A dominant-negative form of POM121 binds chromatin and disrupts the two separate modes of nuclear pore assembly

The nuclear pore complex (NPC) is an intricate biological machine that mediates bidirectional transport of macromolecules across the nuclear envelope (NE) (Fahrenkrog et al., 2004; Gerace and Burke, 1988; Hetzer and Wente, 2009; Weis, 2003). NPCs are embedded at the junction of the two nuclear membranes and are composed of multiple copies of ~30 different proteins termed nucleoporins (Nups) (Alber et al., 2007; Cronshaw et al., 2002; Vasu and Forbes, 2001). Open mitosis in higher eukaryotes entails the synchronous assembly of NPCs and the nuclear envelope (NE) in a tightly regulated, stepwise process, which is coordinated with other mitotic exit events (Burke and Ellenberg, 2002; Guttinger et al., 2009; Wozniak and Clarke, 2003). NPCs also form at a separate stage of the metazoan cell cycle, by insertion into the intact NE during interphase (D'Angelo et al., 2006). This second mode of NPC formation enables metazoan cells to double NPC numbers during interphase (Maul et al., 1972) and more closely resembles the assembly pathway of lower eukaryotes, which undergo a closed mitosis (Fernandez-Martinez and Rout, 2009; Ryan et al., 2007).

The individual subunits that make up the massive structure of the NPC comprise three to four integral membrane Nups and 12–15 distinct subcomplexes of soluble Nups in different eukaryotes (Alber et al., 2007; Chadrin et al., 2010; Hetzer and Wente, 2009; Rasala et al., 2008; Schwartz, 2005). The mechanistic details and sequence of events in NPC assembly have been a matter of controversy and debate. Although NPC biogenesis is often described as a self assembly process, there is also evidence for the involvement of additional assembly factors, such as importin-β (Harel et al., 2003a; Ryan et al., 2007; Walther et al., 2003). The Nup107–160 complex is known to bind to chromatin through the adaptor protein ELYS and form the initial seeding sites, or ‘prepores’, for postmitotic NPC assembly (Franz et al., 2007; Gillespie et al., 2007; Rasala et al., 2006; Rasala et al., 2008; Rotem et al., 2009), but subsequent steps remain obscure. Moreover, extensive redundancy has been suggested to exist in the roles of specific proteins in assembly, especially in the case of integral membrane Nups (Stavru et al., 2006a).

Integral membrane Nups are thought to provide stable anchoring points for the mature NPC within the specialized, highly curved pore-membrane domain. In addition, these membrane nucleoporins have been suggested to play specific roles in NPC assembly, such as recruiting soluble Nups to the initial assembly sites and triggering fusion of the inner and outer nuclear membrane (Antonin et al., 2005; Drummond and Wilson, 2002; Fichtman et al., 2010; Macaulay and Forbes, 1996; Mitchell et al., 2010; Stavru et al., 2006b). A persistent problem has been the lack of overall phylogenetic conservation in sequence and topology of the integral membrane Nups. NDC1, which has six membrane-spanning segments, is the best conserved membrane Nup that is present in yeast and metazoa, but apparently absent from protozoa (Mansfeld et al., 2006; Stavru et al., 2006a). In yeast, NDC1 is a component of both the NPC and the spindle pole body (SPB), and fulfils the essential function of enabling the insertion of newly formed SPBs into the intact NE (Chial et al., 1998; Lau et al., 2004). gp210 is a type I transmembrane protein with a large luminal domain located between the inner and outer nuclear membranes, a single membrane-spanning segment and a short cytoplasmic tail (Gerace et al., 1982; Wozniak et al., 1989). This protein is evolutionarily conserved in metazoa and found in some protozoa and plants, but is missing in fungi (Stavru et al., 2006b). POM121 is the least conserved of the three vertebrate membrane Nups. It has the opposite topology to gp210, with a single membrane-spanning segment close to its N-terminus and the bulk of the protein facing the NPC channel. POM121 orthologues have only been identified in vertebrates (Antonin et al., 2005; Funakoshi et al., 2007; Hallberg et al., 1993).

There is conflicting evidence regarding the vital function for each of the three vertebrate membrane Nups in NPC assembly (Antonin et al., 2005; Cohen et al., 2003; Doucet et al., 2010; Drummond and Wilson, 2002; Funakoshi et al., 2011; Mansfeld et al., 2006; Stavru et al., 2006a; Stavru et al., 2006b). Most recently, POM121 has been proposed to play a crucial, rate-limiting role in interphase NPC assembly but not in the postmitotic process (Doucet et al., 2010). This contrasts with a previous suggestion that POM121 functions as part of a postmitotic checkpoint mechanism that coordinates NPC assembly with the formation of a fully enclosed NE (Antonin et al., 2005). Multiple inner nuclear membrane proteins have been suggested to bind directly to chromatin and drive membrane recruitment by collaborative and possibly redundant interactions (Anderson et al., 2009; Ulbert et al., 2006). In addition, reticulons and other membrane-bending and deforming proteins play a crucial role in ER expansion into an enclosed NE (Anderson and Hetzer, 2008), and might also be required for the insertion of NPCs into membranes in the de novo (interphase) mode of assembly (Dawson et al., 2009).

Collectively, the experimental results from different systems and partially conflicting reports indicate that both modes of NPC assembly are intricate, multi-step processes involving many components. As pointed out by Stavru et al. redundancies and built-in back up mechanisms can easily obscure the function of a given protein in such complicated settings (Stavru et al., 2006a). This is particularly relevant for experiments relying on RNAi-mediated knockdown of specific Nups in mammalian cells, or immunodepletion in cell-free reconstitution systems.

We sought an alternative strategy to study the role of POM121 in nuclear assembly. We reasoned that introducing a soluble fragment, lacking the transmembrane segment of POM121, into an in vitro assembly system might disrupt crucial interactions of the endogenous protein and help to reveal its functions. Here we describe such dominant-negative interactions of a soluble POM121 domain, affecting both the postmitotic and interphase modes of NPC assembly.

Cell-free systems derived from amphibian egg extracts have long served as an experimental model to study postmitotic NPC assembly (Antonin et al., 2005; Forbes et al., 1983; Harel et al., 2003b; Lohka and Masui, 1983; Rotem et al., 2009; Walther et al., 2003) and can also be used to follow NPC insertion into intact membranes (D'Angelo et al., 2006; Harel et al., 2003a; Vollmar et al., 2009). To test for a potential dominant-negative effect in the nuclear reconstitution assay, we expressed several fragments of Xenopus laevis POM121, lacking the transmembrane segment, as recombinant GST fusion proteins in bacteria (supplementary material Fig. S1A). Purified, soluble fragments were introduced into the full assembly system, containing chromatin, cytosol and membrane components. This analysis focused our attention on one central fragment, encompassing amino acids 143–415, which had a strong inhibitory effect on nuclear assembly. This region shows 41% identity to rat POM121 and contains an internal sequence of ~35 more highly conserved residues (Fig. 1A) with a recently characterized bipartite nuclear localization signal (NLS) (Doucet et al., 2010). The purification and characterization of the soluble GST-POM121^143–415 fusion protein is described in supplementary material Fig. S1B,C. We refer to this soluble fragment as POM121^DN.

When POM121^DN was added to the full assembly system at t=0, a dose-dependent inhibitory effect on nuclear assembly was observed (Fig. 1B). Control nuclei stained strongly with mAb414, an antibody that recognizes a group of phenylalanine–glycine repeat nucleoporins (FG Nups) in mature NPCs, and accumulated a fluorescent classical NLS import cargo (Fig. 1B, top row). No NPC staining or import were observed in reactions that contained 10 μM or higher concentrations of the POM121 fragment, a result that was comparable to full inhibition by the calcium chelator BAPTA. To determine whether the aborted assembly intermediates were enclosed by continuous, fully fused membranes we performed an identical set of assembly reactions with fluorescently labeled membrane vesicles (Fig. 1C). After 1 hour of assembly, both the control nuclei and BAPTA-inhibited intermediates were encircled by a continuous membrane stain, whereas a discontinuous membrane rim was observed in the presence of POM121^DN (Fig. 1C, bottom row). This suggested a possible defect in membrane vesicle fusion on the surface of chromatin.

We next performed a time-course experiment to follow the kinetics of membrane recruitment to the chromatin templates in the cell-free assembly system. Membrane recruitment and fusion were largely complete by 20 minutes in both the control and BAPTA reactions (Fig. 2A). A substantial delay in membrane recruitment was observed in assembly reactions containing 10 μM POM121^DN, with a discontinuous membrane rim staining seen at 30 minutes. This suggests that the discontinuous membrane staining observed at 1 hour (Fig. 1C) could be the result of a failure to recruit a specific subtype of membrane vesicles (see Discussion). Nuclear assembly intermediates that were incubated in the presence of the POM121^DN fragment were also fragile and showed an increased sensitivity to handling of the solution in the samples removed between 30–60 minutes of incubation. To circumvent this problem we used the anchored chromatin assembly method, in which chromatin templates are attached to a solid support before being exposed to egg extract (Macaulay and Forbes, 1996; Rotem et al., 2009). This type of assembly results in thin, elongated chromatin units of similar dimensions in control and inhibited assembly reactions (Fig. 2B, insets), and in membrane recruitment kinetics that are similar to those observed in solution (data not shown). We chose the 45 minute time point and prepared anchored chromatin assembly reactions for imaging by field emission scanning electron microscopy (FESEM). Both the control and the BAPTA-inhibited chromatin templates were fully enclosed by membranes (Fig. 2B, left and middle panels). Abundant NPCs were present in the control, whereas no NPC-related structures were observed on the smooth membrane surface of BAPTA intermediates. By contrast, anchored chromatin templates assembled in the presence of POM121^DN had a very uneven surface (Fig. 2B, right panel) with large areas of exposed chromatin (arrowhead), occasional areas covered with flattened fused membranes (thin arrow) and additional membranes vesicles docked onto chromatin or other membranes (thick arrow). No NPCs were observed in the areas covered by fused membranes. This indicated that nuclear assembly and NPC formation were disrupted by POM121^DN at an early stage.

The earliest characterized step in postmitotic NPC assembly is the seeding of chromatin by ELYS and the Nup107–160 nucleoporin complex, which can occur even in the absence of membranes (Rasala et al., 2008). We first investigated whether the recombinant POM121^DN–GST fusion protein itself was binding to chromatin and blocking access of the ELYS–Nup107–160 seeding module. Anchored chromatin templates were incubated in membrane-free cytosol and analyzed by immunofluorescence staining. An anti-GST antibody was used to detect the fusion protein, resulting in a strong signal on the surface of chromatin (Fig. 3A). Immunogold staining for FESEM demonstrated the specificity of the anti-GST antibody and also that the POM121^DN binding sites are dispersed over the three-dimensional surface of chromatin (supplementary material Fig. S2A). To determine whether chromatin binding by POM121^DN prevented seeding, we followed several specific components of the seeding module. Immunoblot analysis of the chromatin-bound fraction demonstrated that ELYS, Nup133 and Nup107 were all recruited to chromatin at normal levels when POM121^DN was present in cytosol (Fig. 3B, left). This was confirmed by immunofluorescent staining with an anti-Nup107 antibody (Fig. 3B, right).

Next, we examined whether the binding of recombinant POM121^DN to chromatin affected the recruitment of the endogenous full-length POM121 from the membrane fraction to the nuclear periphery. GST or the POM121^DN fusion protein were added to the full assembly system containing membranes, and nuclei were separated from cytosol by centrifugation through a sucrose cushion. Immunoblot analysis demonstrated that the amount of full-length POM121 recruited to chromatin was reduced when POM121^DN was present in the reaction, whereas gp210 remained unaffected (Fig. 3C).

Taken together, our results suggest that the POM121^DN fragment exerts its dominant-negative effect by direct binding to chromatin, but this binding does not prevent the seeding of chromatin by ELYS–Nup107–160. Conversely, POM121^DN binding might depend on previous seeding by ELYS–Nup107–160. To test this possibility, we immunodepleted ELYS from egg cytosol (supplementary material Fig. S3A) and used the anchored chromatin immunofluorescence assay to follow Nup107 and POM121^DN. As expected from previous work (Franz et al., 2007; Rotem et al., 2009), the depletion of ELYS abolished Nup107 binding to chromatin (Fig. 3D, bottom left panels). However, POM121^DN still bound strongly to the chromatin periphery, as demonstrated by anti-GST staining (Fig. 3D, right panels). We conclude that POM121^DN and ELYS–Nup107–160 are independently recruited to chromatin.

Previous work in the Xenopus assembly system by Antonin and colleagues has led to the suggestion that POM121 coordinates NPC assembly with NE fusion (Antonin et al., 2005). In that study, immunodepletion of POM121-containing membrane vesicles, or solubilization of the membranes followed by immunodepletion of the protein itself, prevented the formation of a continuous NE. This phenotype is similar to that observed after the addition of the POM121^DN fragment in our experiments (Figs 1, 2). Surprisingly, Antonin et al. also found that co-depletion of the Nup107–160 complex and POM121 suppressed the inhibitory effect on membranes, leading to a fully enclosed NE, albeit with no NPCs (Antonin et al., 2005). We therefore investigated whether depletion of the ELYS–Nup107–160 seeding module would ‘rescue’ NE sealing when nuclear assembly is conducted in the presence of POM121^DN. Mock-depleted and ELYS-depleted cytosol were used in reconstitution reactions containing chromatin and fluorescently labeled membranes, with or without the addition of the POM121^DN fragment. Both of the control reactions showed a continuous membrane staining pattern at the nuclear periphery (supplementary material Fig. S3B, left). The addition of POM121^DN inhibited the formation of a continuous NE and this effect was not reversed by the depletion of ELYS–Nup107–160 (supplementary material Fig. S3B, bottom right). Thus, the inhibitory effect of the POM121^DN fragment is distinct from the effects of immunodepleting the endogenous membrane protein. In addition, the overall distribution of POM121^DN on the three-dimensional surface of chromatin remained unchanged when ELYS was depleted from cytosol and chromatin seeding was inhibited (supplementary material Fig. S2B).

We and others have previously shown that importin-β negatively regulates several distinct steps in postmitotic NPC assembly, including chromatin seeding by ELYS–Nup107–160 (Harel et al., 2003a; Lau et al., 2009; Rotem et al., 2009; Walther et al., 2003). We next investigated whether importin-β was also capable of regulating the binding of POM121^DN to chromatin. Indeed, the addition of excess importin-β into cytosol prevented POM121^DN binding in both the immunoblot and immunofluorescence anchored-chromatin assays (Fig. 4A,B). The binding of the recombinant POM121 fragment to chromatin was more sensitive to this negative regulation by importin-β than the binding of the endogenous ELYS–Nup107–160 complex (Fig. 4A). The inhibitory effect of importin-β was readily reversed by an excess of the RanQ69L mutant, which is locked in the GTP-bound form, and this was quantified in the immunofluorescence binding assay (Fig. 4B). A comparison of the immunofluorescence signals in this assay indicated that importin-β was able to reduce POM121^DN binding by ~88% and an excess of RanQ69L–GTP returned binding to the initial levels (Fig. 4B, right). Thus, under physiological conditions, the high concentration of RanGTP near the surface of chromatin would be expected to tip the regulatory balance towards POM121 binding. These results suggest that importin-β regulates two independent chromatin binding modules: the membrane-anchored POM121 and the soluble ELYS–Nup107–160 complex (see Discussion).

To determine whether the negative regulatory effect of importin-β on POM121^DN binding reflects its interactions with the endogenous membrane protein, we decided to re-investigate the effect of adding excess importin-β to the full assembly system (Harel et al., 2003a). We used fluorescently labeled membranes to follow vesicle targeting and fusion on the surface of chromatin and two specific antibodies to follow POM121 and gp210 (Fig. 4C). As previously reported, excess importin-β blocked nuclear assembly at an early stage and resulted in docked, unfused membrane vesicles, visualized by a general membrane dye. Anti-POM121 staining revealed that the endogenous membrane protein was strongly affected by importin-β and did not appear to be substantially recruited to the surface of chromatin. By contrast, anti-gp210 clearly stained membrane vesicles on the surface of these aborted assembly intermediates (Fig. 4C, left, +importin-β columns). Very similar results were obtained when cytosol was omitted from the assembly reactions (Fig. 4C, right), supporting the notion that endogenous POM121 has an intrinsic capability to be targeted to chromatin and that this targeting is regulated by importin-β.

Our results suggest a complex regulatory interplay between importin-β, RanGTP and POM121, which are all targeted to chromatin at the exit from mitosis. Recent work by two groups (Doucet et al., 2010; Yavuz et al., 2010) has pointed to two classical bipartite NLS motifs in the central region of Xenopus laevis POM121 as the crucial determinants for interaction with importin-β. One of these bipartite NLSs, designated NLS1 by Yavuz et al. is present within the POM121^DN fragment as part of a highly conserved 35 amino acid strip (Fig. 1A) (Yavuz et al., 2010). A possible interpretation of our data might be that the crucial determinants for POM121^DN action are all present within this conserved strip and that in vivo, importin-β modulates the targeting of the full length membrane-anchored POM121 by masking this region.

To investigate the functional role of the NLS1 motif, we first examined whether the POM121^DN fragment could affect the function of NPCs after their assembly. Nuclei were assembled for 1 hour and the recombinant fragment was then added and incubated in the assembly reaction. Mature NEs with functional NPCs were formed, as determined from the normal rim staining by mAb414 and the accumulation of the M9–GFP cargo, imported by transportin (Fig. 5, FG Nups and M9-GFP import). However, a classical NLS–BSA cargo was excluded from the nuclei to which POM121^DN was added at this late time point (Fig. 5, second row). This suggests a specific perturbation of the importin-α/β pathway. Next, we focused on the conserved 35-amino-acid strip containing the NLS motif, and expressed it in a soluble recombinant form. When this NLS1-strip^279–314 was added to nuclear reconstitution reactions after nuclei had assembled, it also blocked the importin-α/β pathway (Fig. 5, third row). When NLS1-strip^279–314 was added at t=0, it specifically blocked importin-β-mediated import, but it was not sufficient to inhibit NE and NPC assembly (supplementary material Fig. S4A). Thus, the conserved NLS strip appears to mediate the importin-β-interactions of POM121^DN, but does not account in itself for the dominant-negative effect on NPC assembly.

Based on our findings with the NLS1-strip, we predicted that inactivating the NLS motif in POM121^DN would abolish the interaction with importin-β, but not the inhibitory effect on NE and NPC assembly. We generated point mutations in the bipartite NLS (supplementary material Fig. S4B) and tested the resulting POM121^DN* fragment in an import assay with pre-assembled nuclei. As expected, the recombinant fragment with a mutated NLS no longer blocked the importin-α/β pathway (Fig. 5, compare second and fourth rows). Next, we tested both the original and mutated recombinant fragments for their effect on nuclear assembly in reconstitution reactions containing fluorescently labeled membrane vesicles. Both POM121^DN and POM121^DN* prevented the formation of a continuous NE (Fig. 6A). Staining with mAb414 and import assays confirmed that no NPCs were formed in these reconstitution reactions (data not shown). We next tested the ability of the POM121^DN* fragment to bind chromatin in the anchored chromatin assay. POM121^DN* produced a specific staining pattern on the surface of chromatin and this binding was insensitive to the addition of excess importin-β (Fig. 6B). We conclude that the POM121^DN* protein, containing the inactivated NLS, still binds chromatin and is a potent inhibitor of NE and NPC assembly in the postmitotic mode. Lastly, we tested the localization of POM121^DN and POM121^DN* when added after the assembly of functional nuclei and NPCs. The POM121^DN fragment strongly accumulated inside the nucleoplasm, whereas POM121^DN* was not targeted to pre-assembled nuclei (Fig. 6C). Thus, NLS-mediated entry through existing NPCs and chromatin binding are two separate functions, although they map to the same region of POM121.

The analysis of nine different internal fragments of Xenopus POM121 in the anchored chromatin binding assay is summarized in supplementary material Fig. S5. This analysis suggests that the chromatin binding activity of the POM121^DN fragment results from the combined effect of at least two separate regions flanking the NLS1 motif. These putative chromatin binding elements are not immediately adjacent to the NLS itself and do not overlap the conserved 35 amino acid strip containing this motif (see Discussion).

To determine whether the POM121^DN fragment can also perturb the interphase mode of NPC assembly, we used the inhibition-and-rescue approach in the Xenopus reconstitution system (Harel et al., 2003a) (see also D'Angelo et al., 2006). Fully sealed, poreless NEs were first assembled in the presence of 5 mM BAPTA, and inhibition was subsequently relieved by dilution in fresh cytosol (Fig. 7A, control). Rescued nuclei were round and contained decondensed chromatin and functional NPCs, as determined by mAb414 staining and import assays (Fig. 7A, middle panels). Because NPCs are inserted into pre-existing double nuclear membranes in this scenario, we could identify whether the soluble POM121^DN fragment affected assembly under interphase settings. When POM121^DN was added to cytosol in the rescue phase (following 1 hour of assembly in the presence of BAPTA), chromatin remained condensed and the NE did not expand (Fig. 7A, right). However, these rescued intermediates stained strongly with mAb414 and contained at least some functional NPCs, because M9–GFP import was restored (Fig. 7A, bottom right).

The observed inhibitory effect of POM121^DN in this experiment can be interpreted in different ways. The initial insertion of functional NPCs into the sealed nuclear membranes might allow access and binding of the dominant-negative fragment to chromatin and thus prevent chromatin decondensation. The POM121^DN fragment might also be acting at the membrane level by preventing further recruitment of membrane components, thus blocking nuclear membrane expansion.

To test for a potential effect on the membranes, we first assembled normal nuclei for 1 hour and subsequently added the recombinant proteins. We then monitored the expansion of the NE upon continued incubation in egg extract. All of the nuclei were round and contained decondensed chromatin (Fig. 7B). The average surface area of nuclei in the control reaction was nearly doubled during the second hour of incubation and continued to increase within the third hour (Fig. 7B, top row and +GST histograms). By contrast, the addition of the POM121^DN fragment effectively blocked all further expansion of the NE (Fig. 7B, middle row and +POM121^DN histograms).

This dramatic effect on nuclear expansion could theoretically be attributed to the specific blocking of the importin-α/β pathway by POM121^DN. A recent study has demonstrated that nuclear size is strongly correlated to titratable factors in Xenopus egg extracts, namely importin-α, which is responsible for the rate of import of lamin B3 (Levy and Heald, 2010). To investigate whether the observed block in NE expansion was due to perturbation of the importin-α/β pathway, we tested the effect of the mutated POM121^DN* fragment on pre-assembled nuclei (Fig. 7B, bottom row and +POM121^DN* histograms). The fragment with the inactivated NLS had an identical effect to the POM121^DN fragment and blocked all further expansion of the NE. The addition of the smaller NLS1-strip^279–314 fragment decreased the rate of nuclear growth, but did not block NE expansion altogether (supplementary material Fig. S6). Thus, the dominant-negative POM121^DN fragment completely blocks NE expansion in the interphase assembly mode and does so independently from the NLS-mediated interaction with importin-α and importin-β.

The results of mAb414 staining (Fig. 7A,B) suggest that new NPCs were inserted into a pre-existing NE in the presence of POM121^DN, but further expansion of the membranes was inhibited. To directly visualize the results of the inhibition of membrane expansion in the interphase assembly mode, we prepared samples for imaging by FESEM. Normal assembly was carried out for the minimum time required to form an enclosed envelope. GST or POM121^DN were then added and the assembly reactions were further incubated to allow NPC insertion into the intact NEs. Mature NPCs were dispersed over the surface of control nuclei with plenty of exposed membrane areas in the intervening spaces (Fig. 8, control). By contrast, densely packed NPCs covered the entire surface of nuclei assembled in the presence of POM121^DN, with only minimal exposed nuclear membranes in between (Fig. 8, right). Morphometric analysis of comparable areas of flat NE showed a large difference in the average density of NPCs in such experiments (Fig. 8, histograms).

In summary, the POM121^DN fragment has a dominant-negative effect on both modes of assembly in the cell-free reconstitution system. The addition of this soluble fragment of POM121 uncouples NE formation or expansion from NPC assembly and results in two drastically different and abnormal phenotypes. This suggests that POM121 plays an important mechanistic role in both the postmitotic and interphase modes of NPC assembly.

In this study, we used the Xenopus nuclear reconstitution system to investigate the role of POM121 in nuclear pore and NE assembly. We describe the dominant-negative effects of a soluble POM121 fragment on both the postmitotic and interphase modes of assembly. The region of Xenopus POM121 that contains the dominant-negative fragment has been previously implicated in forming interactions with importin-α and importin-β and some nucleoporins (Rasala et al., 2008; Yavuz et al., 2010). A similar internal domain of human POM121 induces the ectopic recruitment of soluble Nups to mitochondria, when targeted to the outer mitochondrial membrane (Stavru et al., 2006b). Recent work has identified direct interactions of the same general region of human POM121 with the β-propeller domains of Nup160 and Nup155 (Mitchell et al., 2010). It is therefore logical to assume that our POM121^DN fragment contains binding elements that participate in the Nup–Nup interactions that build up the NPC. However, our results also point to other important interactions mediated by this domain (Fig. 9).

The strongest inhibitory effect of the soluble POM121^DN fragment is observed in the postmitotic mode, when it is added to nuclear reconstitution reactions at t=0. The POM121^DN fragment is targeted to the surface of chromatin and disrupts membrane recruitment, preventing the formation of a continuous NE. FESEM analysis reveals large areas of exposed chromatin and only partial enclosure by membranes, with no sign of NPCs (Fig. 2B). These observations are in good agreement with the previously reported results of immunodepleting POM121- or NDC1-containing vesicles from egg extract (Antonin et al., 2005; Mansfeld et al., 2006). Several distinct subpopulations of membrane vesicles have been described in Xenopus egg extracts that differ in their ability to bind chromatin and fuse to other membranes (Antonin et al., 2005; Drummond et al., 1999; Sasagawa et al., 1999; Vigers and Lohka, 1991). Our results show that both the dominant-negative fragment and excess importin-β reduce the amount of endogenous POM121 that is targeted to chromatin, but do not affect gp210 (Fig. 3C, Fig. 4C). This is consistent with a block in the recruitment of membrane vesicles that are enriched in POM121 and NDC1 (Antonin et al., 2005). Importantly, the presence of the POM121^DN fragment is also sufficient to block FG Nup recruitment (as determined by mAb414 staining) and the formation of NPC-related structures (identified by FESEM) in the postmitotic mode of assembly.

These results suggest that the central domain of POM121 is involved in three important functions: (1) targeting to chromatin; (2) soluble Nup recruitment; and (3) the completion of nuclear membrane assembly. Antonin et al. have previously suggested that POM121 is an essential component of a checkpoint mechanism that coordinates early steps in NPC assembly with fusion events required for NE formation (Antonin et al., 2005). Our current results support this view of a functional link between NE and NPC assembly, although it is not necessary to regard this as a true cell-cycle-dependent checkpoint mechanism (Stavru et al., 2006a). Interestingly, our results differ from those of the depletion of POM121 from membranes, in that co-depletion of ELYS–Nup107–160 does not ‘rescue’ the formation of a continuous NE. This suggests that the blocking of chromatin attachment sites by POM121^DN is a major facet of the dominant-negative effect, and implies that other membrane proteins might provide the chromatin targeting function when POM121 is absent.

Our results show that POM121^DN binds to chromatin and prevents its enclosure by a continuous NE. This inhibitory effect is specific, because ‘seeding’ by ELYS–Nup107–160 is not affected. Previous studies have implicated multiple integral membrane proteins, including POM121, in direct binding to chromatin (Anderson et al., 2009; Ulbert et al., 2006). However, these interactions have been generally assumed to be collaborative and redundant. Our observations suggest that POM121^DN binds to specific sites on chromatin, which are essential for nuclear assembly, but are distinct from the ELYS binding sites. We mapped the chromatin binding elements within the POM121^DN fragment to two separate regions flanking the NLS1 motif, but not immediately adjacent to it. These regions are not well conserved in vertebrates and do not contain any known DNA binding or other motifs. To date, no specific DNA sequences have been implicated in the targeting of any nucleoporin to chromatin, although the chromatin binding elements in ELYS have been identified as an AT-hook motif and one additional region (Rasala et al., 2008). Thus, an important future challenge will be to define the precise molecular interactions, either protein–protein or protein–DNA, that initiate postmitotic NPC assembly on chromatin.

Our data indicate that POM121 and ELYS–Nup107–160 are independently recruited to chromatin in the postmitotic assembly mode. Strikingly, both of these chromatin-binding modules are negatively regulated by importin-β. This explains the previously reported inhibition of membrane vesicle fusion by excess importin-β (Harel et al., 2003a). Because the regulatory interaction between importin-β and POM121 can be reproduced in the absence of cytosl, our results also support the existence of a membrane-associated sub-population of importin-α (Hachet et al., 2004). Recent work has identified five functional NLS motifs in the same general region of one of the human POM121 genes and suggests that these are essential for targeting the protein to the inner nuclear membrane during interphase (Funakoshi et al., 2011). Pulldown experiments performed by Yavuz et al. with different fragments of Xenopus POM121 suggest that the two bipartite NLSs of the Xenopus protein form strong interactions with importin-α. Importin-α in turn, mediates RanGTP-sensitive interactions with importin-β and FG Nups (Yavuz et al., 2010). This is in good agreement with our results, showing that inhibition by importin-β of chromatin binding by POM121^DN is readily reversed by RanGTP. However, our results also show that the POM121^DN* fragment, possessing an inactivated NLS, is still a potent inhibitor of NE and NPC assembly. This indicates that the dominant-negative effects of POM121^DN on assembly are not the result of NLS-mediated interactions.

We suggest that in the postmitotic assembly mode the NLS1 motif of POM121 provides a means for regulating the rate of membrane recruitment to the surface of chromatin. Our findings do not exclude the possibility that the POM121 NLSs might play a separate, import-related function in interphase assembly (Doucet et al., 2010; Funakoshi et al., 2011; Yavuz et al., 2010) (see also Fig. 9C). Although ELYS–Nup107–160 was previously suggested to recruit POM121 to chromatin (Rasala et al., 2008), we now show that both of these modules need to bind chromatin before postmitotic NPC assembly can commence (Fig. 9B). The interaction of POM121 with Nup160 (Mitchell et al., 2010), is likely to represent a subsequent step in the assembly process, which might be required to induce fusion of the inner and outer nuclear membrane and the formation of an initial pore channel (Fichtman et al., 2010).

The POM121^DN fragment also interferes with NPC insertion into intact membranes in the in vitro assembly system (Figs 7, 8). These experimental conditions serve as a model for the interphase mode of assembly in metazoan cells (D'Angelo et al., 2006; Doucet et al., 2010; Harel et al., 2003a; Vollmar et al., 2009). When POM121^DN is added to a pre-formed NE, NPC insertion does occur, but it is uncoupled from nuclear membrane expansion. Direct visualization by FESEM reveals densely packed NPCs with very minimal patches of exposed nuclear membrane surface. This is somewhat reminiscent of the phenomenon of ‘NPC clustering’, reported in yeast (Doye and Hurt, 1997; Gorsch et al., 1995; Heath et al., 1995; Pemberton et al., 1995; Siniossoglou et al., 1996; Wente and Blobel, 1994). This term usually refers to the formation of tight clusters of NPCs interspersed with open stretches of poreless membranes. A similar phenotype is observed in the combined deletion of the yeast reticulon RTN1 and YOP1, both suggested to be ER membrane-bending proteins and implicated in NPC assembly (Dawson et al., 2009). We observe an extreme clustering effect, in the sense that it extends across the whole NE and appears to result from a complete block in membrane expansion. Importantly, the dominant-negative fragment does not need to be localized to the nucleus to exert this effect (Fig. 6C, Fig. 7). We hypothesize that the soluble POM121 fragment competes with the endogenous protein by forming specific protein–protein interactions with membrane assembly factors. These might include membrane-associated proteins such as reticulons, or soluble components of the cytosol (Dawson et al., 2009; Rafikova et al., 2009).

Doucet et al. have recently proposed that POM121 plays a rate-limiting role only in interphase NPC assembly. These authors go on to suggest that NPCs are assembled by two different mechanisms during interphase and at the end of mitosis (Doucet et al., 2010). We believe this conclusion could be premature. Our results point to a distinct and crucial role for POM121 in postmitotic NPC assembly, at least in the Xenopus nuclear reconstitution system. Although we describe a very different dominant-negative phenotype in the interphase assembly mode, this might be due to the obvious differences in membrane topology between the two conditions. Little is known about the assembly steps that follow ELYS–107–160 and POM121 recruitment. Thus, it remains possible that NPC assembly per se, is brought about by the same mechanism in the two different cell cycle phases. The data reported here provide support to this view, by demonstrating that NPC biogenesis is intimately linked through POM121 to nuclear membrane formation and/or expansion in both of the assembly modes.

Commercially obtained antibodies included rabbit polyclonal anti-GST (ab9085; Abcam), rabbit polyclonal anti-histone H3 (ab1791; Abcam), fluorescently labeled monoclonal Ab414 (A488-120L; Covance), Alexa Fluor 568 goat anti-rabbit (Invitrogen) and 12 nm gold-conjugated goat anti-rabbit (Jackson ImmunoResearch). Previously described, affinity purified antibodies included anti-human-Nup133, anti-Xenopus-ELYS and anti-Xenopus-Nup107 (Rotem et al., 2009), anti-hNup93, anti-Xenopus gp210 and anti-Xenopus-POM121 (Harel et al., 2003a; Harel et al., 2003b). A second polyclonal anti-Xenopus-POM121 (Antonin et al., 2005), directed against a different region of the protein, was a gift from Wolfram Antonin (Friedrich Miescher Laboratory of the Max Planck Society, Tübingen, Germany).

Amino acid positions in the full-length coding sequence of X. laevis POM121 are those of GenBank accession number AY676874 (Antonin et al., 2005). Coding sequences for the different constructs were inserted into pGEX6P-3 and expressed as GST fusion proteins in the Escherichia coli strain BL21 (DE3) RIL. Fusion proteins were purified in the presence of a protease inhibitor cocktail (#11-836-156-001; Roche) on glutathione–Sepharose 4B beads (GE Healthcare), followed by size-exclusion chromatography on a Superdex GL 10/300 S200 column (GE Healthcare). The conserved strip of aa 279-314, containing the bipartite NLS motif, was cloned into a modified zz-pET28 vector (Vasu et al., 2001) to produce a combined zz-histidine-tagged protein. A zz-6xHis-T7-tag control protein was also purified, for comparison. Point mutations in the NLS1 motif of the POM121^DN construct were generated by site-directed mutagenesis (QuikChange II; Stratagene). Full-length histidine-tagged human importin-β was expressed and purified as previously described (Harel et al., 2003a). In some experiments untagged Xenopus importin-β was tested and produced identical results to the human protein (see Rotem et al., 2009). Expression, purification and loading of RanQ69L with GTP were performed as previously described (Rotem et al., 2009). The histidine-tagged M9–GFP construct was a gift from Michael Elbaum and was expressed and purified as described by Lachish-Zalait et al. (Lachish-Zalait et al., 2009).

Preparation of demembranated sperm chromatin, Xenopus egg extracts and separation into cytosolic and membrane fractions were performed as previously described (Harel et al., 2003a). Membrane-free cytosol was prepared by a 1:1 dilution in ELBS (10 mM Hepes pH 7.6, 250 mM sucrose, 50 mM KCl, 2.5 mM MgCl₂) and subsequent centrifugation, as previously described (Rotem et al., 2009). For the immunodepletion of ELYS from cytosol, 180 μg of affinity purified anti-ELYS or preimmune IgG were bound to 40 μl of protein A–Sepharose beads (GE Healthcare) and used for two consecutive rounds of depletion, as described by Rotem et al. (Rotem et al., 2009). Nuclei were reconstituted by mixing Xenopus egg membrane vesicles and cytosolic fractions at a 1:20 ratio, with an ATP regeneration system, 5 μg/ml nocodazole and sperm chromatin at room temperature (Harel et al., 2003a; Macaulay and Forbes, 1996). Recombinant proteins in ELBS were added at equal concentrations for each set of reactions, with the total volume of addition not exceeding 15% of the reaction.

Pre-labeled fluorescent membranes were prepared essentially as previously described (Hetzer et al., 2000), using CM-Dil (Molecular Probes), and were added to assembly reactions at t=0. The recruitment of specific membrane components was followed by direct immunofluorescence with anti-gp210 and anti-POM121 (Harel et al., 2003b). NPCs were stained with directly labeled mAb414 (Covance), and chromatin with Hoechst 33258 (Sigma-Aldrich). TRITC–NLS–BSA and M9–GFP were added to reconstitution reactions after 1 hour of assembly and samples were removed for analysis after an additional incubation of 15 or 20 minutes. For the separation of nuclear and cytosolic fractions, 25 μl reactions were diluted with 75 μl ELBS, overlaid on 300 μl cushions of 30% sucrose in ELBS and centrifuged in a horizontal rotor at 4000 g at 4°C for 8 minutes. Complete recovery of nuclei or nuclear intermediates was verified by Hoechst 33258 staining and the separation of unbound membrane vesicles was determined in equivalent reactions containing labeled membranes. Normalized samples were immunoblotted with specific markers. Poreless, membrane-enclosed intermediates were formed by the addition of 5 mM 1,2-bis(o-aminophenoxy)ethane-N,N,N′,N′-tetraacetic acid (BAPTA; Calbiochem) to the reconstitution mixture at t=0. Rescue from BAPTA inhibition was achieved by gentle dilution of the inhibited reaction into two volumes of cytosol, supplemented with recombinant proteins in ELBS and further incubation for 1 hour. To monitor membrane expansion in the interphase mode, nuclei were assembled for 1 hour and samples were removed for staining with Hoechst 33258 and mAb414. At this point the remaining assembly mixture was split into three equal volumes and recombinant proteins were mixed into these reactions. Incubation was continued and samples removed for staining at the 2 and 3 hour time points. Both GST and a zz-histidine-tagged control protein were tested at high concentrations and found to have no effect on nuclear growth rate compared with normal assembly reactions. Nuclei were stained and recorded under identical conditions and the surface area of the NE of 30 randomly chosen nuclei in each category was quantified. The cross-sectional area of each nucleus was measured and multiplied by four to estimate the total NE surface area. See Levi and Heald for validation of the accuracy of this method of measurement (Levi and Heald, 2010). Samples from all assays were fixed in 3% formaldehyde (Electron Microscopy Sciences) in ELB and analyzed by epifluorescence microscopy.

Anchored chromatin templates for immunofluorescence and immunoblotting were prepared essentially as previously described (Rotem et al., 2009). Briefly, chromatin was decondensed by nucleoplasmin and allowed to settle onto poly-lysine-coated coverslips, washed in ELB (10 mM Hepes pH 7.6, 50 mM KCl, 2.5 mM MgCl₂) and blocked in 5% BSA-ELB for 20 minutes. Membrane-free cytosol supplemented with an ATP-regeneration system, 5 μg/ml nocodazole and recombinant proteins, was added to the chromatin-coated coverslips and incubated for 25 minutes at room temperature in a humidified chamber. The coverslips were washed three times with ELBK (10 mM Hepes, pH 7.6, 100 mM KCl, 2.5 mM MgCl₂) and then fixed and processed for indirect immunofluorescence, or scraped with SDS-PAGE sample loading buffer for immunoblot analysis. Coverslips were mounted in Fluoromount-G (SouthernBiotech) and recorded on an Olympus BX61TRF motorized microscope, equipped with a UPlanSApo 100× NA 1.4 oil immersion objective and a DP70 digital camera. Images were acquired using the DP Controller and Manager software (Olympus) and figures were prepared in Adobe Photoshop. Quantitative analysis of anti-GST immunofluorescence was performed with OpenView software, as previously described (Rotem et al., 2009; Tsuriel et al., 2006). Hoechst 33258 staining was used to create masks of chromatin surface area, defining relevant pixels in antibody-labeled images. Normalized fluorescence intensity was calculated after the subtraction of non-specific staining and compiled from two independent experiments. Twenty-five randomly chosen, non-overlapping chromatin templates were analyzed in each category.

Anchored chromatin templates on silicon chips (Ted Pella, Redding, CA, USA) were prepared as previously described (Rotem et al., 2009) and incubated with reconstitution mixtures, containing cytosol and membranes, for 45 minutes in a humidified chamber. For immunogold labeling, anchored chromatin binding reactions in membrane-free cytosol were probed with anti-GST followed by 12 nm gold-conjugated goat anti-rabbit (Rotem et al., 2009). For the visualization of NPC density after membrane expansion arrest, assembly was carried out in solution, reactions were diluted in 1 ml ELBS and spun down onto silicon chips for 10 minutes at 1000 g. Morphometric analysis of NPC density was performed by counting NPCs in 6–10 approximately flat NE areas of 1×1 μm in two separate experiments. Samples were fixed and further processed for FESEM by critical-point drying on a CPD030 apparatus (Bal-Tec) and sputter coating with 2 nm of chromium (EMITECH K575X) (Rotem et al., 2009). Samples were then viewed on a Zeiss ULTRA plus field emission scanning electron microscope, using an in-lens detector for secondary electrons to provide the surface structure and an energy selective backscattered (EsB) detector for backscattered electrons to localize gold particles.

The authors thank Douglass Forbes, Ulrike Kutay, Wolfram Antonin and Michael Elbaum for the kind gift of reagents. We also thank Anca Savulescu for critical reading of the manuscript.