences could be the diversity of the keratin head domains, because their contribution might not always be equal. However, different experimental conditions, such as the location of the deletion sites as well as sequence and position of the polypeptide tags used in several of these studies, might also be responsible for the discrepancies.

To analyse the function of the head domain more precisely, we constructed a head deletion mutant of KI 8 and several head deletion mutants of K8. The shortest deletion included the El and VI region but retained the entire HI region (for nomenclature of subdomains, see Chipev et al., 1992). The longest deletion included ten amino acids of the conserved Hl motif. We show that IF formation is impaired in all assemblies involving any of the mutants, although some short and irregular IF were obtained from the mutant that retained the entire Hl region. Tetramers containing mutant keratins were noticeably destabilized, suggesting a function of the head HI sequence in the association of two dimers. Results on hybrid molecules indicate that exchanging head domains for their partner’s keratin head domain partly compensates this effect.

AK8 ( 1 -74), ΔK18 ( 1-56). ΔK18 AflII and K8 AflII were created by site-directed mutagenesis as described by Nakamaye and Eckstein (1986). ΔK8 ( 1 -66) and ΔK8 (1-64) were constructed according to the method of Kunkel (1984). All mutagenesis reactions were performed in M13mpl8 using Escherichia coli strains JMI0I and XL 1 blue. Mutants were identified by restriction analysis and the mutant sequence confirmed by DNA sequencing. After cloning into the bacterial expression vector pINDU (Magin el al., 1987; Hatzfeld and Weber, 1990a.b) all mutants were completely sequenced. To obtain the chimeric cDNAs coding for K18 (8 head) and K8 (18 head), K8 AflII and K18 AflII were digested with AflII and Hindlll, creating the K18 head fragment in Ml 3mpl8 and the K8 rod and tail fragment. These two fragments were ligated and K8 (18 head) cloned into the expression plasmid using the BamHI and Hindlll restriction sites. The K8 head fragment in MI3mpl8 was ligated to the two K18 rod and tail fragments resulting from the AflII Hindlll digestion of K18 AflII. A positive clone carrying both fragments in the correct orientation was selected and cloned into the expression plasmid using the Hamill and Sai l restriction sites. All cloning procedures were carried out as previously described.

Recombinant wild-type keratins K8 and KI 8. as well as K8 AflII, KI 8 AflI I and the chimeric keratins K8 (18 head) and KI 8 (8 head), were purified (see Hatzfeld and Weber 1990a,b. 1991). Since chimeric keratins were very prone to proteolysis, protein inhibitors (PMSF, E64, trypsin inhibitor. EDTA; see Hatzfeld and Weber 1990a,b) were added to all buffers in all steps. All deletion mutants were purified from inclusion body preparations by Mono Q anion exchange chromatography in 8.5 M urea solution using the same conditions as for wild-type keratins followed by reverse-phase chromatography on a C4 column (Bio-Rad. Richmond CA, USA; Hatzfeld el al., 1987).

Protein concentrations were determined according to the method of Bradford ( 1976) or by amino acid analysis.

Equimolar amounts of type I and type II keratins, or keratin mutants, were combined in 8.5 M urea, 50 mM Tris-HCl, pH 8, 5 mM EDTA buffer at a total protein concentration of ∼250 μg/ml. Samples were dialysed overnight against standard filament assembly buffer (50 mM Tris-HCl, pH 7.5, 5 mM EDTA) or low salt buffers (30 mM Tris-HCI, pH 7.5, 5 mM EDTA or 10 mM Tris-HCl. pH 7.5. 5 mM EDTA) using microdialysis filters (Millipore GmbH. Eschborn, FRG) (see Hatzfeld and Weber, 1990b, 1991). Structures formed during dialysis were analysed after negative staining or rotary shadowing (Henderson el al., 1982; Hatzfeld and Franke, 1985; Hatzfeld and Weber, 1990a,b).

For analysis of tetramers, samples were dialysed against 2 M Gu-HCI, 50 mM Tris-HCI, pH 8.0, 5 mM EDTA.

Analytical gel filtration chromatography was performed using the SMART micro FPLC facility (Pharmacia, Uppsala. Sweden) and a G200 gel filtration column (3.5 mm×3 10 mm. Pharmacia) equilibrated in 2 M Gu-HCl, 30 mM Tris-HCI, pH 7.5. Protein oligomers were separated at a flow rate of 40 μl/minute at room temperature. Elution profiles were monitored at 230 and/or 280 nm. Peak fractions were collected, the protein was precipitated with methanol/chloroform according to the method of Wessel and FlUgge (1984) and analysed on 10% SDS-polyacrylamide gels (Laemmli, 1970).

Sedimentation equilibrium runs were perfonned on a Beckman XL-A analytical ultracentrifuge using the An-Ti60 rotor (Beckman, Palo Alto, CA) at 20°C with an A 230 of 0.2 to 0.3 and 12.000 rpm for peak fractions 1 and 2 and 14.500 rpm for peak 3 of the gel filtration chromatography. Peak fractions obtained after gel filtration chromatography were analysed immediately after collection. The standard analysis software of the manufacturer was used for molecular mass determination. A partial specific volume of 0.73 ml/g and a density of l g/m1 were assumed in all cases.

To analyse the influence of the head domain on keratin filament assembly we constructed several deletion mutants of K8 and one deletion mutant of K18. Prior to selecting the truncation sites, we compared the sequences of different type II keratins in the region close lo the rod domain (Fig. 1). Whereas the end domains E1 and the variable region V1 of the head domains vary strongly in size and sequence, even within one subclass of keratins, the region directly in front of the rod domain (HI) is highly conserved in type 11 but not in type 1 keratins (Steinert and Roop, 1988; Chipev et al., 1992). We also compared the extent of the head deletions introduced into different keratins (Fig. 1) in previous studies (Lu and Lane, 1990; Bader ct al., 1991; Wilson et al., 1992). Analysis of the sequences and the deletions reported so far prompted us to assume that an intact H1 domain might be required and perhaps even be sufficient for IF formation. To test this hypothesis, we constructed three different mutants of the type 11 keratin K8 (see Fig. 1): (1) t.K8 ( 1-64) retains the entire conserved H1 region plus an additional four amino acids (MRGS) encoded by the vector; (2) ΔK8 ( 166) lacks only two amino acids of this homology region. Its starting point corresponds exactly to the t.K7 mutant of Lu and Lane (1990) except that it lacks the 17 amino acid tag, which these authors added at the amino terminus. (3) ΔK8 (1-74) lacks 10 amino acids (∼25%) including three invariant residues of the Hl region. We also constructed a K 18 deletion mutant, ΔK18 ( 1-56). which lacked∼75% of its head domain (for sequences see Fig. 1).

All deletion mutants were cloned into the prokaryotic-expression vector pINDU (Magin et al.. 1987). The recombinant proteins were purified from inclusion body preparations by anion-exchange chromatography in urea solution. 1n some cases this was followed by reverse-phase chromatography using a C4 column (see Materials and Methods). Fig. 2 illustrates the purity of the proteins used.

The filament-forming capacity of the truncated keratins was analysed by dialysis against standard keratin filament buffer (50 mM Tris-HCI, pH 7.5) followed by negative staining and/or rotary shadowing. Whereas wild-type K8 plus K18 assembled into regular IF with only a little protofilamentous material remaining under these conditions, the combination of the two mutant keratins αK8 (1-74) and t.Kl8 (1-56) yielded only soluble oligomers that appeared as rod-like molecules of ∼50 nm in length after rotary shadowing (Fig. 3A and B). This suggests that either the K8, the K18, or both, head domains are essential for IF formation. We then tested the assembly characteristics of the deletion mutants in combination with their wild-type partner keratins: K8 plus αK18 (1-56) yielded predominantly soluble oligomers (Fig. 4A). Rotary-shadowed samples revealed rod-like molecules of ∼75-80 nm and -50 nm length, probably representing tetramers and/or dimers (Fig. 4B). In some places longitudinally associated rod-like particles were detected (arrow in Fig. 4B). These structures represent only a minor portion of the sample. AK8 (1-74) plus K18 also formed soluble oligomers in filament assembly buffer (Fig. 4C). Rotary shadowing showed exclusively rod molecules, most of which were 75-80 nm long (Fig. 4D). In contrast, ΔK8 ( 1 -66) plus K 18 as well as AK8 (1-64) plus K18 formed some short and irregular filamentous structures (Fig. 4E-H) but the major portion of the protein oligomers stayed soluble. Rotary shadowing revealed rather short and irregular filament fragments with the typical 21 nm repeat pattern characteristic of IF (Henderson el al., 1982). In addition, annular structures were formed that resembled those assembled from wild-type keratins in low salt buffer (Franke el al., 1982). This suggests that although the capacity for elongation and lateral association was not completely lost as in the AK8 (1-74) mutant, regular filament formation was impaired in ΔK8 (1-64) and ΔK8 ( 1-66). Elevated salt concentrations as well as variations in pH did not noticeably improve filament formation or assembly of higher-order oligomers (insets in Fig. 4F and H).

We used gel filtration chromatography to characterize the soluble oligomers obtained by the mutant keratins in filament buffer. Fig. 5 shows that AK8 ( 1-74) plus K18 eluted as a major peak at ∼22 minutes followed by two minor peaks at ∼28 minutes and ∼58 minutes, respectively. Analytical ultracentrifugation studies of the peak fractions indicated that the peak eluting at ∼22 minutes contained predominantly tetramers (data not shown; for evaluation of gel filtration data by analytical ultracentrifugation; see next paragraph). Thus the polymerization of the ΔK8 ( 1 -74) mutant was arrested at the level of tetramers. The ΔK8 (1-64) mutant plus KI8 contained a large pool of soluble oligomers together with some filamentous structures. These were also analysed by gel filtration chromatography after centrifugation at 13,000 g for 10 minutes (Fig. 5). The soluble protein fraction eluted as a major peak at ∼22 minutes, whereas the ∼28 minutes peak was not present. This suggests that the soluble pool of AK8 (1-64) and KI 8 in filament buffer consists mainly of tetramers. The AK18 (1-56) mutant and K8 formed soluble oligomers in filament buffer that eluted in a major peak at ∼22 minutes and a minor peak at ∼28 minutes, respectively (Fig. 5).

Tetramers are soluble precursors of IF that are stable under certain experimental conditions such as 2 M Gu-HCl or 4-6 M urea (Geisler el al., 1982; Quinlan el al., 1984; Hatzfeld and Franke, 1985; Coulombe and Fuchs, 1990). To characterize the soluble complexes formed by the deletion mutants, samples containing the mutant keratin plus its wild-type partner were dialysed against 2 M Gu-HCI buffer and the structures formed visualized after glycerol spraying and rotary shadowing. AK8 ( I -64) plus K18 formed some short rodlike molecules, most of which were ∼45-50 nm long (Fig. 6A). ΔK8 (I-66) formed rod molecules with an average length of 75-80 nm and 50 nm in combination with K18 (Fig. 6B). ΔK8 ( l-74) plus K18 did not form any regular structures that could be detected at the EM level (Fig. 6C). ΔK 18 ( I -56) plus K8 also assembled into short rod molecules that were comparable to those obtained from ΔK8 (I-66) plus KI 8 (Fig. 6D). Gel filtration chromatography in 2 M Gu-HCI allowed the separation of several different oligomeric species (Fig. 7). Wild-type K8 plus K 18 eluted as a single peak at ∼22 minutes, which contained ∼98% of the total protein. In contrast, AK8 (I-64) plus Kl8 was a mixture of two different oligomers eluting at ∼22 minutes (∼88%) and ∼28 minutes (∼ll%). respectively. The AK8 (I-66) mutant plus Kl8 formed again several oligomeric species eluting at ∼22 minutes (40%), ∼28 minutes (49%) and ∼50 minutes (∼11 %). AK8 (1-74) plus K18 formed very small amounts of oligomers eluting at ∼22 minutes and ∼28 minutes (∼13% both), most of the sample eluted at -50 minutes (-82%). The elution profile of the AK1 8 (1-56) mutant plus K8 reveals peaks at ∼22 minutes and -28 minutes (70% and 28%).

To characterize the oligomeric species separated by gel filtration chromatography, we first tested whether the oligomeric species eluting in a peak was stable or whether it became rearranged with time to a mixture of the different species. Peak fractions of the ΔK8 (1-64) plus K18 sample were rechromatographed after v24 hour incubation at room temperature: -80% of the protein eluted in the same position as before, indicating that the rearrangement of oligomers was a rather slow process. Analytical ultracentrifugation studies were performed to determine the molecular masses of the protein complexes eluting after ∼22 minutes, ∼28 minutes and -50 minutes. Sedimentation equilibrium analysis of peak 1 containing wild-type K8 plus KI8 in 2 M Gu-HCI (see Fig. 7) indicated that the complex was a tetramer of molecular mass 187,000 (Fig. 8A, calculated mass: 204,200). Peak 2 (∼28 minutes) of AK8 (1-64) plus K18 had a molecular mass of 89,000 (Fig. 8B) corresponding to dimers (the calculated molecular mass is 95,200). The major protein peak of AK8 ( 1 -74) and K18 eluting at -50 minutes (see Fig. 7) gave a molecular mass of 48,000 in the sedimentation equilibrium analysis, indicating that a major portion of this sample had failed to form oligomeric structures. Taken together these results indicate that tetramer formation was diminished under standard conditions in all deletion mutants and dimer formation was reduced in the AK8 (1-74) mutant plus K 18.

We also investigated whether type I and type II head domains complement each other in IF formation as observed with rod domains or whether IF formation can take place with either four type 1 or four type II head domains per tetramer. We therefore introduced a unique restriction site, AflII, into the Hl subdomain in the a-helical region that directly precedes the coiled-coil rod domain of K8 and KI8. This allowed us to exchange the head domains between the two keratins. The position of the AflII site was chosen in such a way that no invariant amino acid of the Hl subdomain was mutated in K8. Fig. 9 shows the amino acid sequences of the point mutants K8 AflII and KI8 AflII as well as the hybrid molecules K8 (18 head) and K18 (8 head) around the mutation sites. All mutants were expressed in E. coli and purified using anion exchange chromatography plus single-stranded DNA chromatography as described (Hatzfeld and Weber, 1990a,b). Since both hybrid molecules were very prone to proteolysis at their head domains, inclusion body preparation and anion exchange chromatography were performed in the presence of protease inhibitors. The degradation products were separated from intact protein by ssDNA affinity chromatography. Purified proteins appeared as single bands on SDS-PAGE (Fig. 10, lanes 1-4). N-terminal amino acid sequencing confirmed that these pools contained the intact proteins.

The two point mutants K8 AflII and K18 AflIIcarrying two amino acid exchanges were able to assemble into normal IF when used together (Fig. 11) as well as when used with their wild-type partner keratins (not shown).

To test the ability of the hybrid molecules to assemble into IF. we combined them with their non-hybrid partner keratins. K18 AflII plus K8 ( 18 head) assembled into short aggregated filament structures under standard assembly conditions (Fig. 12A). Assemblies at lower salt concentrations (30 mM Tris-HCI and 10 mM Tris-HCI: Fig. 12B and C) yielded long, almost normal filaments. However. the efficiency of IF formation was lower than with wild-type keratins and much of the protein stayed soluble (see the protofilamentous material in the background of Fig. 12B and C). K8 AflIIplus K 18 (8 head) assembled in standard assembly buffer into short irregular filament fragments with an unusually large diameter (up to 25 nm) (Fig. 12D) and into electron-dense aggregates (not shown). At lower salt concentrations (30 mM Tris-HCl and 10 mM Tris-HCI; Fig. 12E and F) again short and irregular filament fragments were observed.

Thus. replacement of two head domains per tetramer impaired but did not completely inhibit IF formation. Tetramers containing four head domains of KI8 seemed to be slightly more efficient in assembly than tetramers containing four head domains of K8.

To test whether the effect of a ‘foreign’ head domain can be compensated when both keratins carry their partner’s keratin head domains we also analysed the filament-forming capacity of K8 ( 18 head) plus K18 (8 head). These proteins assembled into relatively short and irregular IF as well as into many aggregates. Again. lower salt concentrations were used to reduce aggregation (10 mM Tris-HCI and 30 mM Tris-HCI) (Fig. 12G-K). Rotary shadowing and negative staining revealed that filaments were irregular (see filaments at high magnification in Fig. 12I), had branch points (Fig. 12K). kinks and an uneven surface. Their diameter varied from 6 to 15 nm. Moreover. much protofilamentous material was seen in the background (see arrowheads in Fig. 12I). These experiments indicated that some interactions were impaired in the assembly of hybrid keratins although polymerization was not totally blocked.

The soluble complexes of the point mutants as well as the hybrid molecules were analysed in 2 M Gu-HCl. Negative stain and rotary shadowing showed rod-like molecules in all samples (K8 Aft II plus Kl8 A//II/K8 A//II plus Kl8 (8 head), K18 A/ZII plus K8 ( 18 head) and K8 ( 18 head) plus KI 8 (head) (not shown)). Gel filtration chromatography showed that all protein combinations contained dimers and tetramers in a ratio of-1:2 (Fig. 1,3). These experiments indicate that the hybrid molecules as well as the point mutants form tetramers, which are however less stable in 2 M Gu-HCl than their wild-type counterparts.

The precise function of the head domain in IF assembly seems to vary amongst members of this multigene family. Although head domains are required for regular IF formation, the exact sequences or sequence motifs, their location within the head domain and the stoichiometric amounts required seem to be different for different IF proteins. A direct analysis of the precise function of the head domain is complicated by the fact that it probably acts at several distinct levels of IF formation, involving longitudinal as well as lateral association.

To analyse the function of the head domain in IF assembly, deletion mutants were constructed and their IF forming ability was compared in in vitro assembly studies (Wilson et al.. 1992) and in in vivo transfection experiments (Lu and Lane, 1990; Bader et al.. 1991; Albers and Fuchs, 1989; Raats ct al., 1990). Most studies so far have concentrated on whether or not IF can be produced. Here, we have examined the role of the head domain in the formation of different oligomeric structures, i.e. in dimer and tetramer fonnation and in IF assembly. This was achieved by in vitro reconstitution experiments using conditions that arrest assembly at the level of filament subunits. Biochemical analysis of these subunits allowed us to determine at which level the deletion mutants interfered with assembly. These results are summarized in Table 1. Surprisingly, the head domain seems to play an essential role not only in filament elongation, but also in stabilizing the heterotetramers. Comparison of several head deletion mutants of keratin 8 described here, as well as of some others described in the literature (Lu and Lane, 1990; Wilson ct al.. 1992), suggests an important role for the HI subdomain.

Most headless mutants formed a mixture of dimers and tetramers with almost no residual monomers under the standard conditions used to isolate the K8/K 18 heterodimer (see Fig. 7). Thus, amino-terminal deletions that leave the Hl domain intact do not influence coiled coil dimer formation. In contrast, the llK8 ( 1-74) mutant plus K 18 stayed predominantly monomeric . after overnight dialysis against 2 M Gu-HC1, indicating that dimer formation is affected in this mutant. We therefore conclude that the Hl domain may influence the alignment of the two polypeptide chains. We propose that the H1 domains interact stably with each other. thereby facilitating coiled coil dimer formation. If the Hl domains were the first regions to interact during dimer fonnation they would orient the a-helical domains in a manner allowing coiled coil formation. The two chains would be fixed in a parallel non-staggered orientation that facilitates hydrophobic interactions between the two a-helices, which would then stabilize the dimer molecules. This model is supported by the finding that tetramers of llK8 ( 1-74) plus K 18 obtained in filament buffer dissociate predominantly into dimers when dialysed against 2 M Gu-HC1. Hence. dimers seem to be stable species independent of the presence of Hl, whereas tetramers are destabilized if the Hl region is not present.

Whereas wild-type K8 plus K 18 formed close to 1 00% tetramers in buffers containing 2 M Gu-HCI. assemblies involving head deletion mutants contained mixtures of dimers and tetramers, indicating that tetramer formation was less efficient (for a similar observation on the desmin rod. see Potschka cl al., 1990). The ratio of dimers versus tetramers correlated with the length of the deletion. We therefore conclude that the H1 region helps to stabilize the K8/K 18 heterotetramers. This subdomain might direct the association of two dimers into the staggered tetramer. This could be accomplished by binding the HI region to a specific site in the coil 2a/L2 region. Staggered tetramers are the only soluble precursors that have so far been experimentally verified as stable intermediates (Steinert, 1991a; Steinert et al., 1993). Other arrangements including the antiparallel non-staggered tetramers have been found in hexameric and octameric precursors of IF (Steinert, 1991a,b). Although all head-truncated keratins seemed capable of forming tetramers in filament buffer, some of these tetramers may be ‘misaligned’ dimers that are polymerization incompetent, e.g. tetramers with overlapping coil 2 regions that are present as nearest neighbours in the surface lattice of IF, but generally not detected in a soluble form (Steinert et al., 1993) and which might be polymerization incompetent.

Earlier studies of keratin tetramers in solution showed that proteolytically prepared rod domains still formed tetramers (Hatzfeld et al., 1987; for desmin see Geisler and Weber, 1982; Potschka et al., 1990). These results are compatible with our results, since the proteolytically excised rod domain retains the entire Hl region (for the sequence of the K8/K 18 proteolytic rod, see Hatzfeld et al., 1987).

IF formation may involve a nucleation process followed by an elongation reaction. Nucleation is supposed to be slow compared to elongation (Steinert et al., 1993; Stewart, 1993). Since nucleation seeds have not been isolated in vitro. the nature of these molecules is still unknown, although it has been suggested that the nucleation seed contains four to eight IF molecules (Steinert, 1991b). Annular structures at the EM level have been observed when keratin assembly was arrested at the protofilament level in low salt buffer (Franke et al., 1982). Using the rotary shadowing technique, annular structures can sometimes be detected at the ends of filaments. However, it is not clear whether these structures represent the nucleation seeds from which filaments grow.

All headless mutants interfered with regular IF assembly, but did so at different levels. Thus, 6K8 (1-74) plus K18 assembled only into tetramers. There are several possible explanations for this effect: (a) hexameric or octameric nucleation seeds could be a prerequisite for IF formation. If this were true the ΔK8 ( 1 -74) deletion mutant would block assembly due to its inability to form a nucleation seed. (b) Dimers containing the ΔK8 ( 1-74) mutant could associate into misaligned tetrameric subunits that are polymerization incompetent (see above). (c) Tetramers formed in filament buffer might function as nucleation seeds and polymerization competent subunits. In this case. IF assembly would be blocked due to an important role for the head domain in assembly of filaments from its subunits. The ΔK8 (l-66) and ΔK8 (l-64) mutants plus Kl8 formed some short fibrillar structures. Rotary shadowing revealed short IF with the typical ∼20 nm repeat and annular structures at some ends. Thus, mutants retaining,the HI region are nucleation competent. Since short IF form in the presence of a large pool of soluble subunits (predominantly tetramers, see Fig. 5) the elongation rate seems to be slowed. Hence, the El and VI regions of the head domain seem to be essential for regular elongation.

Our findings that the Hl region is essential for IF formation explains some major discrepancies described in the literature concerning the characteristics of headless keratins. The ΔK7 mutant of Lu and Lane (1990) was assembly competent. This mutant retained almost the entire Hl region (it corresponds directly to our ΔK8 ( I -66) mutant) and carried in addition a 17 amino acid tag at its amino terminus. The assembly characteristics of this mutant resemble those of our ΔK8 (I-66) and ΔK8 (l-64) mutants. Filament elongation occurred in vivo (Lu and Lane. 1990) as well as in vitro (this study). In contrast, the assembly characteristics of ΔK8 (I-74) plus Kl8 rather resemble those of ΔK5 plus K!4 of Wilson et al. (1992). Both mutants lack their H I regions and severely affect assembly in vitro, although the ΔK5 mutant of Wilson et al. (1992) does not completely block assembly at the tetramer stage. Sequence differences among the different keratin pairs causing different stabilities of the pairs (Franke et al., 1983; Hatzfeld and Franke, 1985) might be responsible for these minor variations.

Our observations, indicating that the Hl domain might play a key role during IF formation, are supported by the findings of Steinert and colleagues. They showed that a synthetic 36-residue peptide containing this region disassembles keratin IF (Steinert and Parry, 1983; Chipev et al.. 1992). This destructive effect of the HI peptide was ameliorated by a peptide representing the L2 region, suggesting that the L2 region might be a binding partner for Hl (Steinert and Parry. 1993). Extensive cross-linking studies performed to identify the nearest neighbours in keratin filaments showed that the HI region lies close to the coil 2a/L2 region (Steinert et al., 1993). This is compatible with the arrangement of the two coiled-coil molecules in an antiparallel staggered tetramer, where the coil I domains overlap (for a similar arrangement in the desmin tetramer, see Geisler et al., 1992). Moreover, the coil 2B region was identified as an additional neighbour of HI, indicating that this region might be involved in specific interactions during lateral and longitudinal association of subunits (Steinert ct al.. 1993). The importance of the Hl region in filament assembly is also demonstrated by point mutations in the Hl region of keratin 1. These mutants interfered with regular IF assembly in vivo and thereby caused a genetic skin disease (Chipev et al.. 1992; Yang et al., 1994).

The defect in assembly of amino-terminally truncated keratins can be partially overcome by the head domain of the partner keratins. In contrast to the large soluble pool observed with the head deletion mutants, aggregation and polymorphism of filaments is the phenotype observed with the hybrid keratins. Modulation of the nucleation complex could cause this phenotype. Alternatively, lateral association was affected, giving rise to filaments with variable diameter and branch points. Head domains that do not fit perfectly might bind less specifically and thereby cause formation of polymorphic forms and aggregates. These findings suggest that the head domains influence the solubility characteristics of keratin pairs. This view is supported by the observation that various peptides from the head domains can cause aggregation (Steinert and Parry, 1993; Hofmann and Herrman, 1992). Although deletion analysis reflects the importance of the Hl region of type II keratins, tetramers containing four type I head domains (K8 (18 head) plus KI8) seemed to be slightly more efficient in assembly than those containing four type II head domains (K8 plus KI8 (8 head)). This suggests that type I head domains at least in some keratins fulfil a similar function, although the sequence responsible for an equivalent interaction varies among the members of this IF class.

We thank Klaus Weber for generous support and suggestions. We also thank Heinz-Jurgen Dehne for expert technical assistance, Hans-Peter Geithe for synthesis of oligonucleotides and Uwe PleBmann for N-terminal amino acid sequencing. Gillian Paterson is thanked for typing the manuscript. This work was supported by a grant from the Deutsche Forschungsgemeinschaft.