|Home | About | Journals | Submit | Contact Us | Français|
Nuclear pore complexes (NPCs) perforate the nuclear envelope and represent the exclusive passageway into and out of the nucleus of the eukaryotic cell. Apart from their essential transport function, components of the NPC have important, direct roles in nuclear organization and in gene regulation. Due to its central role in cell biology, it is of considerable interest to determine the NPC structure at atomic resolution. The complexity of these large, 40-60 MDa protein assemblies has for decades limited such structural studies. More recently, exploiting the intrinsic modularity of the NPC, structural biologists are making progress toward understanding this nanomachine in molecular detail. Structures of building blocks of the stable, architectural scaffold of the NPC have been solved, and distinct models for their assembly proposed. Here we review the status of the field and lay out the challenges and the next steps toward a full understanding of the NPC at atomic resolution.
The hallmark of eukaryotic cells is an elaborate endomembrane system that creates membrane-enclosed organelles. The nucleus is the most prominent organelle, as it harbors the genetic material of the cell. NPCs are the only gateways to the nucleus and reside in circular openings in the nuclear envelope where the inner nuclear membrane (INM) and outer nuclear membrane (ONM) of the nuclear envelope (NE) are fused. NPCs are among the largest multiprotein assemblies in the quiescent cell and were first described 50 years ago by electron microscopy (Watson, 1959). Here we review the status of structural characterization of the nuclear pore complex – the results of truly multi-disciplinary efforts. For general reviews the reader is also referred to (D’Angelo and Hetzer, 2008; Lim et al., 2008; Tran and Wente, 2006) and for the mechanism of NPC assembly to (Antonin et al., 2008) and for nucleocytoplasmic transport of proteins and RNAs to (Carmody and Wente, 2009; Cook et al., 2007; Kohler and Hurt, 2007; Pemberton and Paschal, 2005; Stewart, 2007; Weis, 2003). The role of the NPC in gene regulation and nuclear organization is addressed in (Akhtar and Gasser, 2007; Heessen and Fornerod, 2007).
Electron microscopy has been the best technique to observe the overall structure of the NPC. A variety of cell types from different organisms have been imaged. In its internal symmetry, shape, and size the NPC seems conserved throughout evolution, though at the molecular level there are differences, as noted below. In internal symmetry, the first electron micrographs of the NPC showed that it forms an octagonal ring whose central channel is less electron dense that the eight lobes that surround it. In shape, scanning electron microscopy experiments (SEM) have recorded some of the most stunning NPC images (Figure 1). While the architectural core is grossly symmetric about the plane of the membrane, the peripheral components on the nuclear and cytoplasmic faces are distinct. These peripheral components recapitulate the eightfold symmetry about the transport axis exhibited by the architectural core. On the cytoplasmic side, eight knobs, thought to be attachment sides for fibrous extensions, are visible in NPCs from multicellular species (Kiseleva et al., 2000). In yeast, these features are less pronounced, but likely are also present (Kiseleva et al., 2004). On the nucleoplasmic side, a ring termed the nuclear basket is suspended from eight filaments that join it to the NPC. Lastly, in size, the diameter of the NPC appears to be similar in all eukaryotes, about 90-120nm (Akey and Radermacher, 1993; Beck et al., 2007; Fahrenkrog et al., 2000; Hinshaw et al., 1992; Stoffler et al., 2003). However, there is still considerable uncertainty about the height, determined to be ~30-50nm (Alber et al., 2007b; Elad et al., 2009).
Further work has led to progressively more detailed reconstructions of the NPC. Cryo-electron microscopy that relies on averaging images from many NPCs has been employed to study the core NPC structure, the part that spans the distance between the faces of INM and ONM (Akey and Radermacher, 1993; Hinshaw et al., 1992). These studies have shown that the scaffold ring structure – the electron dense material near the nuclear membrane – has alternating thicker and thinner regions, hence it is often called the spoke ring (Akey and Radermacher, 1993; Hinshaw et al., 1992). The scaffold structure appears to penetrate the pore membrane to also form a perinuclear ring structure. Using cryo-electron tomography, the best pictures of complete NPCs have been achieved, extending even to a resolution of ~ 6 nm (Beck et al., 2007; Elad et al., 2009). With this technique, details of the ring structures become apparent. The scaffold can be divided into three main ring elements: a central spoke ring is sandwiched between a cytoplasmic ring and a nucleoplasmic ring. The rings appear to float on top of one another, indicating that material connecting them is less electron-dense than the rings themselves. Alternatively, this may be due to technical difficulties, such as the ‘missing cone’ problem or poor resolution in the Z-direction.
The central transport cavity of the nuclear pore complex shows no distinct structural features, consistent with the perception that it is filled by an aqueous meshwork formed by natively unfolded FG-domains, which are long polypeptide sequences found in several nucleoporins that contain phenylalanine-glycine (FG) repeats but are otherwise hydrophilic. These extensions are thought to form a distinct, semi-permeable environment that prevents the diffusion of large molecules, unless they are bound to nuclear transport factors that facilitate entry into this central cavity.
In addition to the central channel, the scaffold itself likely harbors additional, peripheral channels. The spoke ring appears porous in cryo-EM/-ET structures, with gaps of ~ 9 nm diameter close to the NE membrane (Hinshaw et al., 1992; Stoffler et al., 2003). Peripheral channels have been discussed in several studies, and postulated to transport small proteins and ions (Kramer et al., 2007). It is unclear, however, how this type of transport could be restricted to the peripheral channels, when the central channel could allow it as well. Alternatively, it also has been suggested that the peripheral channels transport membrane proteins destined for the INM. These are inserted into the ER membrane following translation and stay membrane-anchored until they reach their final destination (the ER, ONM, and INM are all contiguous). Perhaps these membrane proteins pass the NPC into the nucleus via these peripheral channels (Powell and Burke, 1990; Zuleger et al., 2008). The nucleoplasmic domains of INM proteins are limited in size to ~ 40 kDa, about as large as cavities of the observed channels.
While the general NPC architecture is well established, the cryo-EM/-ET structures do not permit the assignment of individual proteins, since their boundaries are not visible at this resolution. For this, higher resolution methods are required.
A characteristic of the NPC is its high degree of modularity, which manifests itself at several levels. First, the NPC is organized around a central eightfold rotational symmetry. Second, only ~30 nucleoporins, composed of a limited set of domain topologies, build the NPC. Third, nucleoporins have various dwell times at the NPC, with only a fraction being stably attached at all times. And finally, the stably attached nucleoporins are arranged into subcomplexes, each of which assembles in multiple copies to build the entire NPC (Figure 2). This modularity is the basis for approaching structural determination of the assembly at atomic resolution (Schwartz, 2005).
Two studies, using S. cerevisiae (Rout et al., 2000) and rat hepatocytes (Cronshaw et al., 2002) as starting material, determined an inventory of nucleoporins. In both studies, cell extracts were enriched for NPCs by fractionation and analyzed by mass-spectrometry to identify the proteins purified thereby. The set of proteins found in both organisms is largely identical and comprises ~ 30 different gene products. The nucleoporins can be broadly classified into three categories (Figure 3). ~10 contain disordered N- and/or C-terminal regions that are rich in phenylalanine-glycine (FG) repeats. These FG-repeat regions emanate into and form the transport barrier in the channel of the NPC. ~15 nucleoporins have a distinct architectural functions and form the NPC scaffold structure. Three nucleoporins have transmembrane domains and anchor the NPC in the circular openings in the NE. Immunogold-labeling of all nucleoporins shows that the majority of the nups, notably scaffold nucleoporins, are symmetrically localized around a two-fold symmetry axis in the plane of the NE, perpendicular to the eightfold rotational symmetry about the main transport channel (Rout et al., 2000). Based on simple hydrodynamic and volumetric calculations the size of the NPC was estimated to range from 66 MDa in S. cerevisiae (Rout and Blobel, 1993) to 125 MDa in vertebrates (Reichelt et al., 1990). Calculations based on the stoichiometry of nucleoporins obtained in the proteomic studies, however, indicate that the NPC size is only 44 MDa in S. cerevisiae and ~ 60 MDa in rat. The discrepancy supports the conclusion that the NPC is a porous, lattice-like assembly, rather than a solid entity (Brohawn et al., 2008; Hinshaw et al., 1992), which accounts for the overestimate of mass based on volumetric analysis.
The majority of nucleoporins are organized in discrete subcomplexes each present in multiple copies that arrange according to the symmetry elements of the NPC to form the complete structure. The subcomplexes are biochemically defined and reflect the stable interaction of subsets of nucleoporins. Interestingly, these subcomplexes are also found as entities in mitotic extracts of higher eukaryotes, when the nuclear envelope breaks down during open mitosis (Matsuoka et al., 1999). At the end of mitosis, NPCs reassemble from these subcomplexes in a defined order (Dultz et al., 2008). Each of the eight spokes arranged around the central rotational axis is composed of 5 subcomplexes (Figure 2). Nup82/Nup159/Nsp1 form a subcomplex localized at the cytoplasmic side of the NPC (Belgareh et al., 1998). A second pool of Nsp1 complexes with Nup57 and Nup49 and resides in the center of the NPC, forming the bulk of the central transport barrier (Grandi et al., 1993). The scaffold ring is constructed from two major subcomplexes: the heptameric Y- or Nup84-complex and the heteromeric Nic96 complex. The Y-complex is the best-characterized subcomplex of the NPC and is essential for its assembly, as shown in several organisms (Boehmer et al., 2003; Fabre and Hurt, 1997; Galy et al., 2003; Harel et al., 2003; Walther et al., 2003). It has 7 universally conserved components – Nup84, Nup85, Nup120, Nup133, Nup145C, Sec13 and Seh1 – that assemble stoichiometrically and exhibit the eponymous Y-shape in electron micrographs (Kampmann and Blobel, 2009; Lutzmann et al., 2002; Siniossoglou et al., 2000). In many eukaryotes, notably excluding S. cerevisiae, three additional proteins, Nup37, Nup43, and ELYS/MEL-28, are considered members of the Y-complex, but their architectural role is unclear (Cronshaw et al., 2002; Franz et al., 2007; Rasala et al., 2006). In most models, the Y-complex is thought to symmetrically localize to the cytoplasmic and the nucleoplasmic face of the NPC sandwiching the Nic96 complex. The Nic96 complex is not as well defined as the Y-complex, likely reflecting the fact that it associates less stably. However, Nic96 interacts directly with Nup53/59 (Hawryluk-Gara et al., 2005), and co-immunoprecipitation with Nup188 (Nehrbass et al., 1996) and with Nup192 have been reported (Kosova et al., 1999). Further, the Nic96 complex is the tether to the Nsp1 complex in the center of the NPC. The newest defined subcomplex contains the transmembrane Nup Ndc1, considered to be an anchor for the NPC in the pore membrane. This complex contains Nup157/170 and Nup53/59, which connect the Ndc1 complex to the Nic96 complex (Makio et al., 2009; Onischenko et al., 2009). The other two transmembrane nups, Pom34 and Pom152, are reported to interact with Ndc1 as well, albeit less strongly. Mlp1/2 are attached to the NPC ring via Nup60 (Feuerbach et al., 2002) and likely form the nuclear basket structure (Strambio-de-Castillia et al., 1999) .
An important aspect of the NPC is that it is not a rigidly assembled machine, but a rather dynamic entity. Inverse fluorescence recovery after photobleaching experiments using GFP-tagged nucleoporins showed that different parts of the NPC have drastically different residence times (Rabut et al., 2004). Some mobile components detach from the NPC within seconds, while other components are stable throughout the entire cell cycle. Notably, the components of the structural scaffold – the Y-complex and the Nic96 complex – are stably attached, while FG-nucleoporins are more dynamic. These studies on nucleoporin dynamics are consistent with the very slow protein turnover of scaffold nucleoporins (D’Angelo et al., 2009; Daigle et al., 2001). The scaffold structure of the NPC can be viewed as a docking site for more mobile nucleoporins, which often have functional roles at sites away from the NPC (Kalverda and Fornerod, 2007).
Until about five years ago, very little high-resolution structural information on nucleoporins was available. This was largely due to the technical difficulties of obtaining nucleoporins of sufficient quantity and quality for structural studies, a challenge particularly severe in the case of scaffold nucleoporins. Despite the scarcity of experimental evidence, structural predictions grouped nucleoporins into a small set of fold classes (Berke et al., 2004; Devos et al., 2004; Devos et al., 2006; Schwartz, 2005). First, FG-domains, the primary transport factor interaction sites, are present in about one third of all nucleoporins. Second, coiled-coil domains are present in a number of nucleoporins. Third, scaffold nucleoporins are largely composed of β-propellers, α-helical domains, or a tandem combination of both. Using this simple classification, about 76 % of the mass of the yeast NPC was accounted for.
A total of 13 % of the NPC mass is made up of FG-repeat containing peptide stretches. The repeats are found in terminal extensions of ~10 nucleoporins and make up the physical transport barrier. NTRs specifically interact with the FG-regions, which allow them to enter the central transport channel. How FG-repeat regions exactly form the transport barrier is vigorously investigated, and hotly debated (Frey and Gorlich, 2007; Lim et al., 2007; Peters, 2009; Rout et al., 2003). Systematic deletion of FG-regions from different nups has shown that the total mass of these filaments is more important than any one individual FG-filament, arguing for substantial redundancy in the meshwork (Terry and Wente, 2007). The intrinsic disorder of the FG-filaments stretches is well documented in a series of crystal structures (Bayliss et al., 2000; Fribourg et al., 2001; Grant et al., 2003; Liu and Stewart, 2005). Only short peptide stretches are orderly bound to the convex outer surface of the HEAT-repeats that build NTRs, with the phenylalanine sidechains inserting between neighboring helices. Otherwise, the filaments remain without structure. Little is known about the intervening, non-FG sequences. They are poorly conserved, but are rich in polar and charged residues, probably important for the biophysical properties of the transport barrier.
Coiled coils in the NPC fulfill structural roles. The nuclear basket of the NPC is mainly constructed from the large coiled-coil proteins Mlp-1/2 in yeast and Tpr in vertebrates. Coiled-coils are often used for protein-protein interactions, thus the nuclear basket may serve as a general recruitment platform to bring accessory factors close to the NPC. The desumoylating enzyme Ulp1, for example, is stably associated with the nuclear basket (Li and Hochstrasser, 2000). Lining the central NPC channel are 6 nucleoporins containing coiled-coil regions. The FG-Nup Nsp1 is part of two distinct entities, the Nsp1-Nup57-Nup49 complex (Grandi et al., 1993) and the Nsp1-Nup82-Nup159 complex (Bailer et al., 2001). In both, the proteins are held together by coiled-coil interactions (Bailer et al., 2001) and the Nsp1-Nup57-Nup49 complex is, in addition, tethered to the NPC scaffold via the N-terminal coiled-coil region of Nic96 (Grandi et al., 1995). So far, only a homodimerized 10 kDa fragment of Nup58 (the vertebrate ortholog to Nup57) has been structurally characterized (Melcak et al., 2007). Biochemical analysis suggests that the network involves specific rather than promiscuous interactions, arguing for a specific tethering function for the coiled-coil segments. It will be interesting to see these coiled-coil interactions in atomic detail in order to manipulate them and potentially swap the attached FG-domains within the NPC. Such experiments could provide important insight into the organization of the FG-network, if there is such.
A large portion of the NPC scaffold is build from β-propellers, one of the most abundant classes of proteins, especially in eukaryotes, and with diverse functions (Chaudhuri et al., 2008; Paoli, 2001). A set of Nups were initially identified as β-propellers based on sequence analysis. In yeast, only Sec13 and Seh1 contain the signature WD-40 repeat motif and were among the very first β-propellers to be recognized (Pryer et al., 1993). More nups have since been recognized as β-propellers despite the lack of signature sequence motifs. The N-terminal domain of Nup133 was the first experimentally determined β-propeller of the NPC and after this structure was solved, the additional non-canonical β-propeller domains in the NPC were identified (Berke et al., 2004). To date, 5 of the 8 universally conserved β-propellers in the NPC are structurally characterized (Figure 4). In Nup133, Nup120, and Nup159 (hNup214) the β-propellers are N-terminal and seven-bladed. While forming a distinct entity in Nup133 and Nup159 (Weirich et al., 2004), physically tethered but otherwise not interacting strongly with the C-terminal part of the protein, the β-propeller in Nup120 is fully integrated with an adjacent helical domain to build one continuous oblong domain (Leksa et al., 2009). Seh1 and Sec13 are so far unique variations of β-propellers in that they are open and 6-bladed (Brohawn et al., 2008; Debler et al., 2008; Fath et al., 2007; Hsia et al., 2007). Their partner proteins insert a seventh blade into the β-propeller to complete the domain in trans. The function of the β-propellers is architectural and it is widely assumed that they serve as protein-protein interaction sites. Peripheral β-propellers can recruit accessory proteins, like the mRNA export factor Dbp5 (von Moeller et al., 2009), whereas those more centrally located likely are used to connect subcomplexes.
α-Helical domains make up more than half of the mass of the NPC scaffold. Structural prediction classified the non-coiled-coil α-helical domains into a strongly related group of α-helical solenoids (Devos et al., 2006). α-solenoids are characterized by a two or three helix unit that is repeatedly stacked to form an elongated, often superhelical domain with N and C terminus at opposite ends of the molecule (Kobe and Kajava, 2000). Such regular, α-helical repeat structures are, often in combination with β-propellers, common scaffolds in large protein assemblies such as the clathrin vesicle coat (Edeling et al., 2006), the protein phosphatase 2A holoenzyme (Xu et al., 2006) and the anaphase promoting complex (Herzog et al., 2009), to name a few. Surprisingly, structural characterization of α-helical domain containing Nups has revealed three different α-helical folds, each distinct from a regular α-solenoid arrangement (Boehmer et al., 2008; Jeudy and Schwartz, 2007; Leksa et al., 2009; Schrader et al., 2008b). Nic96 was the first experimentally determined α-helical structure of a scaffold nucleoporin and it showed an unexpected, atypical α-helical topology (Jeudy and Schwartz, 2007; Schrader et al., 2008). The 30 helices of the ~65 kDa domain, excluding the ~200 N-terminal coiled-coil domain, are arranged in a J-like topology, forming an oblong domain. The chain starts in the middle of the elongated domain, zig-zags up on one side of the molecule, folds back over a stretch of 7 helices and then continues past the N terminus to the other end of the molecule (Figures (Figures4,4, ,5).5). Three other α-helical scaffold nucleoporins (Nup84, Nup85 and Nup145C) have since been structurally characterized and shown to adopt the same fold as Nic96, pointing to a common ancestor (Brohawn et al., 2008; see below). A second, distinct α-helical fold has been identified in structures of Nup133 and Nup170, which are more distantly related, but share an extended and stretched α-helical stack (Boehmer et al., 2008; Whittle and Schwartz, 2009), substantially different from the first group. The third was revealed in the structure of Nup120, which forms a domain that fully integrates a β-propeller with an α-helical domain (Leksa et al., 2009). The α-helical segment is built around a central stalk of two long helices wrapped with 9 additional helices in an unprecedented fashion. In summary, the α-helical domains that occur in the NPC fall in different classes that provide a significant challenge for structure prediction methods. One obvious challenge is the exceedingly low sequence conservation, even between orthologs, apparent in the inconsistent nucleoporin nomenclature. Poor sequence conservation is likely due to some degree of malleability of the scaffold structure and the construction from common sequence elements (Aravind et al., 2006). Whether poor sequence conservation is further the result of adaptive evolution, linking several architectural nucleoporins to speciation, is an intriguing possibility that should be explored in more detail (Presgraves et al., 2003; Tang and Presgraves, 2009).
As mentioned above, the four α-helical scaffold nucleoporins Nic96, Nup145C, Nup85, and Nup84 are constructed around a common ~65 kDa domain composed of 28 helices. Notably, this domain has to date only been identified outside the NPC in Sec31, one of the main building blocks of the COPII vesicle coat. The commonality was surprising. Sequence conservation between the five members is so low that no specific structural relationship was inferred previously (Alber et al., 2007, Hsia et al., 2007). This domain, which we termed Ancestral Coatomer Element 1 (ACE1), is a structural manifestation of the likely common origin of the NPC and the COPII vesicle coat (Devos et al., 2004). ACE1 is constructed from three modules, crown, trunk and tail, that together form an elongated molecule of ~ 140Å × 45Å × 45Å. Structural superposition of ACE1 proteins shows that individual modules are closely aligned, while differences in linkers between modules results in significant differences in their relative orientations. These differences, as well as proteolytic susceptibility data, suggest at least modestly flexible hinges connect the modules, especially the trunk and the tail. This likely explains why all crystal constructs except for Nic96 contain either the trunk and crown (Brohawn et al., 2008; Debler et al., 2008; Hsia et al., 2007) or the tail (Boehmer et al., 2008). Even with the structural information in hand, it is difficult to find additional ACE1 proteins. Beyond a few residues conserved between orthologs, ACE1 is not characterized by a distinct sequence motif. The reason for this amazing degeneracy on the sequence level likely is that for folding the ACE1 domain only some general sequence profiles need to be satisfied. For example, helices α5, α7, α15 and α17 are typically hydrophobic, because they are incased by surrounding helices and are largely buried and solvent inaccessible. Thus, a combination of sequence profile evaluation, α-helical prediction, and overall length are currently the only indicators for the ACE1 domain. The two remaining α-helical scaffold nucleoporins without any crystallographic structural information are Nup188 and Nup192. Whether or not they belong to the ACE1 class, remains to be determined, but it appears unlikely.
Structures of large protein assemblies are typically solved by a combination of cryo electron microscopy and x-ray crystallography (Chiu et al., 2006), exploiting the strengths of both methods. With EM-data in the 10-15 Å range, fitting of crystal structures can often be performed fairly reliably, providing in effect a composite structure at atomic resolution. In the case of the NPC, the best tomographic EM data for the full assembly does not extend beyond ~ 58 Å (Beck et al., 2007) and thus, unfortunately, does not lend itself to directly fitting the available crystal structures. A promising step toward closing the resolution gap is the recent EM reconstruction of the Y-complex at ~ 35 Å resolution, which allows at least tentative fitting of crystal structures (Kampmann and Blobel, 2009). Apart from the resolution gap, however, the biggest difficulty currently is the absence of strong experimental data on how the subcomplexes, notably the Y-complex, assemble to form the core scaffold structure of the NPC.
In an alternative approach to the hybrid method of combining experimental EM-and crystallographic data, Alber et al. used a plethora of volumetric and stoichiometric data, combined with distance restraints obtained from a comprehensive co-immunoprecipitation analysis of all nucleoporins, to solve the subcomplex assembly puzzle (Alber et al., 2007a; Alber et al., 2007b). Even though each datapoint has very limited information content by itself, a useful three-dimensional model is generated by combining all the data, conceptually similar to the way an NMR stucture is computed. The resulting draft model at an estimated 5 nm resolution provides a plausible arrangement of Nups in the yeast NPC, information that is not available from the current tomographic studies. According to the computed model, 8 Y-complexes each self-assemble into a cytoplasmic and a nucleoplasmic ring, defining the Z-dimension of the scaffold at ~38 nm (Figure 6). Sandwiched between the two rings are two 8-membered rings of Nup157/170,Nup188 and Nup192. All remaining nucleoporins, except for the filamentous nuclear basket and cytoplasmic extensions, have been positioned as well and decorate the main scaffold. Clearly, this model has severe limitations and will have to be adjusted with more detailed information becoming available. For instance, the Y-complex is modeled as a compact rod in the computer model and the coat is space-filling with very small gaps, both in apparent contrast to results from crystallographic analyses. The exciting possibility of this combinatorial approach however is that it should be possible to integrate currently available and forthcoming high-resolution structures to further refine the model. To date, crystal structures accounting for 23% of the mass of the NPC scaffold are available. Integration of this data could potentially reveal an NPC structure that would come fairly close to the reality.
Blobel and coworkers have taken a different experimental approach to address the subcomplex assembly problem. Their rationale is that the subcomplexes assemble only at very high protein concentrations as they are found in the assembled NPC in the living cell. Such conditions are difficult to reproduce in vitro, but they can potentially be mimicked in protein crystals where the protein concentration is similarly very high. Thus, crystal-packing interactions between dimeric subcomplex fragments have been used to develop an assembly model for the NPC (Debler et al., 2008; Hsia et al., 2007). Besides the overall dimensions of the scaffold, the resulting model is drastically different from the Alber model. Four instead of two 8-membered rings of Y-complexes are stacked on top of each other to form a continuous membrane-proximal shell. The rings are connected by alternating heterooctameric poles of Nup145C•Sec13 and Nup85•Seh1 (Figure 6A). The Nic96 complex is postulated to form a second, inner shell within the NPC that connects to the FG-network. Similar to the Alber model, the Y-complexes are envisioned to directly contact the pore membrane via membrane-inserting ALPS-motif containing amphipathic helices predicted to be present in several Y-complex components (Drin et al., 2007). However, experimental evidence supporting membrane-insertion is so far only available for an ALPS-helix found in the vertebrate Nup133 β-propeller, and the prediction of ALPS-helices is non-trivial. Indeed, several predicted ALPS-helices have been found to be well packed and to contribute to the hydrophobic core of nucleoporins (for example in Nup85 and Nup120), making their involvement in membrane-insertion unlikely.
In a groundbreaking paper, it was postulated that the NPC and vesicle coats share a common ancestor, dating back more than one billion years to the very early eukaryote (Devos et al., 2004). The principal hypothesis was that these assemblies are used to fulfill similar roles specific to eukaryotes, namely stabilizing the highly curved membranes of vesicles and the circular openings in the nuclear envelope. Judged by primary sequence analysis, however, the relationship between the components is largely undetectable. One important pillar of the ‘proto-coatomer’ hypothesis is that Sec13 is a bona fide component of both the COPII vesicle coat and the NPC. Structural evidence supporting the hypothesis was provided by showing that a class of four scaffold nucleoporins shares a specific 65 kDa domain, ACE1, with Sec31 of the COPII vesicle coat (Brohawn et al., 2008; Brohawn and Schwartz, 2009). The COPII coat is organized into a membrane-proximal inner layer built from Sec23•Sec24 heterodimers and Sar1, and an outer coat assembled from Sec13•Sec31 heterodimers (Stagg et al., 2008). Sec31 in the COPII coat and Nup145C in the NPC interact very similarly with Sec13 by insertion of a 7th blade to complete the β-propeller. This interaction mode is recapitulated in Nup85 binding to the Sec13 homolog Seh1, which is also facilitated by insertion-completion of the β-propeller. Since Nup145C, Nup85 and Sec31 all share the same ACE1 domain, the structural data not only suggests a common ancestor, but it also provides clues as to how the NPC assembles.
In comparison to the NPC, the COPII vesicle coat is much simpler, has been extensively studied, and its assembly is quite well understood. Most informative with respect to the NPC assembly is the organization of the Sec13•Sec31 outer coat of COPII. Here, two Sec13•Sec31 heterodimers dock via their ACE1 crown modules to form an edge element (Fath et al., 2007). Four edge elements converge in a vertex, where the N-terminal β-propellers of Sec31 interact. COPII vesicles of different sizes can be assembled from different numbers of edge elements by adjusting the angles at the interaction sites (Stagg et al., 2008). In analogy to the Sec31•Sec31 homodimer interface in the COPII coat, it was predicted on a biochemical and structural basis that Nup145C and Nup84 form a heterodimeric edge element via a crown-to-crown interface in the NPC. Due to the predicted steric conflicts, this arrangement would be fundamentally incompatible with the hetero-octameric pole model discussed above (for details, see Brohawn et al., 2008). We predict that the entire NPC scaffold has an open, lattice-like organization with significant similarity to the COPII coat, also built from vertex and edge elements (Figure 6C). As much as COPII and clathrin lattices seem not to share common construction principles (Fath et al., 2007), so do similarities between NPC and clathrin lattices not extend beyond superficial characteristics.
One important distinction between the computational and the hetero-octameric pole model on one hand and the lattice model on the other is that in the latter the Y-complex and likely also the Nic96 complex components Nic96 and Nup157/170 do not directly coat the membrane, although there may be punctual contacts. By analogy to the Sec23•Sec24 inner coat of COPII, adaptor molecules instead are predicted to directly contact the pore membrane. The transmembrane Nups, especially Ndc1, are prime candidates for this function. Indeed, Ndc1 has already been shown to physically interact with the Nic96 complex via Nup53/59 (Onischenko et al., 2009). Ndc1 is found in all eukaryotes and is essential, though this could also be related to its additional role in the spindle pole body (Winey et al., 1993). Such adaptor-mediated anchoring to the pore membrane is consistent with the membrane-proximal gaps observed in cryo-tomographic reconstructions (Beck et al., 2007; Stoffler et al., 2003).
Obviously, there are some important limitations to the analogy to COPII. First, it is not yet apparent how the verteces are formed in the NPC. In COPII, Sec31 contains an additional N-terminal β-propeller, absent in the ACE1 domains of the NPC, but essential for COPII vertex assembly (Fath et al., 2007). Thus, the vertex might look different in the NPC. Further, the COPII coat is a self-enclosed lattice structure that completely encapsulates the positively-curved membrane vesicle. In the NPC, the lattice can be continuous in lateral direction around the central transport gate, but must terminate on the nucleo- and cytoplasmic side. Further, the circular pore opening has positive curvature only in the axial direction, but negative curvature in equatorial direction. Since all vesicle coats seamlessly wrap positively curved membranes, it is not unreasonable to speculate that a specific architectural element might occur exclusively in the NPC to establish its unique geometrical requirements.
Progress toward the structural characterization of the NPC at high resolution, one of the holy grails of structural biology (Bhattacharya, 2009), has been spectacular since our last review on the subject, a review that was intended as a roadmap toward this ambitious goal (Schwartz, 2005). The past 5 years have seen improvement in images of the entire NPC available, and a flurry of crystal structures of various key parts of the NPC. New computational methods were developed to integrate geometric data from diverse sources, and these were applied to the NPC, in part to demonstrate a general means to tackle other large assemblies (Alber et al., 2008). The path toward a high-resolution structure of the NPC will continue to be highly interdisciplinary. Single-particle reconstruction on subcomplexes has still been used only sparingly, but holds great promise to close the gap between the resolutions provided by crystallography and by tomography. It is likely that integration of structural data, particularly crystallographic data, into computationally derived models will provide further useful information. Finally, developments in super-resolution microscopy promise to allow in vivo measurement of distances within the NPC on the order of 10 nm, a potential source of further useful parameters – parameters so far inaccessible experimentally. These distances would complement other available structural data. Judging by recent progress in the field, dramatic developments in our understanding of the NPC lie just around the corner.
We are grateful to Elena Kiseleva and Ohad Medalia for providing images for Figure 1. Thanks to all members of the Schwartz laboratory for discussion and suggestions. We sincerely apologize to those whose work we have not referenced as a result of unintentional oversight. T.U.S. is supported by the Pew Scholars Program and by NIH grant GM077537.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.