Physical interactions between lamins and emerin were investigated by co-immunoprecipitation of in vitro translated proteins. Emerin interacted with in vitro translated lamins A, B1 and C in co-immunprecipitation reactions. Competition reactions revealed a clear preference for interactions between emerin and lamin C. Structural associations between lamins and emerin were investigated in four human cell lines displaying abnormal expression and/or localisation of lamins A and C. In each cell line absence of lamins A and C from the nuclear envelope (NE) was correlated with mis-localisation of endogenous and exogenous emerin to the ER. In two cell lines that did not express lamin A but did express lamin C, lamin C as well as emerin was mis-localised. When GFP-lamin A was expressed in SW13 cells (which normally express only very low levels of endogenous lamin A and mis-localise endogenous emerin and lamin C), all three proteins became associated with the NE. When GFP-lamin C was expressed in SW13 cells neither the endogenous nor the exogenous lamin C was localised to the NE and emerin remained in the ER. Finally, lamins A and C were selectively eliminated from the NE of HeLa cells using a dominant negative mutant of lamin B1. Elimination of these lamins from the lamina led to the accumulation of emerin as aggregates within the ER. Our data suggest that lamin A is essential for anchorage of emerin to the inner nuclear membrane and of lamin C to the lamina.
The nuclear envelope (NE) forms the boundary between the nucleus and the cytoplasm and physically separates transcription and translation. The nuclear lamina, together with the inner (INM) and outer nuclear membranes (ONM) and the nuclear pore complex are the major components of the NE. The nuclear lamina is a flattened orthogonal network of filaments lying subjacent to the INM. The filaments of the lamina are composed of type V intermediate filament proteins termed lamins. Like all intermediate filament proteins, lamins possess a long central α-helical rod domain flanked by a short globular N-terminal head and a long C-terminal tail domain. Lamins apparently assemble into filaments through a series of hierarchical associations. These include the initial formation of dimers through parallel in register coiled-coil associations through the rod domain, followed by anti-parallel half-staggered lateral associations of dimers to form tetramers. Tetramers then make head-to-tail longitudinal associations to form protofilaments (see Goldberg et al., 1999EF20; Stuurman et al., 1998EF49).
The composition of the lamina varies according to cell type and stage of differentiation. Two broad classes of lamins are expressed in vertebrates and these are referred to as A-type and B-type. These lamins vary in primary sequence and biochemical properties as well as in their expression patterns. B-type lamins are expressed in all germ cells and somatic cells although different B-type lamins are expressed in each (reviewed by Vaughan et al., 2000EF51). Lamins B1 and B2 are the major B-type lamins expressed in mammalian somatic cells and these proteins are the products of separate genes (reviewed by Gant and Wilson, 1997EF17). Lamins A, C, C2 and A▵10 comprise the A-type lamins and all are alternatively spliced products of a single gene. Lamin A and C are the most abundant A-type lamins and differ in that lamin C lacks a 90 amino acid C-terminal extension possessed by lamin A but has five unique amino acids at its C-terminus (reviewed by Quinlan et al., 1995EF44). Both lamin A and lamin C are expressed only in differentiated cells and during mouse development appear at the time of organogenesis (Rober et al., 1989EF46); however, they are dispensable for development since a lamin A/C knockout mouse survives to adulthood (Sullivan et al., 1999EF50).
Recently, a number of different autosomal dominant diseases have been shown to be caused by mutations in the gene encoding lamins A and C (reviewed by Flier, 2000). These diseases include an autosomal dominant form of Emery-Dreifuss muscular dystrophy (AD EDMD) (Bonne et al., 1999EF2; Raffaele et al., 2000EF45), dilated cardiomyopathy with conduction system disease (Fatkin et al., 1999EF16) and a Dunnigan-type familial partial lipodystrophy (Shackleton et al., 2000EF48; Cao and Hegele, 2000EF6). It is currently unclear why these very different diseases arise through mutations in the same proteins. However, since a majority of disease phenotypes are caused by missense mutations that occur in different parts of the proteins (Flier, 2000; Raffaele et al., 2000EF45), one possibility is that these lamins interact with a number of different nuclear proteins and that different mutations affect different lamin interactions.
Emerin was first identified by positional cloning of a gene on chromosome Xq28 that is mutated in individuals with X-EDMD. The emerin gene encodes a 254 amino acid type II integral membrane protein (Bione et al., 1994EF1). Structural analysis predicts that emerin contains a transmembrane region at the C-terminus and a large hydrophilic N-terminal domain with multiple putative phosphorylation sites (Bione et al., 1994EF1). In addition, emerin contains the LEM domain signature common to a number of integral membrane proteins of the inner nuclear membrane (reviewed by Hutchison et al., 2001EF25). Emerin is a serine rich protein that migrates as a 34 kDa band on SDS-PAGE. It is principally located at the INM in almost every tissue (Nagano et al., 1996EF37: Manilal et al., 1996EF28). In cardiac muscle, emerin has also been located at intercalated disks (Cartegni et al., 1997EF7), although this finding was not substantiated by later studies (Manilal et al., 1999EF31). In skeletal muscle cells grown in culture, a fraction of emerin is located in the endoplasmic reticulum (ER) (Fairley et al., 1999EF15). The majority of lesions in the emerin gene that cause EDMD are null mutations (Nagano et al., 1996EF37, Manilal et al., 1997EF29; Manilal et al., 1998EF30; Mora et al., 1997EF36; Ellis et al., 1998EF14; Yates et al., 1999EF55). However, some mutations result in the production of modified forms of emerin (Manilal et al., 1997EF29; Mora et al., 1997EF36; Ellis et al., 1998EF14; Wulff et al., 1997EF54; Yates et al., 1999EF55). These mutations occur throughout the protein with no obvious hot-spots. Interestingly, some mutations in the N-terminal `nucleoplasmic' domain cause mis-localisation of emerin either to cytoplasmic membranes or to the nucleoplasm (Fairley et al., 1999EF15; Ellis et al., 1998EF14). This finding is consistent with the observation that sequences within the N-terminal nucleoplasmic domain are necessary and sufficient to target emerin to the INM (Östlund et al., 1999EF38).
A possible link between X-EDMD and AD EDMD is that A-type lamins form structural associations with emerin at the INM. To test this hypothesis we have investigated laminemerin interactions in vivo and in vitro. Our data suggest that the organisation of both lamin C in the lamina and of emerin at the INM is dependent upon lamin A.
MATERIALS AND METHODS
Monoclonal antibodies against lamins A/C (JoL2 and JoL5), lamin A (JoL4) and lamin B2 have been described previously (Dyer et al., 1997EF11). In addition, monoclonal antibodies against emerin (MANEM3 and MANEM5; Manilal et al., 1996EF28) and LAP2β (LAP17; Dechat et al., 1998EF10) have been described previously. The polyclonal rabbit antibody against lamin C was raised to the last eight amino acids of lamin C, including an N-terminal lysine as a linker (KHHVSGSRR). The peptide was coupled to keyhole limpet haemocyanin through primary amino groups using glutaraldehyde. The resulting protein-peptide conjugate was dialysed overnight at 4°C against PBS. The conjugate was then used to immunise a rabbit. Immune serum was screened by indirect immunofluorescence and then affinity purified against 10 mg of the lamin C peptide conjugated to CH Sepharose 4B as described (Harlow and Lane, 1988). The antibody recognises the lamin C-specific tail domain of recombinant lamin C, but no other part of recombinant lamin C or any other recombinant lamins on western blots. The antibody detects a single band migrating at 65 kDa in western blots of nuclei isolated from HeLa, human dermal fibroblasts and SW13 cells. In the majority of human cell lines and primary human fibroblasts the antibody stains the NE exclusively (for a complete characterisation of this antibody, see Venables et al., 2001). Polyclonal anti-emerin was from J. Ellis (Ellis et al., 1998EF14). The anti-lamin B1 specific antibody was purchased from Santa Cruz Biotechnology. Monoclonal anti-Ki67 was purchased from DAKO.
Expression and immunoprecipitation of recombinant lamins, NUP-153 and emerin
Plasmid pWW1 hLamin B1 was constructed by subcloning an Nco1-BamH1 fragment of human lamin B1 cDNA (Pollard et al., 1990) into Nco1-BamH1 cut T7 expression vector pWW1 (Whitfield et al., 1990). Cloning of cDNA encoding the first 188 amino acids of emerin from total human skeletal muscle cDNA into the expression plasmid pMW172, and full length emerin cDNA into plasmid pET17xb have already been described (Manilal et al., 1996; Manilal et al., 1999). Plasmids pET-1 hLamin A and pET-1 hLamin C are described elsewhere (Moir et al., 1990; Moir et al., 1991). Nup-153 cDNA (gift from Brian Burke, University of Calgary, Canada) was cloned into pRSET-A using XhoI and PvuII. Recombinant lamins A, C, B1, Nup-153, and the N-terminal 188 amino acids of emerin were expressed using the TNTr Quick Coupled Transcription/Translation System (Promega) under recommended conditions. Reticulolysates containing expressed lamins and emerin were pooled and incubated at 4°C overnight. Pefablocr, leupeptin, pepstatin and aprotinin (Boehringer Mannheim) were added to a final concentration of 1 mM each. Reaction mixtures were pre-cleared using paramagnetic Dynabeadsr M-280 (DYNAL). Lamins A and C were recovered using mAbs Jol2, JoL4 or Jol5 (or the lamin A-specific Jol4) conjugated to Dynabeadsr according to a method previously described (Jenkins et al., 1993). Likewise, emerin was immunoprecipitated using mAbs MANEM3 and -5 (Manilal et al., 1996), and NUP-153 using mAb 414.
Cell culture and preparation of SW13/20
Human cervix carcinoma (HeLa) and Human adrenal cortex carcinoma (SW13) cells (gift from H. Herrmann, Heidelberg) were routinely maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum (FCS, v/v) and 1% penicillin/streptomycin (v/v) at 37°C in a humidified atmosphere containing 5% CO2. Lymphoblastoid, lymphoma and lung carcinoma cells were maintained in RPMI 1640 (GIBCO BRL) containing non-essential amino acids, 15% FCS (v/v) and 1% penicillin/streptomycin. SW13 cells were transfected with pCDNA-EGFP-HLA, a plasmid containing human lamin A fused to EGFP (gift from L. Karnitz, Mayo Clinic), using Lipofectin™ (GIBCO BRL), 15 μg of DNA per 2.5×105 cells and conditions recommended for Lipofectin transfections. Stable clones were selected using 300 μg/ml G418 (Calbiochem).
Cells were scraped from a 75 cm2 flask using a rubber policeman then washed twice with phosphate-buffered saline (PBS) at 4°C. Cell lysis occurred after incubation in cytoskeletal buffer (10 mM PIPES, pH 6.8, 300 mM sucrose, 100 mM NaCl, 3 mM MgCl2 and 1 mM EGTA) containing 0.5% Triton X-100 at 4°C for 5 minutes. Chromatin was removed by digestion with 200 units/ml RNase-free DNase in digestion buffer (10 mM PIPES, pH 6.8, 300 mM sucrose, 50 mM NaCl, 3 mM MgCl2, 1 mM EGTA, 5 units/ml RNase inhibitor (Boehringer Mannheim) containing 0.5% Triton X-100 at 30°C for 30 minutes. Nuclei were washed with extraction buffer (10 mM PIPES, pH 6.8, 250 mM ammonium sulphate, 300 mM sucrose, 3 mM MgCl2, and 1 mM EGTA) twice for 5 minutes each time at 4°C. Nuclear matrix fractions were solubilised in 8 M urea. Nuclei were pelleted after each step by centrifugation at 1000 rpm for 5 minutes within a pre-chilled microcentrifuge. Pefablocr (2 mM), leupeptin (10 μM), pepstatin (1 μM), and aprotinin (0.5 μM) were included throughout.
Construction and expression of GST/GFP fusion proteins
The construction of plasmid pGEX-XLaminB1Δ2+ is described elsewhere (Ellis et al., 1997EF13). This construct expresses amino acid residues 34-420 from Xenopus lamin B1 as a GST fusion protein. GST-XLaminB1Δ2+ fusion protein was expressed then purified from Escherichia coli BL21 cells using a method described previously (Ellis et al., 1997EF13). The pEGFP-XLaminB1Δ2+ plasmid was constructed by subcloning the SalI-NotI insert of pGEX-XLaminB1Δ2+ into pEGFP-C1 (Clontech) cut with SalI and Bsp120L. Plasmid pEGFP-emerin was constructed by sub-cloning full length emerin from pET17xb into pEGFP-C1 (Clontech) cut with BamHI and HindIII. Plasmids pS65T-lamC and pS65T-lamA have been described previously (Broers et al., 1997EF5). HeLa and SW13 cells grown on coverslips to 20% confluence were transfected with relevant plasmids (1-2 μg/coverslip) using the calcium-phosphate method (Graham and van der Eb, 1973EF19). The cells were grown overnight and media replaced, with expression allowed to proceed for a further 36-48 hours.
Lymphoblast/lymphoma cells were centrifuged onto coverslips at 300 rpm for 3 minutes prior to fixation. Cells grown on coverslips were washed twice in PBS and fixed in ice-cold 1:1 (v/v) methanol-acetone for 10 minutes. Nuclear matrices were prepared as described elsewhere (Dyer et al., 1997EF11). Coverslips were washed 3 times, with 0.5% newborn calf serum (NCS, v/v) in PBS. Hybridoma cell culture supernatants (undiluted) containing mAbs Jol2, Jol4 and LN43 (gift from Birgit Lane, Dundee) were used to stain nuclear lamins A/C, A and B2 respectively. Likewise, supernatants of mAbs MANEM3 and -5 were used to immunodetect emerin, and LAP17 (gift from Roland Foisner, Vienna) to detect LAP 2β. Goat anti-lamin B (Santa Cruz Biotechnology) and rabbit anti-ER (gift from Daniel Louvard, Paris) and anti-calreticulin (Calbiochem) were used at recommended dilutions to stain lamin B1 and ER, respectively. Affinity purified rabbit anti-lamin C was used at a dilution of 1/50, as described previously (Venables et al., 2000EF52). Primary antibodies were added for 1 hour at room temperature. Coverslips were washed three times with PBS, then incubated with appropriate Rhodamine (TRITC)-conjugated affinipure secondary antibodies (donkey anti-goat, goat anti-mouse, and goat anti-rabbit; Jackson Immunoresearch) for a further hour at room temperature. After several washes in PBS, coverslips were mounted face down in Mowiol (Calbiochem) containing 1 μg/ml DAPI. Immunostained samples were viewed using a Zeiss axiovert 10 microscope with a plan-APOCHROMAT 63×/1.40-oil immersion lens and equipped with a Digital Pixel Instruments 12-bit CCD camera. Images were captured using IP Lab Scientific Imaging Software (Scanalytics). Additionally, a Zeiss LSM 410 confocal laser scanning microscope was used (63×/1.40 oil immersion lens) for imaging of emerin and ER within SW13 cells.
Gel electrophoresis and immunoblotting
Recombinant proteins and nuclear matrix fractions were resolved on 10% SDS-PAGE and transferred to nitrocellulose according to established protocols (Jenkins et al., 1993EF26). Nitrocellulose membranes were washed with blocking buffer (5% milk powder (w/v), 0.1% Tween-20 in PBS) for 1 hour at room temperature. Undiluted cell culture supernatants containing mAbs Jol2, Jol4 and LN43 were used to detect lamins A/C, A and B2, respectively. Goat anti-lamin B1 (Santa Cruz Biotechnology) was used at 1/200. Affinity purified rabbit anti-lamin C was used at a dilution of 1/100. All primary antibodies were incubated with membranes for 30 minutes at room temperature. Membranes were rinsed with blocking buffer several times then incubated with appropriate HRP-conjugated secondary antibodies (rabbit anti-mouse, DAKO; goat anti-rabbit, BIO-RAD; donkey anti-goat, Jackson Immunoresearch) for 30 minutes at room temperature. ECL reagents (Amersham Life Science) were used for the immunological detection of proteins after membranes were rinsed in PBS.
Lamin-emerin interactions in vitro
We initially investigated interactions between emerin and lamins using co-immunoprecipitation assays in rabbit reticulocyte lysates. Since interactions between emerin and lamins have already been demonstrated (Fairley et al., 1999) we established these assays as competition experiments in order to determine the specificity of emerin-lamin complexes and the relative strengths of emerin interactions with specific lamin subtypes. Initially 35S-met-labelled lamins A, B1 and C, emerin or the nuclear pore protein Nup153 were produced in rabbit reticulocyte lysates. 35S-emerin and 35S-Nup153 were then mixed, together with equal amounts of 35S-lamin A, 35S-lamin B1 or 35S-lamin C (Fig. 1A). We have previously shown that in vitro translated Nup153 interacts with lamins but not emerin. Therefore Nup153 was used here as a control. Emerin was then immunoprecipitated from each mixed lysate using an equal mixture of the anti-emerin mAbs MANEM3 and MANEM5. The immunoprecipitates were resolved on SDS-PAGE and fluorography was performed in order to detect co-immunoprecipitation of the other proteins. All three lamins co-immunoprecipitated efficiently with emerin. However, although Nup153 was present at the same concentration in each mixed lysate, it was co-immunoprecipitated with emerin either at very low levels or not at all (Fig. 1B). Thus we confirmed that interactions between emerin and each lamin subtype occurred. Morevoer, where lamins are capable of forming alternative complexes (e.g. with Nup153), these were not present in the MANEM3/5 immunoprecipitates. This may either be because emerin-lamin interactions out-compete lamin-Nup153 interactions or because alternative complexes are excluded from the immunoprecipitates. However, the absence of Nup153 from the complex indicates that emerin forms a strong and exclusive interaction with each lamin subtype in vitro. To investigate whether emerin displayed a preference for any of the lamins, competition experiments were established between individual lamin polypeptides. In these experiments, 35S-emerin was mixed with equal amounts of 35S-lamins A and C, 35S-lamins A and B1 or 35S-lamins C and B1. Again emerin was immunoprecipitated with MANEM3 and MANEM5 and resolved on SDS-PAGE. The starting mixtures and the immunoprecipitates are shown in Fig. 1C and D, respectively. When emerin was mixed with lamins A and C, significantly more lamin C compared with lamin A was recovered in MANEM immunoprecipitates. When emerin was mixed with lamins A and B1 similar amounts of each lamin was recovered in MANEM immunoprecipitates (see Fig. 1E. Finally, when emerin was mixed with lamins B1 and C significantly more lamin C was recovered than lamin B1. As a negative control, each lamin and emerin was translated in isolation and immunoprecipitated with MANEM3 and -5 in combination. As expected, only emerin was found in immunoprecipitates (Fig. 1F). Since emerin appeared to display a preference for binding to lamin C, densitometry was performed on fluorographs and the intensity of each lamin band was compared with emerin using UVIband software (UVItech Ltd). When lamin C was present in immunoprecipitation reactions, it was recovered at approximately 1:1 ratio with emerin (Table 1). The ratio of lamin B to lamin C was ∼1:3 (Fig. 1D; Table 1) when these two lamins were co-immunoprecipitated with emerin. The ratio of lamin A to lamin C was ∼1:2 (Fig. 1D; Table 1) when these two lamins were mixed with emerin. Finally, the ratio of lamin A to lamin B was ∼2:3 (Fig. 1D; Table 1) when these two lamins were mixed with emerin. These data suggest that, although emerin is able to interact with all three lamins in vitro, its preferred interaction is with lamin C.
Abnormal targeting of emerin in a cell line SW13 with altered expression and organisation of lamins A and C
To investigate lamin-emerin interactions in vivo we compared the distribution of emerin and the lamin B binding protein LAP2β in two cell lines with very different lamin complements. The expression of the different lamins in HeLa cells was investigated by immunoblotting with specific antibodies against lamins A/C, A, C, B1 and B2. All four lamins were expressed at high levels (Fig. 2A). A similar investigation was performed on the adrenal cortex carcinoma cell line SW13 (Paulin-Levasseur et al., 1989EF40). In this cell line lamin A was expressed at greatly reduced levels (compared with HeLa) and was undetectable with some antibodies (JoL4). By contrast, lamins B1 and B2 were expressed at similar levels in SW13 and HeLa, whereas lamin C was expressed at reduced but readily detectable levels (Fig. 2B). Next, the distribution of lamins A and C were compared in HeLa and SW13 cells by immunofluorescence. Using anti-lamin-C-specific anti-sera, lamin C was detected predominantly in the nuclear rim in HeLa cells (Fig. 2C). In SW13 cells lamin C was detected at low levels in the nuclear rim but at high levels in the nucleolus (Fig. 2D). (Note that the nucleolar distribution of lamin C was confirmed by co-staining with Ki67 antibody and in all subsequent experiments co-staining with Ki67 was used to confirm nucleolar localisation of lamin C in the absence of lamin A; data not shown.) Using the lamin-A-specific mAb JoL4, lamin A was detected in nuclear speckles and at the nuclear rim in HeLa cells but was undetectable in SW13 cells (Fig. 2C,D). Lamins B1 and B2 were both localised exclusively at the nuclear rim in HeLa and SW13 cells (data not shown). These data suggested that, in SW13 cells, lamin A was expressed at very low levels (compared with HeLa), whereas a significant fraction of lamin C was mis-localised to the nucleolus.
Next we compared the distribution of LAP2β and emerin in HeLa and SW13 cells. In each cell line, LAP2β was located at the NE (Fig. 2E,F). In HeLa, emerin was detected as a distinct nuclear rim stain indicating its location at the NE (Fig. 2E). By contrast, in SW13 cells, emerin was located both at the NE and within the cytoplasm. To investigate the site of emerin localisation within the cytoplasm, SW13 cells were co-stained with MANEM3 and antibodies against calreticulin. As a control, SW13 cells were co-stained with MANEM3 and the mitochondrial marker p32 (Matthews and Russell, 1998EF32). Confocal microscopy revealed that the cytoplasmic fraction of emerin co-localised exclusively with calreticulin (Fig. 3) but did not co-localise with p32 (data not shown). These data suggest that in SW13 cells a significant fraction of emerin resides in the ER.
To further investigate the behaviour of emerin in HeLa and SW13 cells, cDNA encoding full-length human emerin was sub-cloned into pEGFP (pEGFP-emerin) and expressed in each cell line following transient transfection. As expected, in HeLa cells GFP-emerin was localised exclusively in the NE (Fig. 4E). When the same construct was transfected into SW13 cells the majority of GFP-emerin was localised in the cytoplasm, where it accumulated in large granular structures (Fig. 4A-D). Co-staining with calreticulin revealed that although the granular structures were distinct from the majority of the ER they did contain ER proteins (Fig. 4D). Transfected cells were also co-stained with specific antibodies against lamins C, B1 or B2. Surprisingly, a readily detectable fraction of lamin C (Fig. 4A), but no lamin B1 (Fig. 4B) or B2 (Fig. 4C), relocated from the nucleus to co-distribute with GFP-emerin in some of the cytoplasmic granules.
A number of conclusions can be drawn from these data. First, the absence of lamin A from the NE correlates with a significant fraction of lamin C being mis-localised to the nucleolus and a fraction of emerin residing in the ER. Second, when GFP-emerin is overexpressed in SW13, it forms cytoplasmic aggregates, probably within an ER sub-domain that traps some lamin C but no lamins B1 or B2. Taken together, these data suggest that emerin interacts with lamins A and/or C in vivo.
Emerin is localised in the ER in a range of cell lines that display abnormal levels of expression and distributions of lamins A and C
Birkitt's lymphoma cell lines
A comparison of emerin and lamin C localisation in HeLa and SW13 cell lines indicated that lamin A might organise both proteins at the NE. To investigate the generality of this phenomenon we compared the distribution of emerin and LAP2β in other human cell lines that are deficient for lamin A/C expression or which display altered distributions of lamin C. A Burkitt's lymphoma cell line (Ramos) did not express lamin A and expressed reduced levels of lamin C. Moreover, in immunofluorescence experiments lamin C was localised exclusively in the nucleolus. The levels of expression and distribution of lamins B1 and B2 appeared normal (Table 2). LAP2β was localised exclusively at the NE in Ramos cells. By contrast, emerin was distributed exclusively in the ER (Table 2).
Small cell lung carcinomas
Previous investigations have shown that small cell lung carcinomas express greatly reduced levels of lamins A and C compared with non-small-cell lung carcinomas (Kaufman et al., 1991EF27). Therefore, we compared the distribution of LAP2β and emerin in non-small-cell and small cell lung carcinoma cell lines. The non-small-cell lung carcinoma line used in this investigation (NCl-H125) expressed high levels of lamins A, C, B1 and B2 that were distributed mainly at the NE (Table 2). In this cell line, LAP2β and emerin were both located exclusively at the NE. In common with most examples (Broers et al., 1993EF4), the small cell lung carcinoma cell line used in the study (NL-SCSC2) did not express lamin A or lamin C but expressed high levels of lamins B1 and B2. LAP2β was localised at the NE in the small cell lung carcinoma, whereas emerin was localised in the ER (Table 2).
EDMD cell lines
Finally, we compared emerin localisation in EBV-transformed lymphoblastoid cell lines (LCL) obtained from a control donor and from a patient with autosomal dominant EDMD (AD EDMD). The control LCL expressed relatively low levels of lamins A and C but both were located in the nuclear rim (Table 2; Fig. 5A). Lamin B1 and B2 expression and distribution (Table 2) in this cell line appeared normal. Although some emerin was located in the ER (presumably as a consequence of the relatively low-level expression of lamins A/C), the majority was located at the NE (Fig. 5A). The AD EDMD cell line was obtained from a patient having a missense mutation in the lamin A/C tail (T528K). In this patient expression of lamins A and C was variable, with some cells expressing apparently normal levels of the proteins at the NE, whereas in other cells both proteins were absent (Table 2; Fig. 5B). Again, lamins B1 and B2 appeared normal both in terms of level of expression and distribution (Table 2). The variable level of A-type lamin expression in the AD EDMD patient permitted a side-by-side comparison of emerin distribution in those cells expressing lamins A/C and in those that do not. Cells in which lamins A and C were expressed and localised at the NE (lamin C is shown in Fig. 5B) also contained significant quantities of emerin at the NE. By contrast, in adjacent cells that did not express lamins A and C, emerin was located exclusively in the ER/NE (Fig. 5B; the ER localisation was confirmed by costaining with anti-calreticulin (not shown)). Therefore, in a range of cell lines, localisation of emerin at the NE correlates with expression and localisation of lamins A and C at the NE. In addition, lamin C localisation at the NE may also depend upon lamin A.
Stable and transient expression of GFP-lamin A in SW13 cells causes the relocalisation of lamin C and emerin to the NE
Our immunofluorescence data have revealed a strong correlation between abnormal expression and distribution of lamins A and C and localisation of emerin in the ER. To investigate whether this represented a causal relationship we carried out transient and stable transfection experiments on SW13 with GFP-lamins. Initially, we selected a number of cell lines that had been stably transfected with GFP-lamin A. One such cell line (SW13/20) is shown here, since it is representative. Stable transfection with GFP-lamin A in SW13 resulted in levels of expression of the fusion protein that were approximately fourfold higher than levels of expression of lamin C (not shown). Importantly, all A-type lamins (including endogenous lamin C) were located predominantly in the NE rather than in the nucleolus (Fig. 6A). In SW13/20 cells both LAP2β (Fig. 6C) and emerin (Fig. 6B) were localised exclusively at the NE. Next we performed transient transfection experiments with GFP-lamins A, B1 and C. Typically in these experiments 15% of cells expressed the GFP-fusion protein. When SW13 was transfected with GFP-lamin A, the GFP-fusion protein was localised at the NE (Fig. 7A,B). Importantly, the endogenous lamin C became predominantly co-localised with the GFP rather than in the nucleolus (Fig. 7B). Emerin was also localised at the NE in cells expressing GFP-lamin A but not in surrounding untransfected cells (Fig. 7A, arrowhead; the arrow identifies a mitotic cell lying adjacent to the cell expressing GFP-lamin A; the cytoplasmic emerin observed next to the transfected cell is within the mitotic cell). GFP-lamin B1 was also localised to the NE in transfected SW13 cells but failed to cause the relocalisation of endogenous lamin C or emerin to the NE (data not shown). Finally, when SW13 was transfected with GFP-lamin C, the fusion protein accumulated in nucleoplasmic granules (Fig. 7C,D; note that when GFP-lamin C is transiently expressed in HeLa cells it localises exclusively to the NE (data not shown)). These granules did not influence the distribution of endogenous lamins (endogenous lamin C remained in the nucleolus; Fig. 7C) and emerin remained in the cytoplasm (Fig. 7D). As with previous experiments the nucleolar distribution of lamin C and the ER localisation of emerin was confirmed by co-localisation with Ki67 and calreticulin, respectively (data not shown). These data strongly support the view that localisation of emerin to the NE depends upon the presence of lamins A and C within the lamina. Furthermore, the data also suggest that the presence of lamin A within the lamina is necessary (but possibly not sufficient) for lamin C localisation to the NE.
Dominant negative mutants of lamin B1 selectively eliminate lamins A and C from the lamina and cause emerin to accumulate in cytoplasmic granules
To further investigate the relationship between emerin localisation at the NE and the presence of lamins A and C in the lamina, we used dominant negative mutants of lamin B1 to specifically disrupt A-type lamins. We have previously described the creation of a dominant negative mutant of lamin B1 that is capable of disrupting the lamina of sperm pronuclei assembled in vitro (Ellis et al., 1997EF13). This mutant protein (delta 2+) was fused to GFP and expressed in transient transfection assays in HeLa cells. Following transfection, GFP-delta 2+ accumulated as small nucleoplasmic granules that formed over 48 hours (Fig. 8A-D). Both lamin A (Fig. 8A) and lamin C (Fig. 8B) relocated from the nuclear lamina to the nucleoplasmic granules over the same period of time. By contrast, lamins B1 (Fig. 8C) and B2 (Fig. 8D) remained in the nuclear lamina. To confirm that lamins B1 and B2 remained in the nuclear lamina, transfected cells were extracted in situ with detergents, nucleases and ammonium sulphate. Following this procedure both B-type lamins were retained in the insoluble lamina, demonstrating that their solubility properties were unaffected by the presence of the mutant protein (data not shown). Thus in these transfection experiments, GFP-delta 2+ exerted a dominant effect over A-type lamins, causing their redistribution from the lamina to nucleoplasmic granules, but had seemingly no effect on B-type lamins. Next we investigated the effects of the dominant negative mutants on emerin and LAP2β distribution. When HeLa cells were transfected with GFP-delta 2+ and then stained with anti-emerin antibodies, the majority of emerin was located in cytoplasmic granules rather than the NE (Fig. 9A). By contrast, LAP2β remained in the NE (Fig. 9B).
The cytoplasmic granules observed in transfected HeLa were unlike the ER distribution of emerin observed in lamin A/C-deficient cell lines. Therefore, to investigate the location of the cytoplasmic emerin, transfected cells were stained with anti-calreticulin antibodies (TRITC) and anti-emerin antibodies (Cy5). A typical result is shown (Fig. 10) and reveals that emerin did not co-localise with the majority of calreticulin in transfected cells. Instead. emerin was mainly located in granules lying close to the NE. However, calreticulin did accumulate within these granules (Fig. 10, arrowheads) suggesting that the granules were within the ER.
Lamin C is the preferred in vitro binding partner of emerin
The data presented here demonstrate that lamins A, C and B1 all interact with emerin in vitro. However, in competition reactions the preferred interaction of in vitro translated emerin is with lamin C. This finding, although novel, is consistent with the results of previous investigations. In a recent study, Östlund et al. (1999EF38) reported that the nucleoplasmic domain of emerin is both necessary and sufficient for targeting integral membrane proteins of the ER to the INM. We have shown that this domain of emerin binds to lamins following translation of both proteins in rabbit reticulocyte lysates. Emerin also co-immunoprecipitates from C2C12 myoblast lysates in a complex containing lamins A, C and B1, as well as actin (Fairley et al., 1999EF15). It is not clear which protein (or proteins) within this complex emerin was associating with, since A-type lamins form hetero-oligomeric complexes with B-type lamins in cell lysates (Dyer et al., 1999EF12) and actin also binds to lamin A in vitro (Sasseville et al., 1998EF47). Therefore, it is possible that, in cell lysates, the complex containing emerin, lamin B1, lamin C and actin is formed through association of all three proteins with lamin A. Direct interactions between the nucleoplasmic domain of emerin and lamin A have been detected using the BIAcore biosensor (Clements et al., 2000EF8). Based on the data presented here and previously published evidence we conclude that emerin probably associates with residues conserved between A-type and B-type lamins, possibly as a multimer (Clements et al., 2000EF8), but that its preferred association is with lamin C.
Evidence for a hierarchy of lamina associations mediated by lamin A
We investigated emerin localisation in two different human cell lines that were deficient for synthesis of lamins A and C and two further cell lines that were deficient for lamin A synthesis and in which lamin C was mis-localised to the nucleolus. In each cell line either all or a majority of emerin was mis-localised to the ER. In addition, we used a dominant negative mutant of lamin B1 that selectively eliminates lamins A and C but not lamins B1 and B2 from the NE of HeLa cells. A consequence of eliminating lamins A and C from the NE was that emerin relocated from the NE to the ER, where it formed insoluble inclusions.
Our data suggest that lamin A has a central role in tethering both emerin and lamin C to the NE. The following evidence supports this model. Association of lamin C with the NE in two of the cell lines reported here is dependent upon the presence of lamin A, and in its absence lamin C organisation is disrupted such that most (SW 13) or all (Ramos) is mis-localised to the nucleolus. In SW 13 cells the failure of lamin C to be incorporated into the lamina is a direct consequence of the absence of lamin A since transient or stable expression GFP-lamin A causes significant re-localisation of endogenous lamin C to the NE. The failure of lamin C to be incorporated into the lamina in the absence of lamin A may result from weak associations between lamin C and lamin B filaments or because lamin C is not isoprenylated and carboxy methylated and cannot accumulate at the NE on its own. Thus lamin A may be required to carry lamin C to the NE or to mediate its association with lamina filaments or both. It is unclear why, in the absence of lamin A, lamin C becomes mis-localised to the nucleolus. However, it is possible that the nucleolus is a default location for some proteins that might otherwise form damaging structures (e.g. aggregates) when they are unable to accumulate at their normal sites of assembly.
Evidence that lamin C incorporation into the lamina is dependent on the presence of lamin A has been reported previously. When fluorescently labelled lamin C was microinjected into Swiss 3T3 cells it forms small aggregates in the nucleoplasm which persist for several hours (an analogous situation was observed in this study when GFP-lamin C was transiently expressed in SW 13 cells). When lamin A and lamin C were injected together into Swiss 3T3 cells they are both incorporated into the NE rapidly (Pugh et al., 1997EF43). In agreement with these findings, when lamin C was transfected into cells arrested in S-phase it remained in the nucleoplasm, whereas lamin A was incorporated into the NE under similar conditions. However, if the transfected cells were released from S-phase and permitted to divide, presumably allowing transfected lamin C to interact with soluble lamin A, the lamin C became incorporated into the NE during the following G1 phase (Horton et al., 1992EF24).
We observed that emerin is mis-localised to the ER in four different human cell lines that display abnormal expression or localisation of lamins A and C. Moreover, expression of GFP-lamin A in one of these lines resulted in relocation of emerin from the ER to the NE. In all cell lines employed in the study, lamins B1 and B2 were expressed and localised normally and LAP2β was localised at the NE. Thus, even though emerin can bind to lamin B1 in vitro, the presence of this protein in the lamina is not sufficient to anchor emerin at the INM. Instead these data suggest that the presence of lamins A and C at the NE is necessary for emerin localisation at the INM. The data reported here is entirely consistent with the recent description of the lamin A/C-knockout mouse. In lamin A/C-/- mice, emerin is located mainly in the ER in most tissues (Sullivan et al., 1999). Consistent with this data, when we used dominant negative mutants of lamin B1 to selectively eliminate lamins A and C from the NE in HeLa cells, emerin relocated from the INM to inclusions within the ER.
Our data can be explained by a hierarchical series of associations between lamin A, lamin C and emerin. In this hierarchy, lamin A may mediate the association of lamin C with the lamina and, once present, lamin C may stabilise the association of emerin with the INM. In previous studies associations between lamins A/C and emerin have been reported but the possibility that lamin A and lamin C perform different functions at the INM were not considered. For example, a clear temporal correlation between emerin and lamin A/C association with the reforming NE at telophase has been reported (Manilal et al., 1999EF31). Similar temporal correlations were reported in live GFP-imaging studies (Haraguchi et al., 2000EF21). In addition, one study reported a striking spatial correlation in which lamins A/C and emerin co-associate with the reforming NE in discrete foci at telophase, whereas lamin B1 and LAP2β re-associates throughout the NE (Dabauvalle et al., 1999). This spatial correlation between lamin A/C and emerin incorporation at the reforming NE is all the more striking because A-type lamins do not return to the nucleus until after the formation of a transport-competent envelope, whereas emerin-containing membranes associate with chromatin much earlier. Although all three studies indicated a structural complex involving emerin, lamin A and lamin C at the INM, none of the studies was able to distinguish between different roles for lamin A and lamin C in the complex. Five observations led us to present a new hypothesis to explain lamin-emerin associations at the INM. (1) emerin's preferred in vitro binding partner is lamin C. (2) Emerin is never present in the NE in the absence of lamin C. (3) Lamin C is not incorporated into the lamina in the absence of lamin A. (4) Lamin C can be sequestered into emerin aggregates in the ER when GFP-emerin is overexpressed in lamin A-deficient cells. (5) We have demonstrated previously that lamin A is incorporated into the lamina through association with lamin B filaments (Dyer et al., 1999EF12). Therefore, we propose that the hierarchy of lamina associations involving emerin is as follows: lamin A associates with lamin C as either dimers or tetramers in the nucleoplasm. Lamin A/C then associates with B-type lamina filaments, this association being mediated by lamin A. The association of lamin A with the lamina allows incorporation of lamin C but in the absence of lamin A, lamin C accumulates at default sites in the nucleus. Once incorporated into the lamina, lamin C associates with emerin at the INM. The association between emerin and lamin C may stabilise and tether both proteins at the lamina. Thus in the absence of lamin C, emerin is not stably associated with the lamina and we speculate that, in the absence of emerin, lamin C may not be stably associated with the INM. This model provides an explanation for the accumulation of lamin C in cytoplasmic aggregates when GFP-emerin was overexpressed in SW 13 cells. Presumably, in this instance, emerin aggregates were a preferred location compared with the default location in nucleoli.
Our data suggest that emerin can bind to lamin B1 in vitro. However, we found no evidence to suggest that B-type lamins influences emerin behaviour in vivo. B-type lamins have a number of specific binding partners at the NE, including LAP2β and LBR (reviewed by Vaughan et al., 2000EF51). It is possible that potential emerin binding sites on B-type lamins, organised as lamina filaments, are unavailable because they are occupied by other INM proteins with higher binding affinities. Thus, although emerin can associate with B-type lamins, it does not do so in living cells.
Disease causing mutations in the lamin A/C gene and their relationship to lamina structure
Since the report early last year that mutations in the gene encoding lamins A and C cause AD-EDMD, it has become clear that a large range of lamin A/C mutations give rise to a number of diseases (Bonne et al., 1999; Brodsky et al., 2000; Cao et al., 2000; Fatkin et al., 1999; Raffaele et al., 2000). The mutations so far identified map to different regions of the lamin A/C protein depending upon the disease. Missense mutations causing EDMD map either to highly conserved residues in the lamin A/C tail (Bonne et al., 1999; Raffaele et al., 2000), or to equally highly conserved residues throughout the coiled-coil domain (Raffaele et al., 2000). The spread of mutations causing AD-EDMD suggest that lamin A/C makes a number of molecular interactions at the INM and that different mutations disrupt different interactions. Reported interactions are with emerin (present study; Fairley et al., 1999; Clements et al., 2000), lamin B (present study; Dyer et al., 1999), lamin A to lamin C (present study; Pugh et al., 1997), the chromatin binding protein LAP2α (Dechat et al., 2000), LAPIC (Powell and Burke, 1990), the tumour suppressor protein p110RB (Ozaki et al., 1994) and chromatin (Höger et al., 1991; Glass et al., 1993). If the model presented above is correct, mutations in lamin A/C and perhaps in emerin might give rise to loss or weakened association of lamin C with the lamina in EDMD. This in turn would lead to the possibility of abnormal lamin C complexes forming elsewhere in the nucleus (e.g. the nucleolus). Indeed, in a preliminary investigation of LCLs from 20 EDMD patients, abnormal nucleoplasmic distributions of lamin C were observed in a majority of cases (E. Wang, M.W. and C.J.H. unpublished). Given the range of associations reported for A-type lamins, abnormal nuclear localisation of lamin C may result in significant and deleterious gain of function. Therefore, abnormal nuclear distributions of lamin C rather than an absence of emerin from the INM may promote EDMD and other lamin diseases.
The authors thank Birgitte Lane (University of Dundee) for the supply of LN43, Juliet Ellis (Kings College London) for the supply of rabbit anti-emerin antibodies, and Roland Foisner (University of Vienna) for the supply of LAP17 antibodies. We also thank an anonymous reviewer for their careful and constructive criticism of this paper. This work was supported by grants from the Wellcome Trust to C.J.H. and W.G.F.W., and grants from the CRC to C.J.H. M.W. was supported by grant We1470/12-1 of the Deutsche Forschungsgemeinschaft.