Expression in escherichia coli of human lamins A and C: influence of head and tail domains on assembly properties and paracrystal formation

Lamins are the principal components of the fibrous lamina that underlies the nuclear envelope of eukaryotic cells (reviewed by Gerace, 1986). The lamina fibres are about 10 nm in diameter and are arranged in a dense fibrous mat that sometimes shows a remarkable tetragonal order (Aebi et al. 1986; Stewart and Whytock, 1988). Besides a structural role in maintaining nuclear envelope integrity (Newport et al. 1990; Whytock et al. 1990), the lamina may influence interphase chromatin organization, DNA replication and gene expression (Lebkowski and Laemmli, 1982; Benevente and Krohne, 1986).

Lamins are found in a wide range of species. The mammalian lamin gene family has a complex pattern of expression and, for example, there appear to be at least five distinct lamins (termed A to E) in rat liver (Kaufmann, 1989). In mammals, lamins A, B and C have been more throughly characterised and cDNAs corresponding to each have been cloned (McKeon et al. 1986; Fisher et al. 1986; Hoger et al. 1988; Pollard et al. 1990). Lamins D and E have only recently been described and, although detectable in a number of other somatic tissues, are quantitatively less common than other lamins. In rats, lamin B is expressed in all tissues throughout development, with the possible exception of sperm (Kaufmann, 1989). Recently, two lamin B isoforms have been identified in a murine cell line, but the differences in their expression patterns have yet to be elucidated (Weber ét al. 1990). In man, lamins A and C are virtually identical in sequence except that lamin A has 90 additional amino acids at its C terminus and the two proteins probably arise from alternative splicing of the same gene (McKeon et al. 1986). They have have an identical pattern of expression. They are found in many differentiated tissues and their expression can be induced in non-expressing cells with mitogens (Kaufmann, 1989) or retinioc acid (Lebel et al. 1987). In chickens, a single A-type lamin and two lamin B isoforms have been cloned (Vorburger et al. 1989; Peter et al. 1989). In Xenopus laevis there are at least five different lamins with a complex pattern of expression (Stick, 1988) but there appears to be a single lamin gene in Drosophila melanogaster (Gruenbaum et al. 1988).

Lamin sequences (McKeon et al. 1986; Fisher et al. 1986; Pollard et al. 1990; Hoger et al. 1988; Weber et al. 1990; Krohne et al. 1987; Stick, 1988; Vorberger et al. 1989; Peter et al. 1989; Gruenbaum et al. 1988) show strong homologies to intermediate filament proteins and indicate a three-domain structural model for lamins in which there is a central fibrous rod, having an alpha-helical coiled-coil conformation, flanked by non-helical N- and C-terminal domains (see Steinert and Roop, 1988; or Stewart, 1990, for reviews of intermediate filament protein structure). The lamin N-terminal domain is small compared with most other intermediate filament proteins, whereas the lamin C-terminal domain is comparatively large. Electron microscopy of shadowed single molecules (Aebi et al. 1986) and of paracrystals (Aebi et al. 1986; Parry et al. 1987; Moir et al. 1990) support this model and show a rod-shaped molecule, about 52 nm long, with a prominent globular domain at one end. The sequence homology with intermediate filament proteins is strongest in the rod domain, where there is a strong heptad repeat of hydrophobic residues that is characteristic of alpha-helical coiled-coils (Crick, 1953; McLachlan and Stewart, 1975; Stewart et al. 1989a). However, when compared with intermediate filament protein sequences from all species except invertebrates, the lamins are distinctive in having a 42-residue insertion located 85–90 residues from the N terminus of the rod domain (McKeon et al. 1986; Fisher et al. 1986; Conway and Parry, 1988).

A number of important lamin functions have been localised to defined regions of the sequence. For example, Loewinger and McKeon (1988) expressed human lamin A mutants in CHO cells and identified a putative nuclear localisation signal that has similarities to that of the SV40 large T-antigen. The motif CaaX (a is an aliphatic residue and X is variable) is found at the C terminus of lamins A and B. This sequence is also at the C terminus of ras proteins and, during their maturation, the last three residues are cleaved, an isoprenyl group is added to the sulfur atom of the cysteine, and the carboxy terminus is O-methylated (Gutierrez et al. 1989; Hancock et al. 1989). This cysteine residue is essential for ras protein binding to membranes and their transforming function (Willumsen et al. 1984a,5). Alteration of the equivalent sequence in the lamins also decreases affinity of the proteins for the membranes of the nuclear envelope (Holtz et al. 1989; Krohne et al. 1989). Thus, human lamin A cDNA with altered CaaX sequences transfected into tissue culture cells (Holtz et al. 1989) and lamins injected into Xenopus laevis oocytes (Krohne et al. 1989) are not targeted correctly to the nuclear envelope. Lamin A and lamin B had previously been shown to be isoprenylated (Beck et al. 1990; Wolda and Glomset, 1988) and lamin B is methylated (Chelsky et al. 1987). Interestingly, as part of the normal maturation process, the last 18 residues of lamin A, including the modified cysteine, are removed after the protein has been inserted into the nuclear envelope (Weber et al. 1989). The function of this proteolytic event is unknown. Finally, lamin disassembly during mitosis correlates with phosphorylation of these proteins (Gerace, 1986). The serine residues probably involved have been identified using mitotic cell extracts or cdc2 kinase to phosphorylate expressed or endogenous lamins (Ward and Kirschner, 1990; Peter et al. 1990) and by transfection of mutant lamin cDNAs into CHO cells (Heald and McKeon, 1990). The serine residues, whose phosphorylation promotes disassembly, immmediately flank the rod domain.

In order to study more precisely the molecular basis for the assembly properties of lamins and to explore the interactions between lamins and other components of the nucleus (such as nuclear pores and chromatin), we have expressed the cDNAs for human lamins A and C (McKeon et al. 1985; Fisher et al. 1986) in Escherichia coli. This expression not only permits the production of large quantities of material but also enables specific mutant proteins to be constructed using site-specific mutagenesis. Although we have previously expressed human lamin C cDNA using the pLcII vector system (Moir et al. 1990), the material was obtained as a fusion with part of the bacteriophage lambda ell protein, which we were unable to remove using proteases such as factor X or thrombin because of the lability of the lamins to digestion. To circumvent these difficulties, we have instead used the bacteriophage T7 expression system (Studier et al. 1990) and describe here the production of substantial quantities of human lamins A and C, together with specific molecular fragments obtained using site-specific mutagenesis of the lamin cDNA. The expressed material closely resembled native lamins in a number of key properties and determination of the solubility properties of mutants coupled with electron-microscope examination of the aggregates they form has enabled us to assess the contribution of the different domains to the interaction between these molecules in paracrystals and filaments.

Restriction endonucleases, bacteriophage T4 DNA ligase and the Klenow fragment of DNA polymerase I were obtained from New England Biolabs (Beverly, USA) or Boehringer Mannheim (Mannheim, FRG). Bacteriophage T7 DNA polymerase (Sequenase I) was obtained from United States Biochemical (Cleveland, USA) and radioisotopes from Amersham (Amersham, UK). Oligonucleotides were synthesised by Terry Smith and Jan Fogg (MRC Laboratory of Molecular Biology, Cambridge) using an Applied Biosystems synthesiser. Human lamin A and C cDNA was a generous gift from Drs Frank McKeon and Marc Kirschner (UCSF, California, USA). Site-specific mutagenesis and other cloning methods were performed essentially as described (Quinlan et al. 1989) except that the Sequenase system (United States Biochemical, Cleveland, Ohio) was used to sequence some of the mutants.

We expressed the lamin proteins in E. coli using the T7 RNA polymerase-based pET vectors described by Studier et al. (1990). For convenience of cloning, we inserted the lamin cDNAs into the target plasmid containing the T7 promotor (pET-1) at the Ncol site, which resulted in three additional amino acid residues being added to their N terminus (Gly-Ser-Met). The T7 RNA polymerase gene is located in the bacterial chromosome under control of the lac promoter and is induced with isopropyl-/3-o-thiogalacto-side (IPTG). Coulombe and Fuchs (1990) have also used this cloning strategy for expression of cytokeratins. We found it necessary to include a second plasmid expressing lysozyme (pLysE) to inactivate the small quantities of T7 RNA polymerase resulting from the leakiness of the lac promotor, as has been described for this expression system (Studier et al. 1990). We grew the bacterial strain BS21CDE21) carrying the two plasmids to an absorbance of 0.5–0.6 at 600 nm before inducing expression with 0.4 mM IPTG, and then grew the bacteria for two to three hours prior to harvesting. The different lamin fragments were partitioned between soluble and insoluble (inclusion bodies) phases to varying extents. Protein in inclusion bodies was isolated and purified using the standard protocol of Nagai and Thogersen (1987). Bacterial cell pellets were homogenised or sonicated in 50 mM Tris-HCl, pH 8.0, 25% sucrose, lnm EDTA. Lysozyme from the pLysE plasmid was sufficient to promote lysis. Two volumes of detergent buffer (25 mM Tris-HCl, pH 7.4, ImM EDTA, 0.2 M NaCl, 1% sodium deoxycholate, 1% NP40) were added and the inclusion bodies sedimented by centrifugation (20 000g, 4°C). The crude inclusion body preparations were washed three times with 0.5% Triton X-100, 1–5 mM EDTA, to remove residual membrane-bound proteins. Phenylmethylsulphonyl fluoride (PMSF, 0.2 mM) and pepstatin A (0.2 fiM) were included in all solutions to inhibit proteolysis. The inclusion bodies were dissolved in 8 M urea, 20 mM Tris-HCl or NaHepes, pH8.0, 2mM EDTA and ImM dithiothreitol (DTP). The protein was further purified using either the Mono S (cation exchange) or Mono Q (anion exchange) preparative columns of a Pharmacia FPLC system using the same urea buffer with a 0 to 1 M NaCl gradient. In some cases, the majority of the protein was not in inclusion bodies and was solubilised by sonicating the bacterial cell pellet in 2 M urea, 20 mM Tris-HCl or Hepes, pH 8.0, 1 mM EDTA, 1 mM DTT and purified by ion exchange using the same buffer and a 0 to 1M NaCl gradient. Proteins were stored frozen in 8 M urea, 25 mM Tris-HCl, 1 mM DTT, pH 8, and, when required, were renatured by dialysis against 0.3 to 0.5 M NaCl, 25 mM Tris-HCl, 1 mM DTT, pH 8, at room temperature. Sometimes an intermediate dialysis step against 2 M urea was employed, but this was generally not necessary to obtain fully renatured protein.

SDS-polyacrylamide gel electrophoresis was carried out as described by Laemmli (1970) using 7.5% to 17.5% gradient minigels. Protein concentration was determined using the BioRad protein assay kit (based on the Bradford (1976) dye-binding assay) and by absorbance at 280 nm. The extinction coefficient for lamin A was determined to be 0.2 by refractometry using a Zeiss refractometer and we have also used this value for the other fragments (there is only one tryptophan in the lamin sequence, in the non-helical C-terminal domain). Proteins (lOOpgml^-1) were crosslinked with 11 mM dimethyl suberimidate (Pierce Chemical Company) in 20 mM NaHepes, pH 8.0, 250 mM NaCl, ImM DTP for one hour at room temperature. The reactions were stopped by adding an equal volume of 100 mM Tris-HCl, pH 8.0, 1 M glycine and incubating at room temperature for 10 min before precipitating the samples with chloroform/methanol (1:4, v/v) for gel electrophoresis.

We measured the solubility of the fragments by both dialysis and dilution into appropriate buffers. The absolute solubility at approximate physiological conditions was measured by dialysing samples in urea at about 5mgml^-1 into 20mM Tris-HCl, pH8.0, 300mM NaCl, ImM DTT, then into 20mM NaHepes, pH7.0, 100 mM NaCl, 1 mM DTT. The concentration of the soluble protein was measured by absorbance at 280 nm. The solubility as a function of ionic strength was determined by dialysing the samples first into 20mM Tris-HCl, pH8.0, 300mM NaCl, ImM DTP, then into 20 mM NaHepes, pH 7.0 or 6.5,500mM NaCl, 1 mM DTT. The concentrated samples were diluted to the different salt concentrations and protein concentrations were adjusted so that all had the same starting molarity, representing protein concentrations from 0.7 to 1.3mgml^-1. The absorbance of the soluble material was measured after 3h, after centrifuging for ten minutes at 15000 g’. Samples for electron microscopy were also prepared by dialysis or dilution, often using the 25 #x03BC;l microdialysis buttons used to prepare samples for X-ray crystallography.

Isolated particles were sprayed onto mica in 30 % glycerol, 250 mM NaCl, 20 mM Tris-HCl, 1 mM DTT, pH 8, as described by Stewart and Edwards (1984) and rotary-shadowed with platinum-carbon at a nominal angle of 10 degrees in a Cressington CFE-50 freezefracture system. Paracrystal suspensions were applied to carbon-coated 400-mesh electron microscopy grids and negatively stained with unbuffered 1% aqueous uranyl acetate as described by Stewart (1981). Both negatively stained and shadowed specimens were examined at 80 or 100 kV in Philips EM301, CM12, EM400 and EM420 electron microscopes using standard conditions. Microscope magnification was calibrated by reference to the 2.81 nm spacing of negatively stained sheaths of Methanospirillum hungatei (Stewart et al. 1985).

We produced substantial quantities of recombinant human lamins A and C by expressing cDNA in E. coli using the pET vector system (Studier et al. 1990). In addition to the full-length lamins A and C, we expressed a number of lamin fragments using site-directed mutagenesis as illustrated in Fig. 1. Mutants that lacked the N- or C-terminal non-helical domains (or both) were expressed as well as material corresponding to the mature form of lamin A from which 18 amino acids have been removed from the C terminus (Weber et al. 1990). Mutants lacking the N-terminal non-helical domain began at Glu31, three residues before the putative start of the rod domain heptad repeat, whereas mutants lacking the C-terminal domain ended at Pro393, five residues after the end of the heptad repeat, and corresponded to a deletion of the last 269 amino acids from lamin A (Fisher et al. 1986). All these expressed proteins gave positive Western blots using rat anti-human lamin antiserum (data not shown).

The levels of expression obtained with the T7 vector system varied for the different lamin constructs and with different preparations, but was usually in the range of 5 to 15 mg I^-1 of culture and was comparable to the yields we obtained using the pLcII system (Moir et al. 1990). The lowest level of expression was obtained with lamin A, but even in this case, the expressed protein was still visible in whole bacterial lysates after induction with IPTG (Fig. 2, lanes B and C). The expressed proteins were isolated from bacterial pellets either as inclusion bodies (lamin A, lamin C, lamin AΔ18, lamin ΔC) or as soluble material (lamin CΔN and lamin ΔCΔN) and, after chromatography on ionexchange columns, gave highly purified material as assessed by SDS-PAGE (Fig. 2, lanes D to I). In addition to being much easier to manipulate than the pLcH system, there was much less endogenous proteolysis in the pET system and consequently it was easier to prepare larger quantities of pure material in this way.

As illustrated in Table 1, all the expressed proteins had SDS-PAGE electrophoretic mobilities that corresponded closely to those predicted from their sequence. Shadowed preparations of dilute solutions of the different expressed proteins sprayed in glycerol onto mica showed distinctive rod-like molecular profiles consistent with the fibrous nature of lamins (Figs 3 and 4). These rods seemed generally to resemble similar shadowed preparations of native material (Aebi et al. 1986) as well as other coiled-coil proteins, such as tropomyosin and myosin rod fragments (Stewart and Edwards, 1984). The fragments that retained the C-terminal non-helical domain (lamin A, lamin C, lamin AA18, lamin CΔAN) had a prominient globular region at one end in addition to the rod profile, whereas the fragments in which this domain had been deleted have only the rod (Fig. 4). The rod domain is approximately 50 nm in length. Chemical crosslinking with dimethyl suberimidate produced material in which the apparent M_r had increased by a factor of 4 (Fig. 5), indicating that all the expressed fragments aggregated into four-chain ‘tetramer’ units analogous to those observed with other intermediate filament proteins. As is often observed when crosslinking fibrous proteins (Quinlan et al. 1989), the tetramer gave a series of closely spaced bands on SDS-PAGE. Each of these bands probably corresponded to a crosslink between different residues in the molecule and, because some secondary structure is usually retained in such fibrous proteins in the conditions used for SDS-PAGE, the differently crosslinked species migrate with slightly different mobilities even though they have the same molecular weight. Because human lamins A and C do not have a cysteine in the rod domain of the molecule, we were not able to employ disulphide crosslinking to demonstrate the presence of two-chain ‘dimer’ molecules as has been done for other intermediate filament proteins (Quinlan et al. 1989). However, in some of the fragments (particularly those in which the N-terminal non-helical domain had been deleted: lamin CAN, lamin ACAN) there was an additional crosslinked band at close to double the M_r of the single chain, which probably resulted from crosslinking within two-chain molecules (Fig. 5, arrows)

We measured the solubility of the different lamin fragments under approximately physiological conditions starting with high protein concentrations in urea (5 mg ml^-1). The different samples were dialysed intially into conditions where they are very soluble (20 mM Tris-HCl, pH8.0, 300mM NaCl, ImM DTT), then into approximate physiological ionic strength (20 mM NaHepes, pH7.0 100mM NaCl, ImM DTT). The insoluble material was removed by centrifugation and the concentration of the soluble protein measured by its absorbance at 280 nm. We used SDS-PAGE of the soluble fractions to confirm that this solubility was due to the fragments themselves and not contamination (data not shown). The results are shown in Table 1, on both a mass and a molar basis. The solubility of lamin A and lamin C was roughly similar and was slightly less than that of the lamin AA18. Deletion of either the N-terminal or C-terminal nonhelical domain had a profound effect on lamin solubility under these conditions. The lamin C fragment lacking the 30 residues at the N terminus was more than 20 times more soluble on a molar basis than full-length lamin C, whereas the fragment lacking the 179 residues of the C-terminal non-helical domain was almost 10 times more soluble than lamin A. The rod domain, lamin ΔCΔN, was the most soluble. We further investigated the solubility of the lamin fragments by measuring solubility as a function of ionic strength. For these experiments the starting protein concentration was adjusted so each sample was approximately 1.5#X00D7;10⁻⁶M and the samples were diluted to the appropriate conditions. Fig. 6 shows the results obtained at pH 7.0. The solubility of full-length lamins A and C and lamin A Δ18 mutant showed a strong dependence of solubility on ionic strength. However, the other three fragments, lacking the N or C terminus or both, did not precipitate at any salt concentration at this pH. We again confirmed these results by SDS-PAGE. Similar results were obtained at pH 6.5 (not shown).

The lamin fragments formed a range of ordered aggregates when the pH of the buffer was below 6.5 and the ionic strength below 200 mM, Fig. 7 is a low-magnification micrograph of negatively stained paracrystals formed from lamin A. Lamin A, lamin C and lamin AA18 readily formed these long, thin paracrystals without the addition of divalent cations, over a range of temperatures down to 4 °C, and at protein concentrations above 50 #x03BC;grnl^-1. Fig. 8 shows an enlargement of areas of negatively stained paracrystals formed from different lamin fragments. The axial repeat for lamin A, lamin C and lamin AΔ18 (Fig. 8) paracrystals was 21 nm, close to the value obtained for a mixture of lamins A and C isolated from rat liver nuclear envelopes (Aebi et al. 1986). The fragment that lacked the N terminus of lamin C, lamin CAN, also formed paracrystals with a similar axial repeat but they were less well ordered and always aggregated in clumps (Fig. 8). Moreover, lamin CAN paracrystals only formed when the pH of the buffer was 6 or below.

The lamin AC and lamin ΔCΔN fragments formed different aggregates from those observed with intact lamins. Lamin AC formed networks of filaments with no discernible banding pattern (Fig. 9). The diameter of the filaments varied and was usually in the range 15–30 nm. Lamin ΔCΔN, the lamin rod domain, formed very well-ordered paracrystals with a 45 nm axial repeat, in which there were two 10 nm wide light bands separated by dark bands alternately of about 12 and 13 nm (Fig. 10). In addition, the 13 nm dark band was more dense than the 12 nm band and so, although this pattern was superficially similar to that observed with intact lamins, it was distinctly different in detail.

We have expressed in E. coli cDNAs for human lamins A and C and a number of fragments derived from them. Substantial quantities of material can be produced in this way and a range of biochemical and structural criteria indicated that all these molecules had refolded correctly. Shadowing of isolated molecules (Figs 3 and 4) showed rod-like profiles about 50 nm long with two globular heads at one end for those constructs in which the C-terminal non-helical domain had not been deleted. The absence of these globular heads from constructs lacking the C-terminal non-helical domain confirms the structural assignment made from sequence data (McKeon et al. 1986; Fisher et al. 1986) of shadowed native lamin molecules (Aebi et al. 1986). Chemical crosslinking (Fig. 5) indicated that all the constructs assembled into 4-chain units, indicative of the dimérisation of two 2-chain molecules in a manner analogous to that seen with other intermediate filament proteins (reviewed by Stewart, 1990).

The formation of ordered aggregates (Figs 7–10) also indicated that molecular structure had been preserved, at least to the extent necessary for the usual molecular interactions to occur. Moreover, four of the fragments, lamin A, lamin C, lamin AΔ18 and lamin CΔN, formed paracrystals with a characteristic 21 nm axial repeat (Figs 7 and 8) that were very similar to those obtained using a mixture of lamins A and C prepared from rat liver nuclear envelopes (Aebi et al. 1986). The contrast in the staining pattern seen in the micrographs of negatively stained paracrystals probably derived from two sources: because the molecular length (50 nm) was greater than twice the axial repeat of the paracrystal, there would be a gap between successive molecules that would allow greater stain penetration and so give rise to a band of increased density across the paracrystal at this point as proposed by Moir et al. (1990). The contrast deriving from such a gap-overlap structure would be coupled with stain exclusion from the globular C terminus to accentuate the alternation between high and low stain density along the paracrystal. The paracrystals derived from the lamin rod domain (lamin ΔCΔN) fragment had a different staining pattern that may indicate an alternative set of molecular interactions (Fig. 10) when both N- and C-terminal nonhelical domains are absent. In the lamin rod paracrystals, there were still alternate dark and light bands indicating a gap-overlap structure and there was clear dyad symmetry, suggesting an antiparallel arrangement of molecules. However, the axial repeat distance had doubled. Some caution may be warranted in making a detailed comparison between the staining pattern seen with rod paracrystals and that from the other paracrystals, since the resolution in electron micrographs of the latter was rather low as a result of disorder in the aggregates, and it was conceivable that it may not have been possible to detect the difference between successive 21-nm repeats in this case. In other words, it may be that the true axial repeat of the lamin paracrystals was twice that observed but that this was obscured by strong internal pseudo-halving of the pattern. We were unable to form paracrystals with the fragment that lacked the C-terminal non-helical domain but retained the N terminus (lamin AC). This was probably a consequence of its readily forming filament networks or alternatively the presence of the large C-terminal globular domain inhibited filament formation in the other constructs. The actual filaments formed by this fragment were somewhat wider than intermediate filaments and may not necessarily represent precisely the same structure as that taken up by the lamins in the fibrous lamina.

Because in vivo lamins A and C are always expressed together and in equal amounts, it was not immediately clear if it was necessary for both to be present in a manner analogous to that observed for the two classes of cytokeratin (reviewed by Steinert and Roop, 1988; Stewart, 1990). However, either lamin A or lamin C alone formed paracrystals, implying that heterodimer or heterotetramer formation between the A and C isoforms was not required, at least for this level of assembly. Furthermore, the post-translational modifications of the lamins isolated from nuclear envelopes (such as phosphorylation, isoprenylation and proteolysis as described in the Introduction) also appear not to be an absolute requirement for this level of assembly, although they could possibly modulate the process in vivo. The mature form of lamin A, that lacked the last 18 amino acids at its C terminus, formed paracrystals identical to lamins A and C under the same conditions (Fig. 8) and had similar solubility properties, indicating that this post-translational modification was unlikely to result in a dramatic change in the interactions between this isoform and other lamins in vivo.

The influence of the N-terminal and C-terminal globular domains on lamin solubility and aggregation properties were somewhat different from that seen with other intermediate filament proteins. Results with proteolytic fragments of the intermediate filament proteins desmin, vimentin (Kaufmann et al. 1985; Traub and Vorgias, 1984) and expressed GFAP (Quinlan et al. 1989) have shown that the N terminus is required for filament formation. The role of the C-terminal domain in intermediate filament formation is less clear. GFAP fragments expressed in E. coli that completely lack this domain could not form filaments (Quinlan et al. 1989) unless extra material was attached to their N terminus. However, proteolytic fragments of desmin that lack half of the C-termina] domain still make filaments (Kaufmann et al. 1985) and cytokeratin 8 that retained only the first six amino acids of this domain also made filaments (Hatzfeld and Weber, 1990). The role of the C-terminal domain appears marginal in these intermediate filament proteins. By contrast, the solubility data that we obtained indicate that both the N- and C-terminal domains play a role in lamin assembly. Although deletion of the N-terminal domain increased lamin solubility most dramatically, deletion of the C-terminal domain still produced an almost tenfold increase (Table 1). The differences in solubility indicate that the strength or nature of the interactions at low ionic strength between these mutant molecules was different from that between the whole lamins. The appearance of the aggregates formed when the C terminus was absent (Fig. 8) confirms that this domain has an important influence on lamin assembly. It is interesting to note that, although the aggregation properties of both lamins and intermediate filaments are modulated by phosphorylation, the lamins are phosphorylated in both N and C globular domains (Ward and Kirschner, 1990; Peter et al. 1990; Heald and McKeon, 1990), whereas the intermediate filament proteins are phosphorylated only in the N-terminal domain (Inagaki et al. 1988, 1989; Geisler and Weber, 1988). The presence of additional phosphorylation sites in the lamin C-terminal domain would be consistent with the greater influence of this domain on solubility compared to cytoplasmic intermediate filaments. The aggregation of the lamins probably involves a hierarchy of interactions analogous to those proposed for intermediate filaments (Aebi et al. 1986; Stewart, 1990). All the fragments apparently formed tetramers in crosslinking assays including the rod domain (Fig. 5) so that the effects of the N and C terminus (and possibly phosphorylation) may be to modulate assembly after this point.

It is also significant that the influence of ionic strength on the solubility of whole lamins was the opposite to that observed with other intermediate filament proteins. Lamin solubility increased rapidly with increasing ionic strength (Fig. 6) whereas the solubility of GFAP (and probably other cytoplasmic intermediate filaments) decreases with increasing ionic strength (Yang and Babitch, 1988). In this respect, the solubility properties of the lamins more closely resembled myosin. This solubility behaviour further indicates that the molecular interactions involved in lamin aggregates are not precisely the same as those in other intermediate filaments, which could reflect the different nature of the lamina and cytoplasmic filament networks. The different interactions may also account for the apparent failure of lamins to incorporate into cytoplasmic assemblies whereas different cytoplasmic proteins (GFAP, vimentin, desmin) can be found in the same filament (Quinlan and Franke, 1982, 1983).

We are now using the large quantities of specific lamins produced using the pET expression to explore in greater detail the molecular interactions involved in lamin assembly and to examine the interaction of lamins with other components of the nucleus during the cell cycle.

We particularly thank Sue Whytock for invaluable assistance and advice with the figures. We thank Simon Atkinson for suggestions on protocols and for helpful discussions, and our colleagues, in particular Rob Cross, Roy Quinlan and Nigel Unwin, for their comments and criticisms. Claudio Villa provided technical assistance. R.D.M. was supported by the Alberta Heritage Foundation for Medical Research and A.D.D. holds an MRC studentship.