Assembly of the nuclear envelope (NE) in telophase is essential for higher eukaryotic cells to re-establish a functional nucleus. Time-lapse, FRAP and FRET analyses in human cells showed that barrier-to-autointegration factor (BAF), a DNA-binding protein, assembled first at the distinct `core' region of the telophase chromosome and formed an immobile complex by directly binding with other core-localizing NE proteins, such as lamin A and emerin. Correlative light and electron microscopy after live cell imaging, further showed that BAF formed an electron-dense structure on the chromosome surface of the core, close to spindle microtubules (MTs) prior to the attachment of precursor NE membranes, suggesting that MTs may mediate core assembly of BAF. Disruption of the spindle MTs consistently abolished BAF accumulation at the core. In addition, RNAi of BAF eliminated the core assembly of lamin A and emerin, caused abnormal cytoplasmic accumulation of precursor nuclear membranes and resulted in a significant delay of NE assembly. These results suggest that the MT-mediated BAF accumulation at the core facilitates NE assembly at the end of mitosis.

The nuclear envelope (NE) is a cellular structure that encloses chromosomes and provides a physicochemical framework for genetic activities such as gene expression and DNA replication. The NE is a dynamic structure in eukaryotes, disassembling at prophase and reassembling at telophase in each mitosis. The reassembly of the NE is crucial to re-establish a functional nucleus for the next interphase and to ensure proper progression of the cell cycle. Several lines of evidence have shown that defects in NE architecture can cause diseases in animals. Recently, a variety of human disorders known as `nuclear envelopathy' or `nuclear laminopathy' have been reported (reviewed by Nagano and Arahata, 2000; Wilson et al., 2001; Worman and Courvalin, 2005). These disorders involve mutations in NE components. Mutations in the gene encoding lamin A (official protein symbol, LMNA) cause 12 human diseases (reviewed by Mattout et al., 2006) including Emery-Dreifuss muscular dystrophy (Bonne et al., 1999; Clements et al., 2000) and Hutchinson-Gilford progeria syndrome (Eriksson et al., 2003; Mounkes et al., 2003). Emery-Dreifuss muscular dystrophy can also be caused by mutations of the gene encoding emerin (official protein symbol, EMD) (Bione et al., 1994). The fact that defects in the NE can cause many different diseases indicates that the NE is involved in a wide variety of cellular functions. Thus, understanding the dynamic architecture of the NE will provide insight into the regulatory mechanisms for these cellular functions.

Molecular mechanisms of NE formation have been well described in in vitro experimental systems (Hetzer et al., 2000; Harel et al., 2003; Hachet et al., 2004); however, compared with NE formation in vitro, NE reformation in vivo is poorly understood. It is believed that it begins with attachment of precursor membranes to telophase chromosomes, followed by fusion of the membranes and reassembly of integral membrane proteins, the NPC, and nuclear lamina into the NE. Chromatin-binding integral membrane proteins, such as lamin B receptor (LBR) and LAP2β (official symbol, LAP2B) are proposed to posses a role in early assembly of the NE (Gant and Wilson, 1997; Ulbert et al., 2006). Assembly of LEM-domain NE proteins such as emerin and LAP2β depends on barrier-to-autointegration factor (BAF), a chromatin-binding protein, in human cells (Haraguchi et al., 2001).

BAF forms dimers that bind nonspecifically to double-stranded DNA in vitro (Lee and Cragie, 1998) and can be phosphorylated in its N-terminal region by the vaccinia-related kinases (Nichols et al., 2006; Gorjánácz et al., 2007). BAF was first discovered as a cellular protein that prevents retroviral DNA from suicidal auto-integration and ensures its integration into host DNA (Lee and Cragie, 1998: Jacque and Stevenson, 2006). It was later described as a binding partner of LEM-domain NE proteins (Furukawa, 1999; Lee et al., 2001; Shumaker et al., 2001). BAF is proposed to have a function in NE assembly in human cells (Haraguchi et al., 2000) and C. elegans (Margalit et al., 2005; Gorjánácz et al., 2007) as well as in Xenopus oocyte extracts (Segura-Totten et al., 2002). Recently, it has also been reported to be important for the normal progression of S phase in human cells (Haraguchi et al., 2007).

In human cells, BAF accumulates with emerin, at the central regions of the assembling nuclear rim close to the spindle microtubules (MTs) during telophase when the NE is reforming. This region is named the `core' region (Haraguchi et al., 2001). In our previous study, we found that emerin, LAP2β and lamin A accumulated in the core region during telophase in a BAF-dependent manner (Haraguchi et al., 2001). LAP2α (official symbol, LAP2A) also accumulates at the core (Dechat et al., 2004). Although the core structure seems important for NE assembly of the LEM-domain proteins and lamin A, its structural basis and dynamics in living cells are largely unknown. To understand the structural basis and biological significance of the core region, we have examined physical and chemical properties of the core region by time-lapse microscopy of living cells, FRAP and FRET analyses, and EM analysis combined with live-cell imaging. In addition, we used RNAi to selectively remove individual components of the core structure, BAF, lamin A and emerin, to examine their functions in living cells. Here we report that BAF forms an electron-dense stable structure at the core region of the telophase chromosome in a spindle-MT-dependent manner, and propose that this MT-dependent BAF accumulation at the core coordinates efficient rendezvous of nuclear membranes and chromosomes.

BAF assembles first to the `core' region of the reforming telophase nuclear envelope

Assembly of NE proteins at the end of mitosis is an essential step in the reformation of a functional nucleus in higher eukaryotic cells. In order to elucidate underlying molecular mechanisms of NE reformation in human cells, we examined spatial and temporal sequences of NE assembly by monitoring the behavior of NE-GFP fusion proteins in living cells. The NE proteins tested were BAF (which binds both chromatin and LEM-domain proteins); LAP2β, emerin and MAN1 (membrane-bound LEM-domain NE proteins); LAP2α (LEM-domain protein localized in the nucleus); lamin A and lamin B1 (major components of the nuclear lamina); LBR (classical integral nuclear membrane protein); and nucleoporin Nup35 (official protein symbol, NUP53). Time-lapse observation of HeLa cells expressing each one of these NE fusion proteins showed that BAF, LAP2α, LAP2β, emerin, MAN1 and lamin A accumulated at a specialized region of the telophase chromosome mass close to the spindle (the `core' region; see Fig. 1A and images indicated by arrows in Fig. 1B). This accumulation occurred during a limited period of telophase, as described previously (Haraguchi et al., 2000; Haraguchi et al., 2001; Dechat et al., 2004). In contrast with these core-localizing proteins, lamin B1, LBR and Nup35 did not show clear accumulation at the core region (Fig. 1B).

The fluorescence intensity of NE proteins at the core region is plotted in Fig. 1C; the timing of core localization was determined as the time, in seconds after the metaphase-anaphase transition, at which 80% of maximum accumulation had occurred (Fig. 1D). Among the core-localizing proteins, BAF accumulated earliest at the core region, approximately 434 seconds after the metaphase-anaphase transition. LAP2α, lamin A and emerin accumulated approximately 30 seconds later, and LAP2β and MAN1 accumulated after approximately another 30 seconds (Fig. 1C,D). That BAF accumulated first was confirmed by observation of HeLa cells expressing mRFP-BAF transiently co-transfected with GFP-emerin or GFP-lamin A. In these cells, mRFP-BAF accumulated in the core region earlier than GFP-emerin or GFP-lamin A (data not shown). These results indicate that BAF appears at the core region prior to the accumulation of other core-localizing NE proteins. Since BAF binds both emerin and lamin A in vitro (Lee et al., 2001; Holaska et al., 2003), the early accumulation of BAF at the core region suggests the possibility that BAF provides a structural foothold to allow assembly of nuclear lamins and membrane-bound NE proteins. The time course of assembly to the reforming NE at telophase is schematically summarized in Fig. 1E.

BAF forms an immobile structure at the core region

From our time-lapse analysis, we hypothesized that BAF might form a foothold on the chromosome surface to allow assembly of other NE proteins. To test this hypothesis, we examined the mobility of BAF at the `core' region during telophase by FRAP analysis of HeLa cells stably expressing GFP-BAF. We previously reported that GFP-BAF during interphase was highly mobile with a half time recovery of 270 mseconds and an immobile fraction (IM) of approximately 3% (Shimi et al., 2004). GFP-BAF was also highly mobile during metaphase (data not shown). By contrast, GFP-BAF became highly immobile during telophase with an IM of 72±16% (n=7) (Fig. 2A,B), and remained immobile until the core region disappeared at the end of M phase; GFP-BAF regained its high mobility in early G1 phase (data not shown). This result indicates that physical attributes of BAF are regulated during the cell cycle, with BAF forming an immobile structure at the core region specifically in telophase, a characteristic also exhibited by LAP2α (Dechat et al., 2004).

To further examine physical properties of the core structure, we determined the mobility of other core-localizing NE proteins, such as lamin A and emerin, at the core region. These NE components showed a similar mobility to that of BAF with an IM of 74±14% (n=5) for lamin A and 46±7% for emerin (n=5) (Fig. 2C,D). These results suggest that BAF may form a stable higher-order complex together with these NE proteins at the core.

BAF directly interacts with BAF, lamin A and emerin at the core region

Because BAF, lamin A and LEM-domain NE proteins all become immobile at the core region, the core structure may consist of a higher-order complex of BAF with lamin A and the LEM domain NE proteins. To test whether BAF directly interacts with these NE proteins at the core region, we used fluorescence Förster resonance energy transfer (FRET) analysis. In HeLa cells transiently coexpressing mVenus-BAF and mCFP fusions of the second protein, FRET signals were detected by acceptor photobleaching (Fig. 3). The results showed that mVenus-BAF produced FRET signals with mCFP-BAF (Fig. 3A,B), mCFP-lamin A (Fig. 3C,D) and mCFP-emerin (Fig. 3E,F). In addition, time-lapse ratio imaging between mCFP-BAF and YFP-BAF showed that FRET signals between BAF molecules dramatically increased in the core region during telophase (Fig. 3I). As BAF forms a dimer or dodecamer in vitro (Zheng et al., 2000), this high FRET signal may represent a higher-order complex structure. LAP2α fused with mCFP also produced FRET signals with mVenus-BAF in the core (data not shown), consistent with the previous finding that LAP2α is colocalized with BAF at the core structure (Dechat et al., 2004). A mutant emerin that does not bind to BAF (emerin-m24) either in vivo (Haraguchi et al., 2001) or in vitro (Lee et al., 2001) was used as a negative control. No FRET signals were detected between mCFP-emerin-m24 and mVenus-BAF (Fig. 3G,H), suggesting that the detected FRET signals are the result of specific interactions. Estimated FRET efficiencies are summarized in Table 1. These results suggest that BAF probably forms a polymer or a higher-order complex structure with BAF itself in the core region and also directly binds emerin, lamin A and LAP2α, assembling them into the core complex.

BAF forms an electron-dense complex on the surface of the telophase chromosome mass

To examine the core structure at high resolution, we used electron microscopy (EM). Because the core structure forms transiently, for only a few minutes, during telophase, we selected specimens by observing living HeLa cells expressing GFP-BAF. During live-cell observation, the cell was fixed when the core structure formed and embedded in epoxy resin to prepare EM specimens (Fig. 4A; see Materials and Methods). Three-dimensional fluorescence microscopy images of the fixed cell were also obtained (Fig. 4B). Thin EM sections were compared with fluorescence microscopy images to identify the cell observed during live-cell observation (Fig. 4C,D) and details of the core structure were investigated at high resolution (Fig. 4E,F). Images shown in a, b, and c in Fig. 4F correspond to the regions indicated in Fig. 4E. This method provides an opportunity to combine the temporal information and molecular selectivity of fluorescence live-cell imaging with the high-resolution imaging of EM: this method is designated live correlative light electron microscopy (live CLEM). Examples fixed at later stages are shown in Figs 5 and 6. The reliability of this method was confirmed in an immunoelectron micrograph for GFP-BAF localization using anti-GFP antibody (supplementary material Fig. S1): GFP-BAF was localized in the core region of telophase cells as seen in the live CLEM imaging technique.

We applied live CLEM to cells fixed at approximately 4 minutes (n=4), 7 minutes (n=17) and 10 minutes (n=3) after the metaphase-anaphase transition, (Figs 4, 5 and 6, respectively). The cell shown in Fig. 4 was fixed at 4 minutes 15 seconds after the metaphase-anaphase transition, before BAF became concentrated at the core. Comparison of EM images with those from fluorescence microscopy shows that BAF was localized to only a limited region of the chromosome mass (superimposed with the green region in panel b of Fig. 4F); this region represents a trailing edge of the separating chromosomes, which might correspond to the telomeres, as suggested by Foisner and colleagues (Dechat et al., 2004). In the central region of the chromosomes where BAF localizes, spindle MTs approached and penetrated into the chromosome mass (superimposed with the orange lines in regions b and c of Fig. 4F). At this time, the NE (superimposed with the red hatches) was not being formed in the central region of the chromosome mass where the mitotic spindle was present; rather, the most peripheral regions (`non-core' region in Fig. 1A) of the chromosome mass were being enveloped by the NE (see panel a of Fig. 4F; also see supplementary material Fig. S2A,B). The purple arrowheads indicate vesicular membranes. These vesicular membranes are presumably part of the ER tubules, the precursors of nuclear membranes (Anderson et al., 2007). It should be emphasized that almost no NE was formed in the central region (b and c in Fig. 4F) where spindle MTs are present, whereas the NE was well defined in the peripheral regions where spindle MTs are few or not present at all.

The cell shown in Fig. 5 was fixed at 7 minutes 10 seconds after the metaphase-anaphase transition, when distinct core localization of BAF was observed. Strikingly, an electron-dense area, with a width of 24-64 nm, overlapping with the localization of BAF (superimposed with green) was clearly seen on the surface of the chromosome mass (Fig. 5G). At this time, well-defined NE could be observed in this region of the chromosome mass, but the NE was not yet fully assembled at the sites of the electron-dense areas (panel a in Fig. 5G). Vesicular membranes could be seen near these electron-dense areas (indicated by purple arrowheads in Fig. 5G); some vesicles with a diameter of approximately 50 nm appeared to attach to the BAF-containing electron dense areas (see panel c in Fig. 5G and panel b in Fig. 6G). In addition, vesicular membranes at the site of the reforming NE at the core region appeared to be attached to MTs (see arrowheads in panel c in Fig. 5G and panel b in Fig. 6G). It could also be seen that some MTs had penetrated into the chromosome mass in the gap between two BAF-containing electron-dense areas (panel b in Fig. 5G), and others attached to the surface of the electron-dense area (panel d in Fig. 5G). At the same time, most of the non-core region of telophase chromosomes was overlaid with NE (supplementary material Fig. S2C,D), suggesting that NE reformation in the core and non-core regions are regulated by distinct mechanisms.

It should be pointed out that the core region is not fully covered by NE at 7 minutes; complete reassembly of the NE does not occur until approximately 10 minutes (Fig. 6). This is surprising because nuclear transport started at 7-8 minutes (Fig. 5H, arrows; also see Haraguchi et al., 2000). At this time, the NPCs that have been targeted to the NE were localized exclusively outside the core region (Fig. 1B; also see supplementary material Fig. S2C,D). Collectively, these results suggest that the NPCs on the reforming NE at 7-8 minutes after the metaphase-anaphase transition are functional and that nuclear transport through the NPCs may begin in the absence of a fully reassembled NE.

The cell shown in Fig. 6 was fixed at 10 minutes 11 seconds after the metaphase-anaphase transition. At this time, most of the chromosome mass was surrounded by the NE; the BAF-containing electron-dense surface (green) was also enveloped within the NE (red). This is consistent with previous findings that emerin-containing and LBR-containing membranes fuse at this time (Haraguchi et al., 2000). Most spindle MTs have disappeared (panels a and b Fig. 6G), but some MTs were still present close to the chromosomes. In some cases, MT cables proximate to the electron-dense chromosome areas were disrupted by the presence of the NE (panel c Fig. 6G). Importantly, electron-dense structures formed on the surface of the telophase chromosome mass in HeLa cells not expressing exogenous GFP-BAF (supplementary material Fig. S3), indicating that this structure is a native one, and not an artefact of GFP-BAF.

A proposed order of NE formation at the core region is summarized in Fig. 6H. The NE forms directly on the chromosomes outside the core region, but forms on the BAF-containing dense structures at the core region. It appears that accumulation of the electron-dense material eliminates MT attachment to the chromosome mass as the electron-dense core grows. The BAF-containing electron-dense structure then disappears after the NE covers the entire chromosome mass.

BAF has a direct role on NE formation at the core region, and lamin A and emerin affect the lifetime of the core structure

To determine the contribution of BAF to core formation and subsequent NE formation, we examined the effects of BAF knockdown by RNAi in HeLa cells. BAF was selectively removed in cells treated with BAF siRNA (Fig. 7A,D). In those cells, core localization of emerin and lamin A was effectively inhibited (Fig. 7B); examples of microscopic images are shown in Fig. 7C. In addition, we applied live CLEM analysis to BAF siRNA-treated HeLa cells expressing GFP-BAF, and observed cell structures especially focusing on the electron-dense structure, the NE, chromosomes, the MTs and membranes (Fig. 7E, Figs 8 and 9; also see supplementary material Fig. S4). In BAF siRNA-treated cells (see Fig. 7D), the electron-dense structure at the periphery of the central region of the chromosome mass was lost (n=4; Fig. 7E, compare arrowheads in cells at 7 minutes after the metaphase-anaphase transition with arrows in the control cells), and chromosomes remained condensed (Fig. 8, compare G or M with B), indicating that BAF is required for formation of the core structure and for chromosome decondensation. In addition, the NE at the core region had not formed in BAF siRNA-treated cells even at 12 minutes after the metaphase-anaphase transition (n=10; see Fig. 8I,J) whereas it was almost fully reassembled in luciferase siRNA-treated control cells (n=4; Fig. 8D). In the BAF-depleted cells it was also frequently observed that relatively long MTs remained in the chromosomal region at this late period (compare Fig. 8J with 8D; also compare Fig. 9F with 9C). Finally, abnormal inclusion of cytoplasm inside the nucleus was also often seen these cells (Fig. 8P,Q; pale regions in nucleus).

Interestingly, in addition to these phenotypes, BAF depletion caused a significant reduction of tubular membranes near the core region where the spindle MTs persisted, in contrast to an abundance of tubular membranes in the presence of BAF (compare Fig. 9E,F with 9B,C). Instead, in BAF RNAi cells, abnormal piles of three-dimensionally extended ER membranes in the cytoplasm were generated (Fig. 9G,H; also see Fig. 8K,L, compare these with Fig. 8E): in many cases these abnormal membranes seemed to be NE-like membranes assembling away from the chromosomes (Fig. 9G,H). These findings are schematically summarized in Fig. 9I. Taken together, these results indicate that BAF has a direct effect on NE formation at the core region, as suggested previously (Haraguchi et al., 2000), and suggest that the role of BAF in NE formation is to coordinate the recruitment of NE precursor membranes to the chromosome surface.

RNAi knockdown of lamin A did not eliminate core localization of BAF or emerin, but did reduce the frequency of core localization (Fig. 7B). To test whether this reduced frequency reflected a shortened life of the core structure, we determined the duration of the core structure (Fig. 7F) in cells with specific RNAi-mediated knockdown of BAF, lamin A or emerin, as shown in Fig. 7D. Its duration was significantly shortened from 8 minutes 9 seconds (n=10) in control luciferase siRNA cells to 1 minute 42 seconds (n=7) in lamin A siRNA cells, suggesting that lamin A stabilizes the core structure (Fig. 7F). By contrast, in emerin siRNA cells, the core structure remained intact significantly longer (more than 20 minutes) (n=7) compared with control RNAi cells (n=7) (Fig. 7F), often persisting up to early G1 phase (data not shown), suggesting that emerin destabilizes the core structure.

Furthermore, knockdown of these proteins also affected localization to the NE in the subsequent interphase. Knockdown of BAF disrupted NE localization of emerin and lamin A, knockdown of lamin A disrupted NE localization of BAF and emerin, and knockdown of emerin disrupted NE localization of BAF and lamin A (Fig. 7G). These effects are specific because the same treatment did not affect lamin B localization (Fig. 7G). Taken together, these results suggest that BAF has a direct role on core formation and that lamin A and emerin affect the stability of the core structure.

The spindle MTs mediate BAF assembly to the core region

BAF forms an electron-dense core structure close to the MT-attaching regions (Fig. 5). Therefore, the MTs could be involved in BAF assembly to the core region. To test this idea, we first examined whether BAF colocalizes with the spindle MTs during mitosis. Indirect immunofluorescence staining of cells fixed with trichloroacetic acid (TCA) showed that in mitotic cells BAF localizes to the spindle MTs from metaphase to telophase (Fig. 10A). Since this result shows colocalization of BAF with the spindle MTs for the first time, we used BAF knockdown by BAF RNAi to confirm our initial observation; BAF RNAi treatment effectively removed the signal, indicating that BAF, at least in part, colocalizes with the spindle MTs (supplementary material Fig. S5). To further test a direct role of the spindle MTs in core formation, we examined the effect of nocodazole, a MT-depolymerizing reagent, on BAF assembly to the core region. In cells treated with nocodazole during anaphase, the spindle MTs disappeared and BAF core assembly was lost (Fig. 10B), indicating that the spindle MTs are required for BAF assembly to the core. Taken together, these results suggest that BAF is associated with the spindle MTs during metaphase through anaphase and assembles to the core region in early telophase, in a spindle-MT-dependent manner, where it forms a stable complex with other core-localizing proteins such as lamin A and emerin. Based on these findings, we propose a model, schematically illustrated in Fig. 10C, in which spindle MT function organizes BAF spatially, thereby engendering BAF-directed NE assembly.

In this report, we have demonstrated that the BAF-dependent core structure is an immobile complex that appears as an electron-dense structure directly on chromosomes with a width of 24-60 nm. This study shows for the first time that the spindle MTs are involved in NE reformation at the end of mitosis: spindle MTs mediate BAF assembly to the core region where BAF recruits selective NE components such as lamin A and emerin. The core structure is important in establishing a normal interphase NE and, consequently, normal nuclear function.

Live CLEM observation revealed that the timing of NE reformation is different in the core region from that in the non-core region: NE reformation is significantly later in the core region (compare Figs 4, 5, 6 with supplementary material Fig. S2). This suggests that there are at least two different mechanisms for NE assembly; one that acts at the core region, and a second that acts at the non-core region. RanGTP would appear to be involved in NE reformation at the non-core region because it is known to be involved in the assembly of the NPCs in in vitro Xenopus oocyte extracts (Zhang et al., 2002; Walther et al., 2003) and our observations indicate that NPCs assemble exclusively in the non-core region. We propose that BAF is a factor involved in NE assembly at the core region.

Live cell imaging of various ER- or NE-associated integral membrane proteins have shown that most membranes are excluded from the area of the spindle MTs until early telophase (Ellenberg et al., 1997; Yang et al., 1997; Haraguchi et al., 2000; Anderson and Hetzer, 2007; Anderson and Hetzer, 2008). From these observations, we speculate that the mitotic spindle, or their associated proteins, may have an inhibitory effect on NE reformation, presumably by somehow repelling precursor membranes. As a consequence of this, fewer precursor membranes would be available for NE reformation in the core region. Here, we report that BAF associates with the spindle MTs, and that its targeting to the core region requires spindle MTs (Fig. 10B). These findings strongly support the idea that BAF acts to direct core NE reformation by recruiting selective NE protein-bearing membranes in the presence of the `inhibitory' spindle MTs. This hypothesis is further supported by the fact that loss of BAF by RNAi resulted in removal of most of the membranes from the area near the core region during telophase (Fig. 9E,F), and abnormal assembly of NE-like membranes in the cytoplasm far from the chromosomal region (Fig. 9G,H), and also caused lamin A and emerin to disperse into the cytoplasm (Fig. 7C,G). In addition, abnormal inclusion of cytoplasm inside the nucleus was frequently observed in cells lacking BAF (Fig. 8P,Q).

As BAF is a key molecule for core formation, BAF targeting to the core region must be an important step for core formation. Recently, Gorjánácz and colleagues have reported that BAF-1 (C. elegans BAF homolog) is localized in a chromosomal region similar to the `core' during NE formation in dividing cells of early C. elegans embryos, and its `core'-like localization is abolished by siRNA depletion of VRK-1 (C. elegans vaccinia-related kinase homolog) (Gorjánácz et al., 2007). Their results suggest that phosphorylation of BAF is essential for its `core'-like localization in C. elegans embryos. Human BAF can be phosphorylated in vitro on its N-terminal conserved amino acid residues, Ser4 (Bengtsson and Wilson, 2006) or all three amino acid residues Thr2-Thr3-Ser4, by vaccinia-related kinases (Nichols et al., 2006). Paradoxically, however, initial studies have indicated that phosphorylation of human BAF abolishes its interaction with DNA in vitro (Nichols et al., 2006), reduces its binding to the LEM proteins in vitro (Nichols et al., 2006), and disrupts emerin localization to the NE in vivo (Bengtsson and Wilson, 2006), suggesting that the N-terminal phosphorylated form of human BAF cannot attach to the chromosomal region at the `core'. One possibility to reconcile these results is that BAF phosphorylated by VRK1 binds spindle MTs, and then is able to localize to the core region once it is dephosphorylated at the end of mitosis. This hypothesis is supported by the following findings: (1) BAF associates with the spindle MTs (Fig. 10A); (2) BAF core accumulation is abolished by destruction of the spindle MTs (Fig. 10B); and (3) phosphorylated BAF is localized on the spindle MTs but not in the core region (T.H., unpublished result). However, it remains to be elucidated whether phosphorylation by Vrk1 is required for association with the spindle MTs.

Cell-cycle dependent phosphorylation of emerin also regulates the interaction of emerin with BAF (Hirano et al., 2005; Ellis et al., 1998), suggesting that modification of emerin or other core-localizing proteins may also contribute to the core formation and structural features of the `core'. Since mitotic progression is regulated by the spatial and temporal coordination of several cell-cycle-regulating kinases and phosphatases, the combination of several protein modifications may cooperatively enforce core formation in telophase.

Cooperative actions of BAF and core-localizing proteins in NE organization are supported by several observations. BAF, lamin A and emerin form a three-way complex in vitro (Holaska et al., 2003) and in vivo in C. elegans (Liu et al., 2003; Margalit et al., 2005; Gorjánácz et al., 2007). Retention of lamin A and emerin in the reforming NE requires their initial localization at the core region, and this localization is BAF dependent (see Fig. 7G). Core localization is also required for BAF to establish its own proper nuclear localization (Zheng et al., 2000; Margalit et al., 2005; Gorjánácz et al., 2007). Furthermore, the lifetime of the core structure seems to depend on specific core-localizing proteins such as emerin, lamin A and LAP2α. Loss of emerin by RNAi prolonged core lifetime (see Fig. 7F). A persistent core structure is also observed in HeLaS3 cells expressing high levels of lamin A, and this persistent core structure is enriched in emerin and lamin A (Maeshima et al., 2006). We also observed a prolonged core lifetime in cells overexpressing lamin A or LAP2α (data not shown). By contrast, loss of lamin A by RNAi significantly reduced the lifetime of the core structure (see Fig. 7F). These results suggest that emerin destabilizes the core, whereas lamin A and LAP2α stabilize the core structure. The balance between their competing activities may be important for regulating the formation of the transient core structure, the formation of the mature NE and the eventual formation of a functional nucleus.

In summary, we conclude that BAF assembles on the surface of the telophase chromosome mass in regions where the chromosomes are interacting with spindle MTs, and that this assembly is MT dependent and requires coordinated BAF phosphorylation followed by BAF dephosphorylation. Chromosome-associated BAF makes a higher-order complex with itself, lamin A and emerin, resulting in formation of an electron-dense structure at the core which is required for proper reformation of the NE in this region and for the re-establishment of a functional nucleus. Furthermore, BAF-dependent NE reformation at the core, which involves LEM-domain proteins containing NE precursor vesicles attaching to the BAF complex and then fusing with each other, is mechanistically distinct from NE reformation at the non-core region of the chromosome mass. Understanding the dynamics and the structural details of NE reformation at the end of mitosis will provide new insights into the development of lamin-A-dependent disease and other NE-associated human diseases.

Cells and reagents

HeLa cells were obtained from the Riken Cell Bank (Tsukuba, Japan) and maintained in DME medium containing 10% calf serum. Double-stranded siRNA for BAF, lamin A and emerin were purchased from Qiagen (Hilden, Germany). Mouse monoclonal anti-lamin A/C antibodies TiM92 (IgM) were generated as described previously (Haraguchi et al., 2001). Mouse monoclonal anti-lamin A/C antibodies MMS-107P (IgM) were purchased from Covance (Berkeley, CA). Mouse monoclonal anti-lamin B (101-B7) antibodies were obtained from MatriTech (Cambridge, MA). Rabbit polyclonal anti-emerin antibodies were a gift of the Arahata lab (Tokyo, Japan). Anti-BAF antibodies (PU38143) were generated against recombinant BAF in a rabbit and affinity purified. Plasmids encoding mCFP and mVenus (monomeric versions of ECFP and Venus), which were made by substitution of Ala for Lys (Zacharias et al., 2002), were gifts of Nagai, Sawano and Miyawaki (RIKEN, Wako, Japan). LAP2α-YFP and GFP-lamin B1 were gifts of R. Foisner (Medical University Vienna) and R. Goldman (Northwestern University), respectively.

Plasmid construction

Methods for constructing DNA plasmids encoding LBR-GFP, GFP-emerin, GFP-BAF, GFP-LAP2β and GFP-MAN1 were described previously (Haraguchi et al., 2000; Haraguchi et al., 2001, Shimi et al., 2004). To construct GFP-lamin A, the coding region was PCR-amplified from the RT-PCR product of lamin A using primers lam-1 and lam-2 and inserted into the pEGFP-C1 vector (Clontech Laboratories, Palo Alto, CA). To construct mCFP and mVenus fusion proteins, the coding region of mCFP or mVenus was PCR-amplified from ECFP-A206K/pRSETB or Venus-A206K/pRSETB, respectively, with primers cfp-1 and cfp-2 and inserted into the pEGFP-C1 vector digested with NheI and BglII restriction enzymes to replace EGFP. The resulting mCFP and mVenus vectors were used for construction of mCFP-BAF, mCFP-lamin A, mCFP-emerin, mCFP-emerin-m24 and mVenus-BAF. To construct mCFP-BAF or mVenus-BAF, the coding region of BAF was PCR-amplified from pET15b-BAF (Lee and Craigie, 1998) using primers baf-1 and baf-2 and inserted into the above described mCFP or mVenus vectors, respectively. To construct mVenus-emerin or mVenus-emerin-24m, the coding region of emerin was PCR-amplified from a GFP-emerin construct (Haraguchi et al., 2000) using primers eme-1 and eme-2 and inserted into the mVenus vector. To construct mVenus-lamin A, the coding region of lamin A was PCR-amplified from the GFP-lamin A construct using primers lam-3 and lam-4 and inserted into the mVenus vectors. The coding sequence of human Nup35 was obtained from mRNA in HeLa cells by RT-PCR using primers nup-1 and nup-2, cloned into the TA cloning vector pCR2.1 (Invitrogen), and re-cloned to the pEGFP-C3 vector at the EcoRI sites. The DNA sequences of all fusion plasmids were confirmed using an ABI 310 or ABI 3100 Genetic Analyser (Applied Biosystems, Norwalk, CT). Primer sequences are listed in supplementary material Table S1.

Fluorescence microscopy

Cells were grown in a glass-bottom culture dish (MatTech, Ashland, MA). GFP fusion plasmids (0.1-0.2 μg) were transfected into cells with LipofectaminePlus (Gibco BRL) according to the manufacturer's methods except that the incubation time with DNA was reduced to 1.5 hours.

Time-lapse, multicolor images in living cells, as well as indirect immunofluorescence images in fixed cells, were obtained using the DeltaVision microscope system based on an Olympus wide-field fluorescence microscope IX70 (Applied Precision, Seattle, WA). For temperature control during live observation, the microscope was kept at 37°C in a temperature-controlled room as described previously (Haraguchi et al., 1999). A series of three-dimensional image data were obtained, and deconvolved with the SoftWorx software equipped with the DeltaVision as described previously (Haraguchi et al., 2001; Haraguchi et al., 2004).

FRAP and FRET analyses

FRAP and FRET studies were done using a laser-scanning confocal microscope, Zeiss LSM510 META (Carl Zeiss, Yena) equipped with a Kr laser (Coherent, Santa Clara, CA) (Haraguchi et al., 2002; Hiraoka et al., 2002). FRAP experiments were carried out as described previously (Shimi et al., 2004). Briefly, HeLa cells stably expressing GFP fusion proteins were imaged using a water-immersion objective lens (C-Apochromat 40×/NA1.2) on a Zeiss LSM510 META. Photobleaching was performed with full power of 488 nm light from the argon laser 10 iterations for a defined region of each cell. Image data were collected through a pinhole of 2.37 Airy units (corresponding to an optical section 2 μm thick) with 488 nm argon laser at 0.5% full power.

For FRET experiments, living HeLa cells transiently expressing pairs of NE proteins fused with mCFP and mVenus were imaged using a water-immersion objective lens (C-Apochromat 40×/NA1.2) on a Zeiss LSM510 META. FRET was detected by acceptor photobleaching (Wouters et al., 2001). mVenus was photobleached by 514 nm light from the argon laser, with full power, 50 iterations for a defined area (2.1 μm square) in each cell. After photobleaching, time-lapse images were collected on the META detector from 470 nm to 545 nm through a pinhole of 2.18 Airy units (corresponding to an optical section 1.8 μm thick) by exciting with 413 nm light from the Kr laser at 0.3-0.9% full power. FRET efficiency was determined for fluorescence intensity of mCFP and mVenus after linear unmixing as described previously (Shimi et al., 2004). For time-lapse FRET ratio imaging, a fluorescence microscope IX70 (Olympus) equipped with a Nipkow disc confocal unit, CSU21 (Yokogawa, Japan) was used. Cells were illuminated with a 430 nm diode laser to excite CFP, and image data for CFP and YFP were simultaneously obtained on a CoolSnapHQ CCD camera (Photometrics, Roper, Tucson, Texas) through a dual viewer (Optical Insights), which separates emission light for wavelengths of 480 nm (corresponding to CFP) and 535 nm (corresponding to YFP). Ratio images between CFP and YFP were obtained by MetaMorph software (Universal Imaging Corporation, Downingtown, PA).

Correlative light electron microscope analysis after live observation (live CLEM)

Cells stably expressing GFP-BAF were cultured in a special glass-bottomed dish with an addressing grid (grid size, 175 μm) on the coverslip (Iwaki, Japan). After staining with Hoechst 33342, live cells were monitored on DeltaVision; at the desired time point after the metaphase-anaphase transition, cells were fixed with glutaraldehyde at a final concentration of 2.5% for 1 hour. Three-dimensional images (50-60 focal planes at 0.2 μm intervals) were obtained using an Olympus oil-immersion objective lens (PlanApo ×60/NA1.4) and computationally processed by three-dimensional deconvolution (Agard et al., 1989). Electron microscope (EM) observation of the same cell was carried out as follows. Cells were post-fixed with 1% OsO4 (Nisshin EM, Tokyo, Japan) in phosphate buffer, pH 7.4, for 1 hour, washed briefly with distilled water, and sequentially dehydrated with 50 and 70% ethanol. After staining with 2% uranyl acetate in 70% ethanol for 1 hour, cells were again sequentially dehydrated with 90% and 100% ethanol, and then embedded in epoxy resin by incubating with 50% (v/v) Epon812 (TAAB, Berkshire, UK) in ethanol for 30 minutes and 100% Epon812 for 1 hour. The epoxy block containing the same cell observed by a fluorescence microscope was trimmed according to the grid reference on the coverslip, and sliced to ultra-thin sections with a thickness of 80 nm using a microtome (Leica microsystems, Germany). Thin sections were stained with 2% uranyl acetate (Merck) for 1 hour and a commercial ready-to-use solution of lead citrate (Sigma) for 1 minute. Image data were collected using an electron microscope Hitachi H7600 (80 kV, Hitachi, Japan).

Knockdown by RNAi

Two to five micrograms of oligomeric double-stranded RNA for BAF, lamin A or emerin were transfected into HeLa cells cultured in a 35 mm culture dish using Oligofectamine (Invitrogen, Carlsbad, CA), Lipofectamine PLUS (Invitrogen), or RNAiFect (Qiagen) transfection reagents according to the manufacturer's instructions. The cells were incubated for 3 days in DMEM culture medium containing 10% calf serum.

We are grateful to the Riken Cell Bank for HeLa cells, A. Miyawaki and H. Nagai for mCFP and mVenus, R. Foisner for LAP2α-YFP, A. Gajewski for initial work on FRET between LAP2α and BAF, R. Goldman for GFP-lamin B1, P. Traktman for anti-phosphorylated BAF antibody, and N. Nakamura for anti-GFP antibody. Especially we thank K. L. Wilson for valuable discussion throughout this research. We also thank D. B. Alexander, M. Gorjánácz and I. W. Mattaj for critical reading of the manuscript. This work was supported by grants from the Japan Science and Technology Agency (to T.H.) and the Ministry of Education, Science and Sports of Japan (to T.H. and Y.H.).

Agard, D. A., Hiraoka, Y., Shaw, P. and Sedat, J. W. (
1989
). Fluorescence microscopy in three dimensions.
Methods Cell Biol.
30
,
353
-377.
Anderson, D. J. and Hetzer, M. W. (
2007
). Nuclear envelope formation by chromatin-mediated reorganization of the endoplasmic reticulum.
Nat. Cell Biol.
9
,
1160
-1166.
Anderson, D. J. and Hetzer, M. W. (
2008
). Shaping the endoplasmic reticulum into the nuclear envelope.
J. Cell Sci.
121
,
137
-142.
Bengtsson, L. and Wilson, K. L. (
2006
). Barrier-to-autointegration factor phosphorylation on Ser-4 regulates emerin binding to lamin A in vitro and emerin localization in vivo.
Mol. Biol. Cell
17
,
1154
-1163.
Bione, S., Maestrini, E., Rivella, S., Mancini, M., Regis, S., Romeo, G. and Toniolo, D. (
1994
). Identification of a novel X-linked gene responsible for Emery-Dreifuss muscular dystrophy.
Nat. Genet.
8
,
323
-327.
Bonne, G., Di Barletta, M. R., Varnous, S., Becane, H. M., Hammouda, E. H., Merlini, L., Muntoni, F., Greenberg, C. R., Gary, F., Urtizberea, J. A. et al. (
1999
). Mutations in the gene encoding lamin A/C cause autosomal dominant Emery-Dreifuss muscular dystrophy.
Nat. Genet.
21
,
285
-288.
Clements, L., Manilal, S., Love, D. R. and Morris, G. E. (
2000
). Direct interaction between emerin and lamin A.
Biochem. Biophys. Res. Commun.
267
,
709
-714.
Dechat, T., Gajewski, A., Korbei, B., Gerlich, D., Daigle, N., Haraguchi, T., Furukawa, K., Ellenberg, J. and Foisner, R. (
2004
). LAP2α and BAF transiently localize to telomeres and specific regions on chromatin during nuclear assembly.
J. Cell Sci.
117
,
6117
-6128.
Ellenberg, J., Siggia, E. D., Moreira, J. E., Smith, C. L., Presley, J. F., Worman, H. J. and Lippincott-Schwartz, J. (
1997
). Nuclear membrane dynamics and reassembly in living cells: targeting of an inner nuclear membrane protein in interphase and mitosis.
J. Cell Biol.
138
,
1193
-1206.
Ellis, J. A., Craxton, M., Yates, J. R. and Kendrick-Jones, J. (
1998
). Aberrant intracellular targeting and cell cycle-dependent phosphorylation of emerin contribute to the Emery-Dreifuss muscular dystrophy phenotype.
J. Cell Sci.
111
,
781
-792.
Eriksson, M., Brown, W. T., Gordon, L. B., Glynn, M. W., Singer, J., Scott, L., Erdos, M. R., Robbins, C. M., Moses, T. Y., Berglund, P. et al. (
2003
). Recurrent de novo point mutations in lamin A cause Hutchinson-Gilford progeria syndrome.
Nature
423
,
293
-298.
Furukawa, K. (
1999
). LAP2 binding protein 1 (L2BP1/BAF) is a candidate mediator of LAP2-chromatin interaction.
J. Cell Sci.
112
,
2485
-2492.
Gant, T. M. and Wilson, K. L. (
1997
). Nuclear assembly.
Annu. Rev. Cell Dev. Biol.
13
,
669
-695.
Gorjánácz, M., Klerkx, E. P. F., Galy, V., Santarella, R., López-Iglesias, C., Askjaer, P. and Mattaj, I. W. (
2007
). C. elegans BAF-1 and its kinase VRK-1 participate directly in postmitotic nuclear envelope assembly.
EMBO J.
26
,
132
-143.
Hachet, V., Kocher, T., Wilm, M. and Mattaj, I. W. (
2004
). Importin alpha associates with membranes and participates in nuclear envelope assembly in vitro.
EMBO J.
23
,
1526
-1535.
Haraguchi, T., Ding, D. Q., Yamamoto, A., Kaneda, T., Koujin, T. and Hiraoka, Y. (
1999
). Multiple-color fluorescence imaging of chromosomes and microtubules in living cells.
Cell Struct. Funct.
24
,
291
-298.
Haraguchi, T., Koujin, T., Hayakawa, T., Kaneda, T., Tsutsumi, C., Imamoto, N., Akazawa, C., Sukegawa, J., Yoneda, Y. and Hiraoka, Y. (
2000
). Live fluorescence imaging reveals early recruitment of emerin, LBR, RanBP2, and Nup153 to reforming functional nuclear envelopes.
J. Cell Sci.
113
,
779
-794.
Haraguchi, T., Koujin, T., Segura, M., Lee, K. L., Matsuoka, Y., Yoneda, Y., Wilson, K. L. and Hiraoka, Y. (
2001
). BAF is required for emerin assembly into the reforming nuclear envelope.
J. Cell Sci.
114
,
4575
-4585.
Haraguchi, T., Shimi, T., Koujin, T., Hashiguchi, N. and Hiraoka, Y. (
2002
). Spectral imaging fluorescence microscopy.
Genes Cells
7
,
881
-887.
Haraguchi, T., Holaska, J. M., Yamane, M., Koujin, T., Hashiguchi, N., Mori, C., Wilson, K. L. and Hiraoka, Y. (
2004
). Emerin binding to Btf, a death-promoting transcriptional repressor, is disrupted by a missense mutation that causes Emery-Dreifuss muscular dystrophy.
Eur. J. Biochem.
271
,
1035
-1045.
Haraguchi, T., Koujin, T., Osakada, H., Kojidani, T., Mori, C., Masuda, H. and Hiraoka, Y. (
2007
). Nuclear localization of barrier-to-autointegration factor is correlated with progression of S-phase in human cells.
J. Cell Sci.
120
,
1967
-1977.
Harel, A., Chan, R. C., Lachish-Zalait, A., Zimmerman, E., Elbaum, M. and Forbes, D. J. (
2003
). Importin beta negatively regulates nuclear membrane fusion and nuclear pore complex assembly.
Mol. Biol. Cell
14
,
4387
-4396.
Hetzer, M., Bilbao-Cortés, D., Walther, T. C., Gruss, O. J. and Mattaj, I. W. (
2000
). GTP hydrolysis by Ran is required for nuclear envelope assembly.
Mol. Cell
5
,
1013
-1024.
Hirano, Y., Segawa, M., Ouchi, F. S., Yamakawa, Y., Furukawa, K., Takeyasu, K. and Horigome, T. (
2005
). Dissociation of emerin from barrier-to-autointegration factor is regulated through mitotic phosphorylation of emerin in a xenopus egg cell-free system.
J. Biol. Chem.
280
,
39925
-39933.
Hiraoka, Y., Shimi, T. and Haraguchi, T. (
2002
). Multispectral imaging fluorescence microscopy for living cells.
Cell Struct. Funct.
27
,
367
-374.
Holaska, J. M., Lee, K. K., Kowalski, A. K. and Wilson, K. L. (
2003
). Transcriptional repressor germ cell-less (GCL) and barrier to autointegration factor (BAF) compete for binding to emerin in vitro.
J. Biol. Chem.
278
,
6969
-6975.
Jacque, J. M. and Stevenson, M. (
2006
). The inner-nuclear-envelope protein emerin regulates HIV-1 infectivity.
Nature
441
,
641
-645.
Lee, K. K., Haraguchi, T., Lee, R. S., Koujin, T., Hiraoka, Y. and Wilson, K. L. (
2001
). Distinct functional domains in emerin bind lamin A and DNA-bridging protein BAF.
J. Cell Sci.
114
,
4567
-4573.
Lee, M. S. and Craigie, R. (
1998
). A previously unidentified host protein protects retroviral DNA from autointegration.
Proc. Natl. Acad. Sci. USA
95
,
1528
-1533.
Liu, J., Lee, K. K., Segura-Totten, M., Neufeld, E., Wilson, K. L. and Gruenbaum, Y. (
2003
). MAN1 and emerin have overlapping function(s) essential for chromosome segregation and cell division in Caenorhabditis elegans.
Proc. Natl. Acad. Sci. USA.
100
,
4598
-4603.
Maeshima, K., Yahata, K., Sasaki, Y., Nakatomi, R., Tachibana, T., Hashikawa, T., Imamoto, F. and Imamoto, N. (
2006
). Cell-cycle-dependent dynamics of nuclear pores: pore-free islands and lamins.
J. Cell Sci.
119
,
4442
-4451.
Margalit, A., Segura-Totten, M., Gruenbaum, Y. and Wilson, K. L. (
2005
). Barrier-to-autointegration factor is required to segregate and enclose chromosomes within the nuclear envelope and assemble the nuclear lamina.
Proc. Natl. Acad. Sci. USA.
102
,
3290
-3295.
Mattout, A., Dechat, T., Adam, S. A., Goldman, R. D. and Gruenbaum, Y. (
2006
). Nuclear lamins, diseases and aging.
Curr. Opin. Cell Biol.
18
,
335
-341.
Mounkes, L. C., Kozlov, S., Hernandez, L., Sullivan, T. and Stewart, C. L. (
2003
). A progeroid syndrome in mice is caused by defects in A-type lamins.
Nature.
423
,
298
-301.
Nagano, A. and Arahata, K. (
2000
). Nuclear envelope proteins and associated diseases.
Curr. Opin. Neurol.
13
,
533
-539.
Nichols, R. J., Wiebe, M. S. and Traktman, P. (
2006
). The vaccinia-related kinases phosphorylate the N' terminus of BAF, regulating its interaction with DNA and its retention in the nucleus.
Mol. Biol. Cell
17
,
2451
-2464.
Segura-Totten, M., Kowalski, A. K., Craigie, R. and Wilson, K. L. (
2002
). Barrier-to-autointegration factor: major roles in chromatin decondensation and nuclear assembly.
J. Cell Biol.
158
,
475
-485.
Shimi, T., Koujin, T., Segura-Totten, M., Wilson, K. L., Haraguchi, T. and Hiraoka, Y. (
2004
). Dynamic interaction between BAF and emerin revealed by FRAP, FLIP, and FRET analyses in living HeLa cells.
J. Struct. Biol.
147
,
31
-41.
Shumaker, D. K., Lee, K. K., Tanhehco, Y. C., Craigie, R. and Wilson, K. L. (
2001
). LAP2 binds to BAF.DNA complexes: requirement for the LEM domain and modulation by variable regions.
EMBO J.
20
,
1754
-1764.
Ulbert, S., Platani, M., Boue, S. and Mattaj, I. W. (
2006
). Direct membrane protein-DNA interactions required early in nuclear envelope assembly.
J. Cell Biol.
173
,
469
-476.
Walther, T. C., Askjaer, P., Gentzel, M., Habermann, A., Griffiths, G., Wilm, M., Mattaj, I. W. and Hetzer, M. (
2003
). RanGTP mediates nuclear pore complex assembly.
Nature.
424
,
689
-694.
Wilson, K. L., Zastrow, M. and Lee, K. K. (
2001
). Lamins and disease: insights into nuclear infrastructure.
Cell
104
,
647
-650.
Worman, H. J. and Courvalin, J. C. (
2005
). Nuclear envelope, nuclear lamina, and inherited disease.
Int. Rev. Cytol.
246
,
231
-279.
Wouters, F. S., Verveer, P. J. and Bastiaens, P. I. (
2001
). Imaging biochemistry inside cells.
Trends Cell Biol.
11
,
203
-211.
Yang, L., Guan, T. and Gerace, L. (
1997
). Integral membrane proteins of the nuclear envelope are dispersed throughout the endoplasmic reticulum during mitosis.
J. Cell Biol.
137
,
1199
-1210.
Zacharias, D. A., Violin, J. D., Newton, A. C. and Tsien, R. Y. (
2002
). Partitioning of lipid-modified monomeric GFPs into membrane microdomains of live cells.
Science
296
,
913
-916.
Zhang, C. and Clarke, P. R. (
2000
). Chromatin-independent nuclear envelope assembly induced by Ran GTPase in Xenopus egg extracts.
Science
288
,
1429
-1432.
Zheng, R., Ghirlando, R., Lee, M. S., Mizuuchi, K., Krause, M. and Craigie, R. (
2000
). Barrier-to-autointegration factor (BAF) bridges DNA in a discrete, higher-order nucleoprotein complex.
Proc. Natl. Acad. Sci. USA.
97
,
8997
-9002.

Supplementary information