LAP2α and BAF transiently localize to telomeres and specific regions on chromatin during nuclear assembly

The nuclear lamina, a constituent of the nuclear envelope (NE), and intranuclear lamin complexes are major structural elements in nuclei of multicellular eukaryotes (Goldman et al., 2002; Hutchison, 2002). The nuclear lamina lines the inner nuclear membrane and forms a meshwork of tetragonally organized filaments of type V intermediate filament proteins, the lamins (Stuurman et al., 1998). Vertebrates have three lamin genes encoding seven distinct isoforms (Cohen et al., 2001). B-type lamins are constitutively expressed throughout development and are essential for cell viability (Harborth et al., 2001; Lenz-Bohme et al., 1997; Liu et al., 2000). A-type lamins are not essential, but may have functions in tissue organization and homeostasis in the adult organism (Mounkes et al., 2003; Sullivan et al., 1999). Inherited mutations in the lamin A gene in humans give rise to a heterogeneous group of diseases, termed laminopathies, which affect heart, muscle, adipose, bone and nerve cells (Burke and Stewart, 2002), or lead to premature aging (Cao and Hegele, 2003; De Sandre-Giovannoli et al., 2003; Eriksson et al., 2003). Interestingly, A-type lamins in particular are not restricted to the nuclear periphery but exist throughout the nuclear interior (Hozak et al., 1995; Jagatheesan et al., 1999; Moir et al., 2000).

In addition to lamins, numerous lamin-binding proteins (Burke and Stewart, 2002; Foisner, 2001; Simos and Georgatos, 1992; Ye and Worman, 1994) define the structure and function of lamin complexes. Lamin B receptor (LBR) is a protein of the inner membrane with eight transmembrane domains (Worman et al., 1990) and binds B-type lamins (Simos and Georgatos, 1992; Ye and Worman, 1994). Lamina-associated polypeptide 2 (LAP2) is a family of six alternatively spliced proteins, of which four, LAP2β, -γ, -δ and -ϵ are type II membrane proteins (Berger et al., 1996; Dechat et al., 2000b; Harris et al., 1994). LAP2α is structurally and functionally different, sharing only the N-terminus with the other isoforms. LAP2β binds lamin B (Foisner and Gerace, 1993; Furukawa et al., 1998) at the NE whereas LAP2α interacts with A-type lamins in the nucleoplasm (Dechat et al., 1998; Dechat et al., 2000a). Emerin is an inner nuclear membrane protein (Manilal et al., 1996) that binds both A- and B-type lamins in vitro (Clements et al., 2000; Lee et al., 2001; Sakaki et al., 2001) and is retained in the NE by A-type lamins (Holt et al., 2003; Sullivan et al., 1999; Vaughan et al., 2001). All LAP2 isoforms, emerin and the inner membrane protein, MAN1 (Lin et al., 2000) share a ∼40 amino acid-long structural motif, the lamina-associated polypeptide emerin MAN1 or LEM domain (Cai et al., 2001; Laguri et al., 2001), which mediates binding to BAF (Furukawa, 1999; Shumaker et al., 2001). BAF is a small highly conserved protein in multicellular eukaryotes that binds double stranded DNA without sequence specificity (Zheng et al., 2000).

In higher eukaryotes the nucleus is transiently disassembled during mitosis because of a phosphorylation-dependent disassembly of the lamina and pore complexes (Burke and Ellenberg, 2002; Foisner, 2003), a process facilitated by a microtubule/dynein-mediated deformation and disruption of the nuclear envelope (Beaudouin et al., 2002; Salina et al., 2002). Mitosis-specific phosphorylation of lamins was found to be essential for lamina disassembly (Heald and McKeon, 1990) and lamin-binding proteins are also phosphorylated in mitosis (Dechat et al., 1998; Foisner and Gerace, 1993; Nikolakaki et al., 1997).

After sister chromatid separation the NE and nuclear structure reassemble in a tightly regulated manner, ensuring that the interphase chromatin organization can be re-established in daughter nuclei (Cohen et al., 2001). Nuclear reassembly requires phosphatase activity and, at least for B-type lamins, has been shown to involve the membrane-associated phosphatase PP1 (Steen and Collas, 2001; Thompson et al., 1997). The kinetics of the association of lamins and some lamin-binding proteins with the reforming nucleus have been studied in the past years. LBR and emerin are targeted to chromosomes about 5 minutes after metaphase-anaphase transition, followed by nuclear pore complex assembly (Haraguchi et al., 2000). LAP2β was also found to accumulate at around the same time as emerin (Bodoor et al., 1999; Dabauvalle et al., 1999; Foisner and Gerace, 1993). B- and A-type lamins follow different pathways (Moir et al., 2000). Stable B-type lamin structures are seen at the nuclear envelope after pore complex assembly (Chaudhary and Courvalin, 1993; Dabauvalle et al., 1991; Daigle et al., 2001) followed by the bulk of lamin A that accumulates in the nuclear interior and at the NE later. Addition of protein mutants or antibodies to in vitro nuclear assembly assays or transient expression of mutants in cells has implicated lamins, LBR, LAP2β and LAP2α in nuclear assembly (Gant et al., 1999; Lopez-Soler et al., 2001; Lourim and Krohne, 1994; Pyrpasopoulou et al., 1996; Vlcek et al., 2002), but the molecular mechanisms remain unclear.

Here we perform detailed analyses of the dynamics of LAP2α during nuclear assembly in relation to other NE and chromatin proteins and show several major new findings. Following initial binding to telomeres, LAP2α is the first among a group of lamina proteins, including lamin C and emerin, detectable in the `core' structures on chromatin that have been described previously (Haraguchi et al., 2001). In contrast, LBR, LAP2β and lamin B initially bind to distinct, peripherally located regions of the chromatin bulk, and spread to core structures later. In support of previous studies implicating BAF in NE assembly (Haraguchi et al., 2001; Segura-Totten et al., 2002), we provide evidence that a subfraction of BAF relocalizes to core structures together with LAP2α. We propose that these structures may define chromatin organization in the reassembling nucleus.

GFP-LAP2α, H2B-CFP and CFP-lamin B1 have been described previously (Beaudouin et al., 2002; Daigle et al., 2001; Ellenberg et al., 1997; Vlcek et al., 2002). YFP-LAP2α was constructed like GFP-LAP2α except that pEYFP-C1 was used instead of pEGFP-C1 (Clontech Laboratories, Palo Alto, CA). For construction of CFP-lamin C, oligonucleotides (5′-TCGAGACTAGTGCCGGCGAATTCG-3′ and 5′-GATCCGAATTCGCCGGCACTAGTC-3′) were inserted into pECFP-C3 via XhoI and BamHI sites, introducing SpeI and EcoRI restriction sites. cDNA encoding lamin C (a gift from G. Krohne, Würzburg, Germany) was cloned into modified pECFP-C3 via SpeI and EcoRI. Emerin-CFP was generated by cloning a cDNA encoding emerin amino acids 3-254 fused to a Flag-tag at its N-terminus (Ostlund et al., 1999) into pECFP-N1 (Clontech) via SacI and KpnI sites.

HeLa and NRK cells were routinely maintained in high glucose DME medium supplemented with 10% FCS, 10 mM HEPES, pH 7.0 and 50 μg/ml penicillin and streptomycin (Life Technologies, Paisley, UK) at 37°C and 5% CO₂. For generation of stable HeLa cell clones expressing GFP- or YFP-LAP2α, cells were transfected with plasmids using the standard calcium phosphate method and clones were selected in medium containing 700 μg/ml G418 (Life Technologies) and maintained in medium plus 100 μg/ml G418. Transient transfections were performed with Fugene 6 (Roche, Mannheim, Germany). For live cell microscopy, cells were cultured in two-well LabTek chambers with glass bottoms (Nunc, Rochester, NY) and cell growth was synchronized by arresting cells 24 hours post transfection in 0.5 μg/ml aphidicolin (Sigma-Aldrich, St Louis, MO) for 15 hours, followed by a 7-hour release in complete medium. For imaging, Phenol Red-free DME supplemented with 20% FCS and 0.5 mg/ml acetylsalicylic acid was used.

Cells on plastic dishes were fixed with 3.7% formaldehyde or 2% paraformaldehyde in PBS for 20 minutes at room temperature, followed by incubation in 50 mM NH₄Cl/PBS and 1% Triton X-100/0.1% SDS for 5 minutes each. Samples were incubated in 0.2% gelatin/PBS for 30 minutes prior to antibody incubation. Primary and secondary antibodies were applied in gelatin/PBS for 1 hour each at room temperature. Primary antibodies were goat antiserum N18 against lamins A/C (Santa Cruz Biotechnology, Santa Cruz, CA), rabbit antiserum against LAP2α (Vlcek et al., 2002), antiserum to BAF (Furukawa, 1999), monoclonal antibody 4A794 to TRF2 (Upstate Biotechnology, Lake Placid, NY) and monoclonal antibodies 15-2 and 17 to LAP2α and LAP2β, respectively (Dechat et al., 1998); secondary antibodies used are donkey anti-goat IgG conjugated to Cy3 and goat anti-mouse IgG conjugated to Texas Red (Jackson ImmunoResearch, West Grove, PA) and goat anti-rabbit IgG conjugated to Alexa488 (Molecular Probes, Leiden, The Netherlands). DNA was stained with 1 μg/ml Hoechst dye 33258 (Calbiochem-Behring) for 10 minutes. Samples were mounted in Mowiol and viewed in a Zeiss LSM 510 confocal microscope (Carl Zeiss, Jena, Germany).

Live cell imaging was performed at 37°C maintained by an air stream incubator (ASI 400; Nevtek, Burnsville, VA) in conjunction with an objective heater (Bioptechs, Butler, PA) on a customized LSM 510 confocal microscope fitted with a z-scanning stage, selected photomultiplier tubes (Carl Zeiss), a Kr 413 nm laser (Coherent, Dieburg, Germany) and custom dichroics and emission filters (Chroma, Brattleboro, VT) for fluorescent protein imaging as described elsewhere (Daigle et al., 2001; Gerlich et al., 2001). Mitotic events were monitored with a PlanApochromat 63× N.A. 1.4 oil DIC objective (Carl Zeiss) every 10 seconds for 3 minutes. Image series from a single confocal z-plane representing selected time points were assembled and single confocal z-stacks of each time point were projected with the LSM 510 2.3 software (Carl Zeiss) and combined into movies using ImageJ software (http://rsb.info.nih.gov/ij/). False colors for optimal presentation were used: mainly green for YFP and red for CFP. Rendering of four-dimensional images was by graphical reconstruction. Noise levels were reduced using an anisotropic diffusion filter on individual image slices (Gerlich et al., 2001). Isosurface reconstruction and rendering was carried out using the Amira 2.3 software package (TGS, San Diego, CA, USA). Photobleaching was done as described previously (Daigle et al., 2001).

Mitotic cells harvested from an unsynchronized culture were incubated in 0.075 mM KCl for 20 minutes at room temperature and lysed by adding 0.1% Tween 20. Samples were spun onto coverslips at 500 g for 3 minutes with a Cytospin 2 (Therme Shandon, Pittsburgh, USA), fixed either in methanol (-20°C for 5 minutes) or 2% formaldehyde (10 minutes at room temperature) and processed for immunofluorescence microscopy.

Mitotic NRK cells were lysed in KHM buffer and chromosomes spun out as described (Vlcek et al., 2002). The lysate was pre-cleared by addition of 50 μl 50% protein G-sepharose beads and centrifugation at 400 g for 5 minutes with a tabletop centrifuge. 10 μl protein G-sepharose beads, preincubated in 1 ml of hybridoma supernatant containing antibody to LAP2α or 10 μl untreated beads (control) were added to the lysates and harvested by centrifugation following a 2 hour incubation. Beads were washed in KHM buffer and proteins solubilized in 3× SDS-PAGE sample buffer.

BAF was in vitro translated using the TNT® Quick Coupled Transcription/Translation System (Promega) from a pET15b vector (Novagene) containing full-length BAF (Lee and Craigie, 1998). The translation product was diluted sevenfold in binding buffer (50 mM HEPES, pH 7.4, 50 mM NaCl, 5 mM MgCl₂, 1 mM EGTA, 0.1% Triton, 1 mM DTT and 1 mM PMSF), precleared and used for coimmunoprecipitation. Recombinant LAP2α was expressed (Vlcek et al., 1999; Vlcek et al., 2002), renatured by dialyzing into binding buffer and centrifuged at 1500 g for 5 minutes. 100 μl BAF sample was mixed with 100 μl recombinant LAP2α or 100 μl binding buffer as a control and transferred to 50 μl protein G-sepharose beads, preincubated in 1 ml hybridoma supernatant containing antibody to LAP2α. Following a 30-minute incubation, beads were underlaid with 30% sucrose and harvested by centrifugation.

SDS-PAGE was performed according to Laemmli (Laemmli, 1970). For solubilizing BAF, 6 M urea was added to sample buffer. For immunoblotting, primary antibodies were hybridoma supernatants of LAP2 antibodies (Dechat et al., 1998) and BAF antiserum 3273 (Haraguchi et al., 2001); secondary antibodies, alkaline phosphatase or peroxidase-conjugated goat antibodies. Proteins were detected with the Protoblot Immunoscreening System (Promega) or the Super Signal ECL (Pierce, Rockford, CA, USA).

To investigate the cell cycle-dependent dynamic behavior of LAP2α we performed 3D time-lapse live cell imaging (4D imaging) (Gerlich and Ellenberg, 2003) of HeLa cells stably expressing a yellow fluorescent protein (YFP) fusion construct of LAP2α. Transient co-expression of a cyan fluorescent protein (CFP) fusion construct of histone 2B allowed detection of chromosomes in living cells. Like endogenous LAP2α (Dechat et al., 1998), YFP-LAP2α was localized throughout the nuclear interior sparing only the nucleoli in interphase (Fig. 1A). During metaphase only a small fraction of LAP2α was detected on the chromosomes: most of the protein was found dispersed throughout the cytoplasm. After the onset of anaphase (Fig. 1B; Movie 1 in supplementary material), YFP-LAP2α translocated to distinct regions at the tips of lagging chromosome ends protruding from the bulk of segregating sister chromatids towards the midspindle region (Fig. 1B,C, arrowheads). Isosurface reconstruction of individual 3D data sets confirmed the generation and accumulation of LAP2α-containing spot-like structures on the surface of chromatin protrusions (Fig. 1D). LAP2α-enriched patches were first detected approximately 4 minutes after anaphase onset and then increased significantly in size and intensity as mitosis progressed. The majority of cellular LAP2α was redistributed from the cytoplasm to the chromatin within about 6 minutes after metaphase exit, which was accompanied by the retraction of lagging chromosome ends into a compact chromatin mass. During the next minute the protein concentrated into two brightly stained YFP-LAP2α `core' structures at the inner and the outer side of newly forming nuclei, facing the midspindle and spindle pole region, respectively (Fig. 1B, arrows). When cytokinesis was nearly complete and chromosomes partially decondensed, LAP2α redistributed from these core structures back to the nuclear interior where it remained throughout interphase.

The preferred association sites of LAP2α at the tips of lagging chromosomes observed by live cell imaging suggested localization at telomeric structures. To address this possibility we performed double immunofluorescence microscopy of fixed cells using antibodies to LAP2α and to the telomeric TTAGGG repeat binding factor 2 (TRF2) (Broccoli et al., 1997; van Steensel et al., 1998). Although in interphase cells TRF2 antibody stained dot-like telomeres throughout the nucleoplasm, the antibody had a complex dotted pattern on telophase chromosomes. Here, we observed partial overlap of TRF2 structures with LAP2α on chromatin (Fig. 2A), consistent with a targeting of LAP2α to telomeric regions in the early phases of nuclear reassembly.

To identify LAP2α on individual chromosomes we prepared mitotic chromosome spreads. Both a mouse monoclonal antibody (Fig. 2Ba) and a rabbit antiserum to LAP2α (Fig. 2Bb-d) detected LAP2α at the tips of chromosomes, whereas antibodies to LAP2β (Fig. 2Be), or to lamins (data not shown) stained dots scattered throughout the sample without a preferred telomere association. The LAP2α structures on chromosomes differed significantly in size ranging from small dots (Fig. 2Bb) to larger clusters (Fig. 2Bc,d). Furthermore, most of the chromosomes in the preparations with LAP2α localized at the tips contained two chromatids, representing a metaphase stage, whereas in living cells LAP2α patches were first detected at chromosomes in anaphase after sister chromatid separation. From previous studies we knew that binding of LAP2α to chromosomes is controlled by phosphorylation (Dechat et al., 1998; Vlcek et al., 2002) and that incubation of mitotic cell extracts induces dephosphorylation-dependent association of LAP2α with chromosomes. Thus, we hypothesized that during chromosome spread preparation the cells proceeded partially to post-metaphase stages without chromatid separation owing to the lack of a functional spindle. This may induce dephosphorylation and chromosome association of LAP2α. In line with this model, spreads prepared in the presence of phosphatase inhibitors rarely showed LAP2α at telomeres (data not shown). Thus, the preparation of mitotic chromosome spreads in the absence of phosphatase inhibitors may represent the narrow time window in nuclear assembly during which LAP2α associates at chromosomes.

We next investigated how stably LAP2α was associated with chromosomes during nuclear reassembly by fluorescence recovery after photobleaching (FRAP) analysis in cells expressing GFP-LAP2α (Fig. 2C). In interphase cells, FRAP revealed that photobleached regions in the nucleoplasm rapidly recovered to almost their prebleach intensity within 8 seconds. This suggested that LAP2α structures in the nucleoplasm turn over very rapidly as also shown previously for certain proteins in the NPC (e.g. Nup153) (Daigle et al., 2001) and heterochromatin (HP1) (Cheutin et al., 2003). In anaphase, when LAP2α structures first detectable on chromosomes were bleached, fluorescence still recovered to more than half of the prebleach intensity within 8 seconds. However, part of this recovery was probably due to additional LAP2α binding to chromosomes during ongoing nuclear assembly in the recovery time (see unbleached half of the cell in Fig. 2C) rather than recovery to a steady-state situation. In telophase, when the majority of LAP2α was already associated with core structures of chromatin, bleached areas recovered to only about 10% of their prebleach intensity, suggesting that LAP2α was stably associated with these core structures.

Unlike B-type lamins, A-type lamins interact directly with LAP2α in the nucleoplasm, particularly during G1 phase (Dechat et al., 2000a). To compare the kinetics of LAP2α redistribution to chromosomes with those of A- and B-type lamins, we transiently expressed CFP-tagged lamin C or lamin B1 in stable YFP-LAP2α-expressing cells and performed live cell imaging. Both lamins localized primarily at the nuclear periphery in interphase and were dispersed throughout the cytoplasm in mitosis (Fig. 3A). During nuclear reassembly, telomeric LAP2α structures (arrowheads) were clearly detectable before the accumulation of lamins C or B at the chromatin surface, although small fractions of lamins may surround chromatin throughout all stages of assembly. Lamin C and lamin B (arrows) were both detected at higher concentrations on chromatin approximately 1 minute after the first occurrence of telomeric LAP2α, but their localization was different (Moir et al., 2000). A small subfraction of lamin C colocalized with LAP2α at core structures during early stages, but the majority of lamin C was translocated to the nuclear interior at later stages when LAP2α relocated to the nucleoplasm. Although the translocation of A-type lamins into the nucleoplasm during telophase/G1 has been described (Moir et al., 2000), the early association of a subfraction of A-type lamins with chromosomes in anaphase was not observed. To make sure that our observation was not an artifact due to lamin C overexpression, we performed immunofluorescence microscopy of untransfected, fixed cells using antibodies to lamin C. Endogenous lamin A/C behaved in a manner similar to the ectopic protein, partially overlapping with LAP2α structures at chromosomes during anaphase (Fig. 3B).

In contrast to lamin C, the first enrichment of lamin B on chromosomes did not overlap with telomeric LAP2α. Lamin B accumulated initially at chromatin surfaces next to the spindle pole, forming a cap-like structure opposite of the major telomeric LAP2α structures. From there lamin B spread over the entire surface of decondensing chromatin and formed a continuous envelope only when LAP2α translocated to the nucleoplasm (Fig. 3A).

To test where and when during assembly nuclear membrane proteins associated with chromosomes in relation to LAP2α, we transiently expressed CFP fusion proteins of LBR and LAP2β, both binding partners of B-type lamins (Foisner and Gerace, 1993; Furukawa et al., 1998; Meier and Georgatos, 1994; Worman et al., 1988; Ye and Worman, 1994) and emerin, which binds A-type lamins (Lee et al., 2001; Sakaki et al., 2001). As expected, all proteins localized to the nuclear periphery in interphase (Fig. 4). LAP2β was first detectable at chromosomes slightly after LAP2α during nuclear reassembly (Vlcek et al., 2002) whereas LBR appeared at chromosomes at detectable levels slightly before LAP2α (see Movies 2 and 3 in supplementary material). However, both LAP2β and LBR mostly localized to more peripheral sites of the chromatin bulk, distinct from the LAP2α-containing core regions (see also Dabauvalle et al., 1999; Haraguchi et al., 2000). In contrast to LBR, LAP2β also accumulated on chromatin core structures, before both LBR and LAP2β spread over the entire surface of decondensing chromatin at later stages of assembly. Emerin showed a mixed appearance in terms of its association sites. Although the first emerin staining on chromosomes was detected on the peripheral chromatin regions like LBR and LAP2β, the majority of emerin accumulated subsequently at the LAP2α core structures. Altogether, we observed a temporally and spatially highly coordinated accumulation of lamina proteins with chromosomes during nuclear assembly. LAP2α and LBR became concentrated on anaphase chromatin first, whereas other membrane proteins and lamins were detectable at significant levels slightly later. Furthermore, two groups of lamina proteins with different preferred chromosome association sites could be distinguished. LAP2α, followed by emerin and A-type lamins formed the core region of chromatin, whereas LBR followed by LAP2β and lamin B bound to more peripheral chromatin regions initially.

The initial binding of LAP2α to telomere patches and formation of chromatin-associated LAP2α core structures could indicate an important role of LAP2α early in nuclear assembly. Consistent with this hypothesis, in in vitro nuclear assembly studies we have previously shown that C-terminal LAP2α fragments bind to chromosomes and dominantly inhibit assembly of nuclear membranes and lamin A around chromosomes (Vlcek et al., 2002). As the dominant-negative LAP2α mutants lacked the LEM motif, which mediates binding to the DNA crosslinking protein BAF, and BAF mutants deficient in binding to the LEM domain were found to inhibit the assembly of emerin and lamin A in vivo (Haraguchi et al., 2001), we reasoned that complexes of LAP2α and BAF may be involved in early stages of nuclear assembly. To test this hypothesis, we performed immunofluorescence microscopy of HeLa cells at different mitotic stages using antibodies to LAP2α and BAF. Although the majority of BAF was uniformly localized at the chromosomes in early anaphase, endogenous LAP2α was predominantly cytoplasmic (Fig. 5a). However, when LAP2α was first detectable at the chromosomes, a small fraction of endogenous BAF was also enriched at these structures (Fig. 5b, arrowheads) while still present on the entire chromatin and to some extent also in the cytoplasm. Interestingly, upon formation of telomeric LAP2α structures (Fig. 5c) and core structures in telophase (Fig. 5d), BAF also localized to these structures, whereas it was barely detectable throughout the chromatin or cytoplasm at this time. Thus, cytoplasmic and/or chromosome-associated BAF apparently relocalized to core structures on chromatin with similar timing as LAP2α during nuclear assembly. However, we cannot exclude the fact that BAF is still present in the nuclear interior, but epitopes were masked owing to reorganization of chromatin. Other lamina proteins, such as LAP2β, did not strictly colocalize with BAF at core structures at this stage of assembly, but showed a broader distribution over the entire surface of chromatin (Fig. 5e).

To test whether LAP2α and BAF may indeed bind to each other, we investigated the interaction of these proteins in vitro. Although LAP2α was expected to bind BAF via its N-terminal LEM motif, it was important to demonstrate the binding directly, because previous studies using different Xenopus LAP2β isoforms indicated that their slightly different C-termini modulated binding to BAF at the N-terminal LEM motif (Shumaker et al., 2001). Therefore, BAF was produced by in vitro transcription/translation in reticulocyte cell lysates and mixed with purified recombinant LAP2α. Immunoprecipitation of LAP2α significantly coprecipitated BAF (Fig. 6A). Control experiments in the absence of LAP2α or with a C-terminal LAP2α fragment lacking the LEM domain did not precipitate BAF. Thus, BAF is able to interact with the LEM motif in LAP2α in vitro.

In order to confirm the existence of LAP2α-BAF complexes in cells at early stages of assembly, we performed coimmunoprecipitation assays. We synchronized NRK cell cultures using a thymidine block, harvested metaphase cells by mechanical shake-off and lysed cells in a metal ball homogenizer. As up to 50% of total BAF (Holaska et al., 2003; Lin and Engelman, 2003) and the majority of LAP2α (Dechat et al., 1998; Vlcek et al., 2002) are found in the cytoplasm in metaphase cells, we spun out chromosomes from cell lysates and incubated the supernatant for 5-10 minutes at room temperature in order to allow partial post-metaphase assembly of soluble LAP2α complexes. Immunoprecipitation of these LAP2α structures from the lysates precipitated BAF, whereas neither LAP2α nor BAF were precipitated with control beads lacking antibodies (Fig. 6B). Interestingly, BAF in cell lysates and in soluble fractions behaved as a monomeric protein in SDS-PAGE, whereas BAF coprecipitated with LAP2α was primarily detectable as a putative dimer or oligomer (or a posttranslationally modified species) that migrated at approximately 20 kDa on SDS-PAGE (Fig. 6B). We speculate that LAP2α binding to BAF may change the conformation of BAF (see also Forne et al., 2003) and stabilize dimeric or oligomeric forms that become oxidatively crosslinked during sample preparation, and migrate aberrantly on SDS-PAGE (Zheng et al., 2000).

In this study, we investigated the dynamics of LAP2α during nuclear assembly in relation to major components of the NE, which led us to propose a novel model for the molecular mechanisms in initial phases of post-mitotic nuclear assembly (Fig. 7): LAP2α becomes enriched on telomere patches during anaphase and rapidly accumulates at the initial docking sites. The assembly of LAP2α structures at telomeres and chromatin could then give rise to core structures on the inner and outer regions of anaphase chromatin adjacent to the midspindle and spindle poles, respectively. We cannot rule out, however, that the binding of LAP2α to telomeres and its assembly at core structures are independent events mediated by binding of LAP2α to different chromosomal proteins or complexes.

We also demonstrate an in vitro interaction between LAP2α and BAF via the N-terminal region of LAP2α containing the LEM domain, and provide evidence for the existence of LAP2α-BAF complexes in early post-metaphase stages in situ. Accordingly, a large fraction of BAF colocalized with LAP2α at chromatin core regions. It is still unclear whether BAF at LAP2α core structures originated from the cytoplasmic pool of BAF (Holaska et al., 2003; Lin and Engelman, 2003), which could translocate to core regions in a complex with LAP2α, or whether chromosome-bound BAF redistributed transiently to core structures upon binding of LAP2α. The lack of BAF detection on chromatin outside the core regions in telophase would argue for the latter possibility, however, the epitopes of BAF in the internal chromatin structure may be masked at this time. As we found that various N- and C-terminally tagged BAF forms did not behave exactly as endogenous BAF in terms of homo-oligomerization, DNA binding and interaction with LEM-domain proteins (our unpublished data), we could not test GFP-tagged proteins to overcome potential problems of epitope masking.

Other lamina and NE proteins did not colocalize with LAP2α at telomere patches, but a subset of them, including a subfraction of lamin C, emerin, and to some extent also LAP2β, were targeted to the core regions slightly after LAP2α-BAF (Fig. 7). Targeting of these proteins to chromatin cores may be mediated by direct interaction of lamin C with LAP2α or by binding of the LEM-domain proteins, emerin and LAP2β, to BAF. Another group of lamina proteins showed a different pattern of assembly. LBR association with chromatin occurred at more peripheral sites, distinct from the LAP2α-BAF core complexes. LBR targeting to chromosomes may be the trigger for attaching membranes to chromatin and may initiate nuclear membrane assembly. Other membrane proteins, such as emerin and LAP2β may then diffuse within the lipid bilayer and form stable complexes at the chromatin surface by specific interactions with BAF, HA95 (Martins et al., 2003), heterochromatin protein 1 (HP1, Ye and Worman, 1996), histones (Goldberg et al., 1999; Polioudaki et al., 2001; Taniura et al., 1995) or DNA (for a review, see Vlcek et al., 2001).

Our findings show that LAP2α does not bind uniformly to chromosomes during assembly, as one might have expected from our previous in vitro binding studies (Vlcek et al., 1999). Instead, LAP2α initially associates with telomeres and subsequently forms larger complexes, before it relocates into the nuclear interior in late telophase. It is unclear whether the transient interaction of LAP2α with telomeres is mediated by a telomere-associated protein complex or by a direct interaction with DNA. Based on its previously reported properties as a nucleoskeletal component (Dechat et al., 1998), we hypothesize that the specific targeting of LAP2α to telomeres provides a mechanism by which telomeres are positioned within the reforming nucleus during the establishment of higher order chromatin structure after cell division.

As LAP2α and BAF localized at core structures simultaneously during nuclear assembly, potential functions of this interaction in nuclear assembly and possibly higher order chromatin organization may be envisaged. First, the association of LAP2α with chromosome-bound BAF may target LAP2α from the cytoplasm to chromosomes during early stages of assembly. This, however, seems very unlikely based on our observations that: (1) BAF localized uniformly at chromosomes in metaphase and early anaphase, whereas LAP2α was targeted preferentially to patches in the vicinity of telomeres; (2) the constant LAP2 N-terminus, which is common to all LAP2 isoforms and contains the LEM domain, did not accumulate at chromosomes at any stage of nuclear reassembly in vivo or in vitro (Vlcek et al., 1999; Vlcek et al., 2002); (3) LAP2α did not require its N-terminal LEM domain for chromosome targeting, instead its C-terminus (that does not bind BAF) was found to be essential and sufficient for chromosome interaction (Vlcek et al., 1999). Second, LAP2α may target the cytoplasmic pool of BAF to chromosomes during nuclear reassembly, which would be consistent with our observations. Third, both proteins could be targeted independently to chromosomes and associate in specific subregions of chromosomes. Alternatively, chromosome-bound BAF may be the predominant binding partner for independently targeted LAP2α, whereas cytoplasmic BAF may not be involved in the assembly at all.

We hypothesize that, independent of the targeting mechanisms of LAP2α and BAF, the complex may help to re-organize chromatin during decondensation and nuclear assembly by various mechanisms: (1) the complex may transiently tether telomeres to stable structures favoring ordered chromatin reorganization. This model is supported by low exchange rates of LAP2α in core structures; (2) the binding of BAF to LAP2α may regulate DNA crosslinking activity of BAF, thus resulting in different DNA (de-)compaction states at defined chromosomal subregions; (3) the formation of LAP2α-BAF complexes may provide docking sites for other lamina proteins favoring their ordered assembly.

The essential role of BAF in cell cycle-dependent chromatin organization has been described previously by means of in vivo and in vitro studies in various systems. Drosophila baf null mutants were lethal at the larval-pupal transition and showed grossly aberrant nuclear structures with chromatin clumps (Furukawa et al., 2003). BAF-depletion in Caenorhabditis elegans by RNA interference caused defects in chromatin segregation (Zheng et al., 2000). Furthermore, BAF mutants defective in binding the LEM domain and DNA did not accumulate at chromosomes and inhibited assembly of lamin A, emerin and LAP2β in HeLa cells (Haraguchi et al., 2001). In vitro, BAF favored or inhibited chromatin decondensation during assembly of nuclei in Xenopus extracts depending on its concentration (Segura-Totten et al., 2002). Finally, our previous observations have indirectly shown a role of LAP2α-BAF complexes in assembly. C-terminal LAP2α mutants that bound chromosomes but lacked the N-terminal LEM domain (Vlcek et al., 2002) and were thus unable to interact with BAF, dominantly inhibited chromatin decondensation and nuclear membrane assembly in in vitro assays.

In summary, we suggest a model for nuclear assembly in which transient accumulation of LAP2α and BAF at telomeres and/or chromatin cores mediates chromatin reorganization and NE assembly. However, given that BAF is an evolutionarily conserved protein found in metazoans (Cai et al., 1998; Lee and Craigie, 1998), whereas LAP2α has only been detected in vertebrates (Dechat et al., 2000b; Vlcek et al., 2001), one has to assume that this mechanism is not essential for nuclear assembly in general, but may contribute to vertebrate-specific chromatin organization.

We thank Katherine Wilson, Johns Hopkins School of Medicine, Baltimore, USA, for providing BAF antiserum and plasmids. This study was supported by grants from the Austrian Science Research Fund (FWF, P15312), from the Jubiläumsfonds of the Austrian National Bank (9006) and from the `Österreichische Muskelforschung' to R.F. and the Japan Science and Technology Corporation to T.H. A.G. was a recipient of a fellowship from the Deutsche Forschungsgemeinschaft (DFG), D.G. of an EMBO long-term fellowship and T.D. of a short-term travel fellowship (EurALMF, EC Transnational Access to Major Research Infrastructures Programme).