Actin- and protein-4.1-containing filaments link nuclear pore complexes to subnuclear organelles in Xenopus oocyte nuclei

Nuclear pore complexes (NPCs) are embedded in the nuclear envelope, where they mediate and regulate nucleocytoplasmic transport (reviewed by Fahrenkrog and Aebi, 2003; Vasu and Forbes, 2001). NPCs have eightfold structural symmetry and consist of 29 distinct nucleoporins, each of which is present in 8-32 or more copies (Cronshaw et al., 2002). The `basket' structure on the nucleoplasmic side of the NPC consists of eight filaments, which attach to a distal `ring' structure. Several reports suggest that these `rings' connect to other filaments that extend into the nucleus (Cordes et al., 1993; Cordes et al., 1997; Parfenov et al., 1995). Short NPC-linked filaments have been visualized by feSEM in Xenopus oocyte nuclei; these filaments had diameters of ∼50 nm and appeared to interconnect neighboring NPCs (Ris, 1997). Arlucea and colleagues (Arlucea et al., 1998) examined amphibian oocyte nuclei by transmission electron microscopy, and documented the three-dimensional morphology of eight thin filaments that extend from NPC baskets and form a hollow, open-sided tube. These filaments, and the nucleolus (Kneissel et al., 2001), are unstable in buffers with low Mg²⁺ (Arlucea et al., 1998).

The composition (and existence) of NPC-linked intranuclear filaments have been debated. Nucleoporins Nup98 (Powers et al., 1997), Nup153 (Panté et al., 1994) and Tpr (Hase and Cordes, 2003), all of which are NPC basket components (see Frosst et al., 2002), have been considered as candidates. By indirect immunofluorescence, Nup98 was found in a filamentous pattern throughout the interior of somatic mammalian nuclei (Fontoura et al., 2001), along with epitopes related to, but probably distinct from, Tpr (Hase and Cordes, 2003; Frosst et al., 2002; Zimowska and Paddy, 2002). The putative NPC-linked filaments were suggested to mediate nuclear export and intranuclear trafficking.

Actin is present and functional in the nucleus (Rando et al., 2000; Pederson and Aebi, 2002; Bettinger et al., 2004) with known roles that include chromatin remodeling, mRNA transcription and nuclear export (e.g. Scheer et al., 1984; Goodson and Hawse, 2002; Machesky and May, 2001). Nuclei also contain a specific isoform of myosin I, an actin-dependent motor. Antibodies against nuclear myosin I block transcription by RNA polymerase II when injected into mammalian cells, and also inhibit isolated transcription complexes in vitro (Pestic-Dragovich et al., 2000). The nature and composition of the track(s) and cargo(s) for myosin motors in the nucleus are important open questions, but are likely to involve actin. Somatic cells regulate the amount of actin that constantly enters the nucleus, at two levels: actin has a conserved nuclear export signal (Wada et al., 1998), and actin-profilin complexes are exported by a dedicated receptor, exportin-6 (Stuven et al., 2003). Actin is also required for the nuclear export of certain proteins (Hofmann et al., 2001). Xenopus oocyte nuclei contain 4-6 mg/ml actin (Clark and Merriam, 1977; Clark and Rosenbaum, 1979; Pederson and Aebi, 2002). Actin is also present in the nucleolus (Funaki et al., 1995; Clark and Merriam, 1977), and TEM analysis of Xenopus oocyte nuclei suggested that short bundles of actin extend from nucleoli towards the nuclear envelope (Parfenov et al., 1995). The most important questions about nuclear actin revolve around its polymeric state(s). Nuclear actin does not form long actin filaments (`F-actin') and does not stain with phalloidin (Bettinger et al., 2004). To avoid `fighting' chromatin, nuclear actin is proposed to assume shorter, potentially novel, forms (Pederson and Aebi, 2002; Bettinger et al., 2004). Interestingly cytoplasmic actin is known to form short `protomer' filaments (e.g. at branched intersections with protein 4.1, tropomyosin and spectrin), as well as tubes, sheets and short branched filaments (Pederson and Aebi, 2002).

Actin binds many structural proteins in the nucleus: the intermediate filament protein lamin A (Sasseville and Langelier, 1998), membrane protein emerin (J. M. Holaska, A. K. Kowalski and K.L.W., unpublished data), the nesprin family of filamentous proteins (Zhang et al., 2002b; Zhen et al., 2002) and nuclear-specific isoforms of protein `4.1', an actin-scaffolding protein (Correas, 1991; Krauss et al., 1997; Luque and Correas, 2000). Nuclear assembly in Xenopus egg extracts is disrupted by depletion of protein 4.1, or by exogenous protein 4.1 fragments that bind spectrin/actin, suggesting structural roles for both actin and nuclear 4.1 protein during nuclear assembly (Krauss et al., 2002; Krauss et al., 2003; Bettinger et al., 2004).

Imaging the interior of the nucleus is technically challenging, given that there are rapidly mobile components (Phair and Misteli, 2000; Spector, 2001), stable structural elements (Aebi et al., 1986; Nickerson, 2001; Daigle et al., 2001) and chromosomes, which obscure all other structures and thwart direct visualization. To investigate the structure of putative NPC-linked filaments, we used stage 3-4 Xenopus oocyte nuclei, which contain proportionally little chromatin, yet are quite large and highly active for transcription (Gall et al., 1999) and nucleocytoplasmic transport (see Feldherr and Akin, 1997). Thus, interior structures such as NPC `baskets' and associated filaments can be viewed three-dimensionally by feSEM, without obstruction by chromatin. Our results suggest that nuclear protein 4.1 and actin are components of a stable network of PLFs that attach to Cajal bodies and other subnuclear organelles and ramify throughout the interior of the nucleus.

Female Xenopus laevis were purchased from the UK Xenopus facility (Blades Biological) and maintained in the laboratory under conditions optimal for health. Ovaries were surgically removed and placed in amphibian Ringer's solution [111 mM NaCl, 1.9 mM KCl, 1.1 mM CaCl₂, 2.4 mM NaHCO₃ (Macgregor and Varley, 1983)]. Small pieces of ovary were placed in Petri dishes in buffer A (83 mM KCl, 17 mM NaCl, 10 mM Hepes, 250 mM sucrose, 3 mM (or 0.3 mM) MgCl₂, pH 7.4). Silicon chips pre-coated with poly-L-lysine were placed near the oocytes in the same dish. All solutions were kept ice-cold.

Xenopus oocytes in developmental stage 3-4 were identified as being 480-500 μm diameter and light brown to brown. PLFs were present in nuclei from oocyte stages 3-4, 5 and 6. We used stage 3-4 because they are highly active for transcription and because nucleus isolation was easier and quicker than at stage 6, perhaps because stage 3-4 oocytes have fewer yolk granules. The nucleus was manually isolated from each oocyte using glass needles, and gently and quickly separated (released) from surrounding cytoplasm. Two to three isolated nuclei were transferred to a nearby silicon chip, and their nuclear envelopes carefully opened manually using the glass needle. For nuclear content `spreads' (Fig. 1), nucleoplasm was pushed slightly using the glass needle and carefully spread over the surface of the chip. To examine intact filaments under gentle conditions, each nucleus was first released from the oocyte and transferred to the chip. The nuclear envelope was stripped carefully away from the underlying gelated nucleoplasm using watchmaker's forceps. The intact demembranated ball-like nuclear gel was fixed within 2-3 seconds after isolation by adding fix solution (see below) directly to the chip in the Petri dish, and then transferring the chip to another Petri dish with fresh fix.

For both `spread' and `intact' samples, the chips were fixed and treated as follows, using ice-cold solutions. Chips were transferred from dish to dish without drying, by keeping a drop of the previous solution on the chip. Samples were fixed for 10 minutes in 2% glutaraldehyde, 0.2% tannic acid, 10 mM Tris-HCl (pH 7.4) and 3 mM MgCl₂. Importantly, the elapsed time from nuclear isolation to placing in fix was always 5-10 seconds or less. Fixed samples were postfixed for 30 minutes in 1% OsO₄, washed in double distilled H₂O, stained 20 minutes in 1% uranyl acetate, dehydrated in ethanol and critical point dried from CO₂ using Arklone (Arcton 112, ICI Chemicals, Runcorn, UK) as the transitional solvent. Specimens were then sputter coated with a 4 nm layer of chromium using an Edwards Auto 306 coating unit (UK) with a cryo-pumped vacuum system, and examined at 30 kV in a DS 130F feSEM (Japan).

Latrunculin A (Calbiochem Corp., La Jolla CA USA) and jasplakinolide (Molecular Probes, Eugene OR) were each dissolved in 100% DMSO at a concentration of 250 μg/ml, and stored at –20°C. Intact oocytes were incubated in amphibian Ringer's solution with 2 μg/ml of latrunculin A/0.8% DMSO for 2 hours at room temperature, or overnight at 4°C. Alternatively, oocytes were incubated in 1 μM jasplakinolide/0.4% DMSO for 2 hours at 22-24°C. Control oocytes were incubated in the corresponding amount of DMSO alone (0.8% and 0.4%, respectively). After drug treatment, oocytes were washed in amphibian Ringer's, solution transferred into buffer A (see `Xenopus oocyte nuclei', above), and prepared and fixed as described for feSEM imaging.

Rabbit antibodies against the C-terminal domain of actin were purchased from Sigma Chemical Corp. (catalog number A-2066). Rabbit antibodies against coilin were kindly provided by Prof. Joe Gall (Carnegie Institute, Baltimore, USA) and Prof. A. Lamond (Wellcome Trust Biocentre, University of Dundee, UK). Rabbit antibody against protein 4.1 was purchased from Biogenesis (Poole, Dorset UK; catalog number N-7838-5010) and used at 1:20 dilution for immuno-gold labeling and 1:100 dilution for western blots. For immuno-gold localization, nuclei were manually isolated from stage 3-4 Xenopus oocytes using glass needles, transferred to a silicon chip and nuclear envelopes opened manually as described above. Nuclei were then fixed for 15 minutes at 22-24°C in 3.7% formaldehyde, 3 mM MgCl₂, 0.2% tannic acid and 10 mM Tris-HCl pH 7.0. Fixed samples were washed three times (1 minutes each) in PBS, incubated in PBS containing 1% BSA for 30 minutes, washed in PBS and incubated for 1-3 hours with primary antibody against either actin (diluted 1:100) or protein 4.1 (diluted 1:20) in PBS. Primary antibodies were removed by washing three times in PBS, and samples were incubated for 1 hour with 10 nm gold-conjugated secondary goat anti-rabbit antibody (Amersham Corp.). As negative controls, we used gold-conjugated secondary antibody diluted 1:20 in PBS. All samples were then post-fixed 30 minutes in 1% OsO₄, washed in double distilled H₂O, stained for 20 minutes in 1% uranyl acetate, ethanol dehydrated and critical point dried from CO₂ using Arclone as the transitional solvent. Specimens were finally sputter coated with a 4 nm layer of chromium for feSEM, as described above. Colloidal gold particles were visualized using a solid state backscatter electron detector (UK). For immunoblotting, oocyte nuclei were manually isolated and at least 10 nuclei per lane were resolved by SDS-PAGE on 10% pre-cast Bis-Tris gels (Novex), transferred to PVDF membrane, blocked for 1 hour at 22-24°C with 5% dried milk (in PBS plus 0.1% Tween 20; PBST), washed in PBST three times (5 minutes each) and incubated with anti-actin, anti-coilin or anti-protein-4.1 antibodies for 1 hour at 22-24°C. Blots were then washed, incubated for 1 hour at 22-24°C with sheep anti-rabbit secondary antibodies conjugated to horseradish peroxidase (DAKO A/S, Denmark; diluted 1:10,000 or 1:2000), washed again and visualized by enhanced chemiluminescence (ECL; Amersham Biosciences).

We measured 300 filaments from 27 randomly chosen negatives of five different oocyte nuclei. For spherical bodies (nucleoli and Cajal bodies), we measured the diameter of 200 bodies in 12 randomly chosen images from three different oocyte nuclei. To estimate the spacing of gold-labeled protein 4.1 epitopes, we measured the distances between a total of 23 pairs of gold particles from four non-overlapping images of individual ∼40-nm-thick PLFs.

To image their interior structure, nuclei were first isolated gently and quickly from the oocyte, opened by peeling off a small section of nuclear envelope and fixed in buffer containing 250 mM sucrose and 3 mM MgCl₂ within 5-10 second after isolation (see Materials and Methods). The nuclear contents consisted of small spherical bodies and thick fibers of ∼300 nm diameter (Fig. 1A,B). These 300-nm fibers were not examined further in situ, but are assumed to represent aggregates of thinner (12-120 nm; see below) filaments that were visualized after nuclear contents were spread manually (Fig. 1C-E). After manual spreading, some filaments remained attached to the nuclear envelope (Fig. 1D; see also Figs 5, 6). In samples that were deliberately subjected to moderate (Fig. 1C) or strong (Fig. 1E) spreading forces during isolation, many filaments remained attached to spherical bodies, which retained their structure during spreading (Fig. 1C-E). These spheres were heterogenous in size, but their average diameter (500-700 nm) and range of diameters (100 nm to ∼5 μm; Fig. 1F) corresponded to three well-studied subnuclear organelles: Cajal bodies, nucleoli and snurposomes. In Xenopus oocyte nuclei, Cajal bodies are 0.1 to 3 μm in diameter and are present in only 50-100 copies per nucleus (Morgan et al., 2000). Using antibodies specific for the 80 kDa coilin protein (Fig. 1G), and controls lacking primary antibody (Fig. 1I), a subset of the spherical bodies was confirmed as Cajal bodies in immuno-gold labeling experiments (Fig. 1H). Cajal bodies are proposed to be assembly sites for polymerase-containing `transcriptosome' complexes (Gall et al., 1999). The majority of spheres were coilin negative, and were probably nucleoli and snurposomes based on their morphology, abundance and size distribution (diameters of 0.5-10 μm; Fig. 1F). Each oocyte nucleus contains ∼1500 small nucleoli, which assemble around amplified extrachromosomal rDNA genes (Wu and Gall, 1997; Mais and Scheer, 2001). Oocyte nuclei also contain hundreds to thousands of snurposomes (`speckles'), which are 1-4 μm in diameter and contain RNA splicing factors (Callan and Gall, 1991; Wu et al., 1991). However, other types of subnuclear organelles might also be present.

At higher magnification, filaments of 100 nm (or more) diameter were seen to embed directly into the spheres (Fig. 2). The width and surface morphology of filaments were irregular (Fig. 2B), suggesting that their architecture or associated molecules (or both) were heterogeneous. Meiotic chromatin was absent from many images. When chromatin was detected, it was often associated with small spherical bodies and filaments (Fig. 3A,B). However, we frequently observed chromatin without any associated filaments (data not shown), suggesting that either filaments attached to chromatin weakly, or that the filament network was independent of chromatin.

We detected filaments near the inner membrane and many also extended deep into the nuclear interior (Figs 4, 5). Filaments were typically 30-50 nm in diameter, but ranged from 10 nm to 120 nm (Fig. 4A,B). The smallest filaments, ∼12 nm in diameter, were detected less frequently (Fig. 4B) and were deduced to be more fragile. These ∼12 nm filaments appeared to form `bridges' between larger filaments and also connected some spherical bodies to the inner membrane (Fig. 4C, arrows). Many ∼40 nm filaments attached directly to NPC baskets (Fig. 5A; each white circle encloses 1-3 NPCs). Filaments attached to a single NPC are shown at higher magnification in Fig. 5B (upper and lower panels) and individual NPC basket filaments are indicated by dotted lines in Fig. 6D. Some filaments extended over 1 μm into the nucleus (e.g. Fig. 5D), consistent with TEM-based reports of long filaments in Xenopus oocyte and mammalian somatic nuclei (Arlucea et al., 1998; Cordes et al., 1997). We concluded that our feSEM images of Xenopus oocyte nuclei revealed a three-dimensional network of NPC-linked filaments that varied in width, but had an average thickness of ∼40 nm.

Inside the nucleus, NPC-linked filaments reached lengths of 4 μm or more (Fig. 4B, Fig. 5D), and appeared to be strikingly adaptable as seen by extensive branching, bundling, cross-bridging and merging at junctions involving as many as 3-7 filaments (Fig. 4B,C, Fig. 5D, arrows). Interestingly, some, but not all, ∼40 nm filaments had a structural element that repeated at ∼30 nm intervals (Fig. 5C, arrows). The molecular basis for this repeat element was unknown. Based on the relative mechanical stability of this PLF network during manual stretching we speculated the presence of either nuclear intermediate filament proteins (lamins) or nuclear actin. Lamins were provisionally ruled out by immuno-gold staining; a monoclonal antibody against Xenopus lamin B3 recognized abundant epitopes at the nuclear inner membrane, but no signal on PLFs (E.K., M.W.G. and T.D.A., unpublished observations).

To determine if nuclear actin associated with the PLF network, we incubated oocyte nuclei with (or without) antibodies against actin, followed by gold-conjugated secondary antibodies. Controls showed that these antibodies recognized one major 42 kDa protein in immunoblots of oocyte nuclei, as expected for actin (Fig. 6A). We imaged samples by feSEM plus backscatter detection to determine the position of the gold particles three-dimensionally (see Materials and Methods). Actin was abundant on the PLF network (Fig. 6B,C); staining was specific because very few gold particles were detected in control nuclei treated with secondary antibody alone (Fig. 6E; arrow indicates a single gold particle). Though abundant, the actin signal was discontinuous along any given PLF; this could be explained trivially (only a subset of molecules are detected by immuno-gold labeling), or it might suggest the presence of other structural proteins. Actin-gold epitopes were not present on NPC basket filaments (dotted white lines in Fig. 6D), but were detected on structures immediately distal to basket filaments (Fig. 6D, arrow), which might be broken PLFs. These results suggested that actin was abundant on the PLF network. However, we were skeptical of this result because oocyte nuclei contain large amounts of actin (see Introduction), and its apparent association with PLFs might be non-specific.

To determine if the integrity or morphology of PLFs depended on actin, intact Xenopus oocytes were incubated for 2 hours (or overnight) in 2 μg/ml latrunculin A, which sequesters G-actin, leading to actin filament depolymerization (Spector et al., 1999). Control oocytes were incubated with an equivalent concentration of solvent (DMSO) alone (see Materials and Methods). The latrunculin A-treated and control nuclei were gently opened, fixed and imaged by feSEM (Fig. 7). Control nuclei had typical intranuclear filaments and spherical bodies, as expected (Fig. 7A). In contrast, latrunculin-treated nuclei had few filaments, and the spherical bodies were shrouded by amorphous material (Fig. 7B). Thus, the integrity of the PLF network may depend on actin polymers or actin turnover.

To disrupt actin dynamics, intact oocytes were incubated for 2 hours in 0.4% DMSO plus 1 μM jasplakinolide, which stabilizes pre-existing actin filaments (Lee et al., 1998) and can also induce actin polymerization (Spector et al., 1999). Control nuclei incubated in solvent only (0.4% DMSO) appeared normal (data not shown, see Fig. 7A). Jasplakinolide had no effect on the organization of the PLF network per se (Fig. 8A), but changed the morphology of individual filaments. After exposure to jasplakinolide, PLFs lost their `solid' appearance and had a delicate filigree morphology, including short (∼40 nm) filaments that projected radially from the long axis of each filament (Fig. 8B,C). These short filaments were ∼12 nm thick and were arranged in parallel rows ∼30 nm apart. Some rows (Fig. 8C) were spaced and angled in a manner that might correspond to the structural `repeat' element noted in Fig. 5. After subtracting the calculated thickness (4 nm, all sides) of the metal coat, we estimate that these axial filaments were at least 4-5 nm thick, and possibly thicker (e.g. 8 nm) if the actual metal coat was thinner than calculated. Thus, these ∼40-nm-long jasplakinolide-stabilized structures might be actin filaments. Interestingly the ∼30 nm spacing of the `repeat' element seen in untreated PLFs (e.g. Fig. 5C) might correlate with the ∼30 nm spacing between `rows' of jasplakinolide-induced short axial structures (e.g. Fig. 8C). Because jasplakinolide has many effects on actin, and latrunculin might have unanticipated effects on PLFs, these findings must be interpreted cautiously. Nevertheless, the jasplakinolide and latrunculin studies suggest that actin polymers, and potentially actin turnover, might contribute to the integrity of the PLF network.

To investigate further, we considered the actin-scaffolding protein 4.1, which is required for nuclear assembly in Xenopus egg extracts (Krauss et al., 1997; Krauss et al., 2002), and has properties consistent with a nuclear `matrix' component (De Carcer et al., 1995). Given our immuno-gold actin localization and inhibitor studies, we tested the hypothesis that nuclear protein 4.1 might be a component of the PLF network. Antibodies against protein 4.1 recognized a major ∼80 kDa band in immunoblots of isolated Xenopus oocyte nuclei (Fig. 9A), and minor bands visible only on longer exposures (data not shown). Different-migrating forms of protein 4.1 are prominent in Xenopus eggs (Krauss et al., 2002), potentially the result of cell cycle and developmental differences between oocytes (early meiosis) and egg extracts [post-`fertilization' interphase (see Lohka and Masui, 1983; Wu et al., 1991)]. FeSEM imaging of immuno-gold-labeled Xenopus oocyte nuclei revealed protein 4.1 epitopes on PLFs (Fig. 9B,C). Protein 4.1 epitopes were also highly abundant on a subset of unidentified spherical organelles (data not shown). Along individual PLFs, many 4.1 epitopes were spaced at regular intervals of ∼120 nm (±10 nm) (Fig. 9B,C, and data not shown). Where two PLFs merged or forked, we also frequently saw a pair of protein 4.1 epitopes separated by ∼60-80 nm (data not shown). We concluded that protein 4.1 is present in the PLF network. Moreover, its regular spacing raised the possibility that this actin-scaffolding protein is a structural component of the PLF network.

Our results provide evidence for a three-dimensional network of PLFs that extend throughout the nucleus, and attach directly to subnuclear organelles. These PLFs contain actin, and inhibitor studies suggest that actin polymers of undetermined length and undetermined conformation are required for the integrity of the PLF network. Interestingly, PLFs also contained regularly spaced protein 4.1 epitopes, suggesting that protein 4.1 is a structural component of the PLF network. These results from immuno-gold imaging strongly support and extend recent evidence that protein 4.1 and actin epitopes localize on filamentous structures such as those previously visualized by whole-mount electron microscopy of human fibroblast nuclei (Krauss et al., 2003). Thus, we speculate that a PLF network might also exist in somatic metazoan nuclei.

The PLF network had several striking features, including its stability during mechanical stretching, its apparent ability to fork and branch forming an extensive network inside the nucleus, its direct attachments to NPCs (at one end) and subnuclear organelles (at the other), and its dependence on actin polymerization. PLFs were typically 40 nm in diameter, but ranged from 12-120 nm. The actual diameter of the apparent 12-nm-diameter filaments can be estimated by subtracting the depth of chromium metal coating these structures. If the metal added 4 nm on all sides, the true diameter of the thinnest PLF-associated filaments would be 4-5 nm, too thin to be actin filaments. However, if the actual metal coat was thinner (e.g. 1-2 nm) the protein filaments would be 8-10 nm in diameter, consistent with actin filaments (8-9 nm diameter) (Schoenenberger et al., 1999; Dos Remedios et al., 2003). We favor models in which the thinnest PLF-associated filaments are actin filaments.

Our current data are insufficient to conclude that actin forms the `backbone' structure of ∼40-nm and thicker PLFs, because the actin-gold staining was distributed irregularly on PLFs. In contrast protein 4.1, spaced regularly along PLFs, is a plausible `backbone' candidate.

The composition of the PLF network is an open question. We suggest two classes of PLF-associated proteins: structural components (including but not limited to actin and protein 4.1) and transiently associated proteins. Hypothetical actin-relevant structural candidates might include EAST and NUANCE. The EAST protein localizes to `interchromosomal domain' spaces in Drosophila nuclei. When overexpressed, EAST expands these spaces and recruits actin (Wasser and Chia, 2000). NUANCE is a 796 kDa actin-binding protein found at the nuclear envelope, nuclear interior and nucleoli of human cells (Zhen et al., 2002; Bettinger et al., 2004). Previously proposed components of intranuclear filaments include an unidentified `Tpr-related protein' (but not Tpr itself) and nucleoporins Nup98 and Nup153 (Hase and Cordes, 2003) (see Frosst et al., 2002). Preliminary immuno-gold labeling studies suggest that PLFs are specifically labeled by antibodies against a Tpr-related epitope, but not by antibodies against Nup153 (E.K. and T.D.A., unpublished observations). The future challenge is to distinguish `backbone' from `cargo', if PLFs do indeed mediate diffusional transport in the nucleus.

Previous investigators hypothesized that NPC-linked filaments might facilitate intranuclear diffusion and nuclear export (Fontoura et al., 2001; Arlucea et al., 1998; Paddy, 1998). Soluble nuclear proteins and mRNAs have high rates of diffusion in the inter-chromatin space (Phair and Misteli, 2000; Politz and Pederson, 2000). This diffusional `space' appears to have boundaries (e.g. Wasser and Chia, 2000; Muratani et al., 2002; Reichenzeller et al., 2000). It is not known how the interchromatin space is maintained or protected from occlusion by chromatin. Arlucea et al. proposed that groups of eight pore-linked filaments form hollow open-sided cages that define and protect the diffusional space leading to each NPC (Arlucea et al., 1998). The `solid' appearance of our PLFs does not conflict with this `hollow cage' model, for two reasons. First, our samples were coated with a ∼4 nm layer of metal, which might clog the open spaces between the individual subfilaments that appear to collectively form each 12-120 nm PLF. Second, we imaged nuclei with no washing and minimal disruption of intranuclear structure. Our PLFs therefore might be `filled' with associated endogenous proteins, potentially including diffusing `cargo' and mobile nuclear proteins, which would also become metal-coated during sample preparation. Interestingly, PLFs imaged after treatment with jasplakinolide, which stabilizes pre-existing actin filaments, had the most `open' architecture, as if many PLF-associated components (cargo?) had been lost.

The nucleolus and Cajal bodies are thought to be trafficking `hubs' in the nucleus (Grosshans et al., 2001; Politz et al., 2000; Morgan et al., 2000), and there is evidence that traffic to and from the nucleolus might involve `tracks' of unknown composition (Meier and Blobel, 1992). Our three-dimensional images suggest that PLFs might provide such `highways' or `tracks' for diffusional movement, at least in Xenopus oocyte nuclei, because the PLF network linked major subnuclear organelles to each other and to NPCs. However, we further propose that PLFs might also create or maintain chromatin-free diffusion zones in the nucleus, as discussed below.

In vertebrate cells the nuclear envelope and NPCs disassemble during mitosis, and must reassemble, beginning in late anaphase and telophase (Gant and Wilson, 1997). Nuclear membranes rapidly target and enclose the telophase chromosomes. Similarly, NPCs reassemble rapidly and asynchronously in the nascent envelope (e.g. Haraguchi et al., 2000; Wiese et al., 1997; Ellenberg et al., 1997). All subsequent steps in nuclear assembly – including chromatin decondensation and the assembly of nucleoli and other organelles – depend on NPC-mediated nucleocytoplasmic transport (Wiese et al., 1997; Goldberg et al., 1997). NPC assembly requires fusion between the inner and outer membranes to create the hole [`pore' (see Drummond and Wilson, 2002; Cohen et al., 2003; Vasu and Forbes, 2001)]. Membrane fusion appears to be triggered by key soluble nucleoporins (Harel et al., 2003; Walter et al., 2003) and is followed by an ordered pathway in which many additional nucleoporins self-assemble to form the NPC (Macaulay and Forbes, 1996; Vasu and Forbes, 2001; Kiseleva et al., 2001; Goldberg et al., 1997).

What happens after the new envelope first encloses chromosomes is a mystery, particularly for newly-assembled NPCs, which face a wall of mitotically condensed telophase chromatin (Wiese et al., 1997; Goldberg et al., 1997; Vasu and Forbes, 2001). Similarly, NPCs that assemble during interphase may have a mechanism to gain active access to `interchromatin' diffusional space. Based on a TEM study of NPC assembly in mammalian cells, Maul and colleagues (Maul et al., 1971) deduced that chromatin is actively displaced from the region under each pore, during NPC assembly. In yeast cells, chromatin is actively excluded from regions near NPCs (Ishii et al., 2002). When overexpressed in yeast, a Tpr-related nucleoporin named Mlp1p forms large intranuclear structures that exclude chromatin but are permeable to proteins (Strambio-de-Castillia et al., 1999). Thus, NPC assembly might involve an active mechanism to remodel or displace the underlying chromatin. We propose that the assembly of a protein 4.1-and-actin-containing PLF network could provide a mechanism for building NPC-linked `diffusion channels' through chromatin. Consistent with this model, the actin-binding activity of protein 4.1 is required for nascent assembling nuclei to enlarge (Krauss et al., 2002). We speculate that PLF assembly might originate at NPC baskets and terminate at subnuclear organelles, potentially creating a `smart' network that points towards, and away from, NPCs. A `smart' PLF network might contribute to the rapidity of NES-mediated nuclear export of gold particles in Xenopus oocyte nuclei (Feldherr and Akin, 1997).

Our latrunculin results suggest actin polymers and actin turnover are somehow required for the integrity of the PLF network. Actin is an ATP-binding protein that can hydrolyze ATP when polymerized (Bettinger et al., 2004); assuming that PLFs facilitate diffusional movement in the nucleus, our results might explain, in part, the need for energy during intranuclear movement (Carmo-Fonseca et al., 2002). The apparent conformational flexibility of the PLF network, and intriguing morphology of `stabilized-actin' (jasplakinolide-treated) PLFs with their axially projecting ∼40-nm-long filaments support the idea that actin might form flexible and potentially novel polymers in the nucleus (Pederson and Aebi, 2002). Finally, our evidence for an actin-and-protein-4.1-scaffolded PLF network might explain, in part, why actin is required for both the nuclear export of unspliced HIV-1 genomic RNA (Kimura et al., 2000) and the export of mRNAs encoded by HIV-1 (Hofman et al., 2001; Percipalle et al., 2002). Actin is also required to export endogenous hnRNP complexes from the nucleus (Percipalle et al., 2001; Percipalle et al., 2002; Zhang et al., 2002a) (reviewed by Pederson and Aebi, 2002; Bettinger et al., 2004). Further analysis of the roles of protein 4.1 and actin in the oocyte PLF network, and the identification of `backbone' components of PLFs, are important goals for future work. If PLFs also exist in somatic nuclei, this work will shed `three-dimensional' light on nuclear architecture and transport mechanisms.

We are indebted to Thoru Pederson and Ueli Aebi for their insightful and optimistic review of nuclear actin. We thank Joe Gall (Carnegie Institute of Washington, Baltimore MD, USA) and Angus Lamond (Wellcome Trust Biocentre, University of Dundee, Scotland, UK) for anti-coilin antibodies and interesting discussions. We thank Michael Matunis and Pam Silver for helpful discussions and Steven Bagley for fluorescence imaging. This work was funded by a Wellcome Trust grant and a Russian Foundation for Basic Research grant (to E.K.), and grants from the National Institutes of Health USA (GM48646 to K.L.W.) and Cancer Research UK (to T.D.A.).