Eukaryotic chromosomes are housed within the nucleus, which is delimited by the two parallel membranes of the nuclear envelope (NE). The evolution of this physical barrier endowed eukaryotes with a critical control mechanism segregating the sites of gene transcription and ribosome biogenesis from the site of protein synthesis. This compartmentalization allows cells to strictly coordinate numerous key cellular processes, but it also presents cells with the challenge of selectively managing the transport of a bewildering number of proteins and RNAs between the nucleus and cytoplasm. This is accomplished by the presence of “nuclear pores,” which arise at points where the inner and outer NE membranes conjoin to form circular channels across the nuclear envelope. Within these pores sit large proteinaceous complexes, appropriately named nuclear pore complexes (NPCs), which, in conjunction with soluble transport factors, govern all biomolecular transport into and out of the nucleus. Beyond this fundamental control of transport, the NPC has adopted a host of other activities by acting as a spatial landmark or anchor site for many of the machineries that directly control gene activity and transcriptional processing (reviewed in Ahmed and Brickner 2007; Hetzer and Wente 2009). As a transporter, it must allow small molecules to pass as freely, prevent most macromolecules from crossing, and permit the quickest possible passage of selected macromolecules bidirectionally across the NE. As an anchor, it must allow free communication between the attached control machineries and the chromatin or transcripts that they regulate without hindering nuclear transport. One can thus also consider the NPC as a major way station in eukaryotes, interacting with and regulating DNA, RNA, and membranes and communicating between the cytoplasm, nucleoplasm, and ER lumen. Because of this, the subject of the nuclear pore complex and nuclear transport is a huge one, far beyond the scope of any single review. Our aim here is therefore to give an overview, including references to many excellent reviews that detail particular areas of study.

The sheer size and flexibility of the NPC make it difficult to fully solve its molecular architecture by conventional techniques. Therefore, an orthogonal approach has been taken; large and diverse sets of proteomic data were amassed and a computational method for using these data was developed to define the relative positions and proximities of the yeast NPC’s constituent proteins. A corresponding average protein density map represents the position of every Nup with a precision of ~5 nm, sufficient to resolve the molecular organization of the entire NPC (Alber et al. 2007a,b) (Figure 2A). The resulting map agrees with complementary data in both yeast and vertebrates (reviewed in Strambio-de-Castillia et al. 2010).

The core scaffold somewhat resembles a vesicle coat, as it forms a discrete layer completely following the curve of the pore membrane, effectively coating it (Figure 2). Thus, it defines the size of the central tube/transporter of the NPC and the height of the NPC core, and all other Nups and poms are attached to either the inner or the outer face of this coat (Figure 1). Biochemical studies have shown that the core scaffold Nups compose several subcomplexes that appear to function as “building blocks” during NPC assembly and can even exchange with a soluble pool in mature NPCs (Lutzmann et al. 2002; D’Angelo et al. 2006; Makio et al. 2009). Morphologically, the scaffold is made of two inner rings, sandwiched between and interconnected with two outer rings, such that the nuclear and cytoplasmic halves of the NPC have one inner and one outer ring each (Figure 2). Although there is still some debate on the matter, a consensus remains that in both yeast and vertebrates these inner rings are compositionally, as well as morphologically, distinct (Tran and Wente 2006; Alber et al. 2007b).

Four large Nups, each just under 200 kDa in size, compose the inner rings: Nup188p and Nup192p (primarily composed of α-solenoid-like folds) and the homologous Nup170p and Nup157p (Figure 2) (made from the clathrin-like pattern of an amino-terminal β-propeller followed by an α-solenoid-like domain), proteins proposed long ago to be core components of the NPC (Aitchison et al. 1995b). The inner rings are adjacent to each other at the equator of the NPC in the same plane as the three poms of the membrane ring with which they interact extensively to anchor the core scaffold to the NE (Alber et al. 2007b; Onischenko et al. 2009). Numerous mutations in all four proteins demonstrate extensive genetic interaction networks with each other and with the membrane ring components, which likely underscore the functional importance of the inner ring in being a keystone of the core scaffold and in anchoring the NPC to the pore membrane (Aitchison et al. 1995b; Tcheperegine et al. 1999; Miao et al. 2006). Moreover, on both the nuclear and cytoplasmic sides of each spoke, one copy of the Nup Nic96p is anchored through Nup192p and a second copy through Nup188p, linking the inner ring to the other internal structures of the NPC (Figure 2) (see below).

Structurally, the most extensively studied set of core scaffold proteins are the seven Nups that compose the yeast outer rings, first identified and characterized by the Hurt laboratory; Nup133p, Nup120p, Nup145Cp, Nup85p, Nup84p, Seh1p, and Sec13p form a discrete complex that can be biochemically isolated and is termed the Nup84 complex (Figure 2B) (Siniossoglou et al. 1996, 2000; Lutzmann et al. 2002; Flemming et al. 2009). Importantly, the evolutionary link between NPCs and vesicle-coating complexes is supported by the fact that Sec13p is shared with the Sec13/31 COPII vesicle-coating complex (Siniossoglou et al. 1996; Salama et al. 1997; Devos et al. 2004). Moreover, just like the inner-ring proteins, all Nup84 complex proteins are formed almost entirely by a β-propeller fold (Seh1p, Sec13p), an α-solenoid-like fold (Nup85p, Nup84p, Nup145Cp), or an N-terminal β-propeller and a carboxy-terminal α-solenoid-like fold (Nup133p, Nup120p), again common to vesicle-coating complexes (Devos et al. 2004, 2006) (Figure 2B and Figure 3). Excitingly, crystal structures primarily from the Blobel and Schwartz laboratories have begun to piece this complex together at the atomic level (Figure 2B) (Hsia et al. 2007; Brohawn et al. 2008; Debler et al. 2008; Brohawn and Schwartz 2009; Leksa et al. 2009; Nagy et al. 2009; Seo et al. 2009; Whittle and Schwartz 2009). Electron microscopy studies have shown that the Nup84 complex formed an extended Y structure (Siniossoglou et al. 2000; Lutzmann et al. 2002; Kampmann and Blobel 2009), and pioneering work from the Hurt laboratory reconstituted this complex from bacterially expressed proteins and showed that the two short arms of this Y are composed, respectively, of Nup120p and Nup85p+Seh1p, while Nup133p, Nup84p, and Nup145Cp/Sec13p form the main stalk (Siniossoglou et al. 2000; Lutzmann et al. 2002). There is evidence that this complex is flexible (Kampmann and Blobel 2009), perhaps reflecting the known flexibility of the NPC in response to changes in NE shape and during its assembly (see below).

FG Nups also usually carry small structured domains, which serve to anchor them to the NPC (Alber et al. 2007b). These are generally predicted to be coiled-coil or α-helical by fold analysis and comparison to vertebrate homologs (Melcak et al. 2007), in addition to β-sandwich and (oddly) RNA recognition motif (RRM) folds (Figure 3) (Devos et al. 2006). Nup82p and Nic96p seem to serve as linkers to attach many of the FG Nups to the core scaffold (mainly through the inner-ring Nups), such that the inner surface of the scaffold is lined with FG Nups whose filamentous FG-repeat regions fill the central channel and extend into the nucleoplasm and cytoplasm (Figure 2). On both the cytoplasmic and nucleoplasmic sides of each spoke two copies of Nic96p carry the FG Nups Nsp1p, Nup57p, and Nup49p and another two copies form interactions to additional copies of Nsp1p, such that these FG Nups face both the nucleus and the cytoplasm. At the cytoplasmic side, Nup82p associates with Nsp1p as well as with the cytoplasmically facing FG Nups Nup159p, Nup116p, Nup100p, and Nup42p (Grandi et al. 1995; Belgareh et al. 1998; Bailer et al. 2000, 2001; A. K. Ho et al. 2000; Rout et al. 2000; Alber et al. 2007a,b). There are also the FG Nups Nup145Np, Nup1p, and Nup60p found on the nucleoplasmic side, connecting mainly to the inner-ring Nups. In addition, Nup53p and Nup59p are attached to Nup170p and Nic96p, and both face the pore membrane (Figure 2). The latter two Nups may also belong to the class of FG Nups, as they can bind transport factors and carry degenerate FG-repeat regions that are predicted to be natively unfolded (Marelli et al. 1998; Fahrenkrog et al. 2000b; Lusk et al. 2002; Makhnevych et al. 2003; Alber et al. 2007b). One FG Nup, Nup145p, uniquely cleaves itself in half at its Phe605-Ser606 peptide bond to produce two separate Nups, Nup145Np (carrying the FG-repeat region and the autoproteolytic β-sandwich domain at its new carboxy-terminus) and Nup145Cp (a mainly α-solenoid-like protein that forms a major component of the Nup84 complex) (Figure 3) (Wente and Blobel 1994; Teixeira et al. 1997, 1999; Rosenblum and Blobel 1999). Although not essential in yeast (Emtage et al. 1997), this cleavage event appears conserved in vertebrates (Fontoura et al. 1999; Hodel et al. 2002; Sun and Guo 2008). Nup145Cp and Nup145Np remain linked as a dynamic complex such that Nup145Np can shuttle between the NPC and the nuclear interior, as does its vertebrate counterpart (Griffis et al. 2002; Ratner et al. 2007). The entirety of Nup145Np is highly conserved with the homologous yeast nucleoporins Nup100p and Nup116p, neither of which undergoes autoproteolysis as they lack a homologous counterpart for Nup145Cp. It seems that lineage-specific gene duplications of an ancestral Nup145N-like gene gave rise to the truncated versions Nup100p and Nup116p, and likely other FG Nups also originated from such, sometimes more ancient, duplication events (Mans et al. 2004; Devos et al. 2006).

Interestingly, these mutants correspond to two key components of the cargo-carrying transport factor pathways, namely the β-karyopherins Kap95p and Kap121p and Ran cycle components [Ran, RanGAP, RanGEF, and Ntf2 (responsible for transporting RanGDP into the nucleus)] (Lusk et al. 2002; Ryan and Wente 2002; Ryan et al. 2003, 2007). The reasons for the functional associations between NPC assembly and transport factors are still being elucidated, but similar connections have been seen in vertebrates (D’Angelo et al. 2006). In yeast, Kap121p has been proposed to target Nup53p to the NPC, where it is attached to the core scaffold component Nup170p. Indeed, recent work has revealed the importance of the core scaffold to the early stages of NPC assembly. Thus, when the C-terminal domain of Nup170p is overexpressed, what appear to be intermediates of NPC assembly accumulate both in the cytoplasm and at the NE (Flemming et al. 2009). Similarly, in strains lacking both Nup53p and its paralog Nup59p, depletion of Nup170p or either of two transmembrane nucleoporins that connect with Nup170p—Pom152p or Pom34p—also caused the accumulation of such intermediates in yeast cells (Onischenko et al. 2009).

For an NPC to be inserted into the intact NE, both the inner and the outer NE membranes must approach at a site and fuse to give rise to the pore membrane, upon which the core scaffold and the rest of the NPC can then assemble. It is curious, therefore, that two of the three poms (Pom152p and Pom34p) are not essential and so are dispensable for NPC assembly and that all three poms (including Ndc1p) are not required for NPC assembly in the closely related fungi, Aspergillus (Liu et al. 2009). Taken together, this suggests that there must be other transiently or dynamically associating membrane proteins that play key roles in initiating the NPC assembly process and fusion of the inner nuclear membrane (INM) and outer nuclear membrane (ONM) to form the pore membrane.

Indeed, there has been a growing cadre of proteins that, while not strictly Nups, play a key role in yeast NPC assembly. As well as Ran, Ran cofactors, and the Kaps (above; Lusk et al. 2002; Ryan et al. 2003, 2007), the two yeast reticulons Rtn1p and Rtn2p and their interacting partner Yop1p have been implicated in NPC assembly (Dawson et al. 2009). Rtns and Yop1/DP1 proteins can deform and mold membranes, having been shown to have roles in both dynamically restructuring and maintaining tubular ER (De Craene et al. 2006; Voeltz et al. 2006; Hu et al. 2008) and, in metazoans, even in postmitotic NE shaping (Anderson and Hetzer, 2008b). Reticulons have a segment that can insert into one leaflet of a membrane, which may promote or induce membrane curvature (Oertle et al. 2003; De Craene et al. 2006; Voeltz et al. 2006; Shibata et al. 2008); indeed, they are depleted in regions of flat membrane, such as the NE between NPCs, and are found to concentrate in curved membrane regions such as tubular ER (De Craene et al. 2006; Voeltz et al. 2006; Anderson and Hetzer, 2008a,b). The apparent absence of these proteins in the mature NPC suggests that they play only a transient role at the beginning of the assembly process, perhaps helping the first NE membrane curving and fusion step to make the pore membrane. Similarly, the NE/ER proteins Apq12p and Brr6p are genetically linked to each other and are necessary for normal NPC assembly and distribution. This work indicates that both proteins are involved in maintaining lipid homoeostasis in the ER, which is necessary for proper NPC insertion and distribution in the NE (Scarcelli et al. 2007; Hodge et al. 2010).

Another candidate NPC assembly factor is Pom33p, isolated in a genetic screen for genes that are essential in cells lacking Nup133p (Chadrin et al. 2010). The transmembrane protein Pom33p and its paralog Per33p are found in both the ER and the NE, although Pom33p shows a preferential dynamic localization at NPCs. Pom33, but not Per33, genetically interacts with Nup84 complex components and the interacting proteins Nup170p and Ndc1p and physically forms a direct complex with Rtn1p (Chadrin et al. 2010). These data, plus the fact that depletion of both Nup170 and Pom33 significantly impaired assembly of NPCs, point to a role for Pom33p in NPC assembly or maintenance of the NE (Chadrin et al. 2010). Pom33p thus potentially links the reticulon membrane bending and manipulation machinery with the assembling NPC, which together possibly either help the transmembrane nucleoporins during the initial membrane fusion event required for the start of NPC assembly or facilitate the stabilization of the nascent nuclear pore. Following this initial pore formation, assembly to form the mature NPC must proceed extremely rapidly, as no naturally occurring intermediates have been found.

Of course, the NPC core scaffold is composed almost entirely of homologs of vesicle-coating proteins, whose function is to mold and fuse membranes into curved vesicles. On the basis of this similarity it has been suggested that the Nup84 complex and the Nup170 inner-ring complex (which interacts directly with poms) could be directly involved, after recruitment to the NE, in forming a coat somewhat like those in coated vesicles that produces the nascent pore membrane and pinches the inner and outer NE membranes together in a manner analogous to pinching off a curved vesicle (reviewed in Fernandez-Martinez and Rout 2009; Hetzer and Wente 2009).

In summary, there appear to be several main steps to NPC assembly. Initially, accessory factors collaborate with transmembrane and inner-ring NPC components to accumulate on both the inner nuclear membrane and the outer nuclear membrane to warp the latter into a fused pore. The recruitment of the inner-ring components would recruit the outer-ring components, permitting the assembly of the entire membrane-coating core scaffold in the pore (Alber et al. 2007b). Rapid association of the remaining FG Nups, other NPC components, and the nuclear basket would then complete the process. However, this sequence of events remains strictly speculative, and much remains to be understood about the mechanism of NPC assembly in yeast or in any other eukaryote.

It was clear that not all proteins destined for the nucleus contain an NLS typified by SV40 large T antigen. The diversity of cargoes and complexes that traverse the NPC is huge, ranging from proteins to RNAs and ribonucleoproteins (RNPs), including mRNPs and ribosomes, to viruses. Analysis of the yeast genome revealed family transport factors structurally related to karyopherin β (and more distantly to karyopherin α). Other eukaryotes studied, even the most evolutionarily divergent, seem to retain this same family of Kaps (DeGrasse et al. 2009; Mason et al. 2009). Members of the β-Kap family are generally large (molecular weight of 100–125 kDa) proteins that share ~20% sequence identity with each other. Each is typified by the presence of up to ~20 HEAT repeats (amphipathic helix-loop-helix motifs) that form a large helical solenoid (with a single extended hydrophobic core) (Figure 5) (Cansizoglu et al. 2007). There are apparently 14 Kaps in S. cerevisiae and at least 19 Kaps in humans (Stewart 2003), all of which differentially bind different classes of nuclear transport signals, FG-repeat nucleoporins, and Ran; unlike the Kap60p:Kap95p dimer, all other β-Kaps bind directly to their cargos (Figure 4).

Karyopherins responsible for importing cargoes are often called importins, and exporters are called exportins. The direction of transport for each karyopherin is dictated by its differential interaction with cargoes and Ran (Figure 4). In cells, Ran primarily exists in two forms: in the nucleus, Ran is maintained in its GTP-bound form by a GTP exchange factor (RanGEF; RCC1, Prp20p, or Srm1p in yeast); this protein is chromatin bound, thus signaling to the nucleocytoplasmic transport system the position of the nucleoplasm by virtue of generating a cloud of RanGTP around it. In contrast, the Ran GTPase-activating protein (RanGAP) is localized to the cytoplasm, so that Ran in the cytoplasm predominates in the GDP form. Karyopherins exploit this property during transport. As mentioned above, during an import cycle, Kaps bind to their cargoes in the cytoplasm, and when they reach the high Ran-GTP in the nucleus, are induced to release their cargoes. In contrast, exportin binding to cargoes is enhanced by the formation of a trimeric complex that includes Ran-GTP. Thus, as this complex meets the RanGAP in the cytoplasm, the GTP is hydrolyzed and the complex falls apart. Indeed, the direction of karyopherin-mediated transport through the NPC can be reversed by inversion of the Ran gradient (Nachury and Weis 1999). Most karyopherins are thought to be recycled to their original compartments empty, but in a few instances they are believed to chaperone another cargo on their return journey (Figure 4).

Studies of prototypical interactions among constituents of these transport pathways have shed considerable light on the structural basis of transport (Figure 5). In the classical pathway, the NLS binds to a long region on the inside of the Kap60 superhelix, made of alternating α-helical turns. Kap95p, which is also made of alternating α-helical turns, forms a spiral with two surfaces, and the inner surface wraps around an extended N-terminal domain of Kap60p (Figure 5) [a.k.a the importin β-binding (IBB) domain] (Cingolani et al. 1999). The interaction of Kap95p with FG Nups is mediated as the repeated Phe residues on the FG-repeat regions (see below) insert into complementary repeated pockets formed from the crevices between adjacent α-helical repeats all along the outer surface of Kap95p’s spiral. RanGTP binds to Kap95p (Lee et al. 2005) on the inner surface of Kap95’s amino-terminal solenoid spiral, which causes conformational changes that lead to release of Kap60p (and cargoes) (Figure 5).

Perhaps it is not surprising that structurally related Kaps can bind to structurally related NLSs, but it also appears that Kaps can recognize more than one type of NLS. For example, while Kap121p was originally shown to bind to noncanonical Lys-rich NLSs (Rout et al. 1997; Kaffman et al. 1998b; Leslie et al. 2002), it, like Kap104p, also imports proteins through rg-NLSs, which are reminiscent of structurally distinct RNA-binding motifs (Dreyfuss et al. 1993; Lee and Aitchison 1999; Leslie et al. 2004) characterized by repeats of Arg and Gly amino acid residues. Moreover, multiple cargo domains exist in Kap114p, and it has been proposed that this Kap is capable of importing multiple cargoes simultaneously (Hodges et al. 2005).

Although some of the cargoes for many Kaps have been defined, there are an estimated 1500–2000 proteins that transit the NPC during their life cycle, and as a field, we have identified only a handful of the cargoes that they each recognize. So, while it has been proposed many times that evolution has likely exploited their overlapping specificities and potential complexity to regulate classes of cargoes by regulating the karyopherins, it remains for the field to more comprehensively define Kap-cargo complexes to demonstrate how much this is the case and to fully appreciate how they may have done so.

Protein export from the nucleus is mediated by at least three β-karyopherins. The first (“classic”) nuclear export signal was defined in vertebrate cells in HIV-Rev protein. Rev binds specifically to unspliced and singly spliced HIV mRNA and ensures that it is exported efficiently. Studies to define this process identified a short leucine-rich region within Rev that is necessary and sufficient for nuclear export. This sequence is recognized by the karyopherin Xpo1/Crm1 (Stade et al. 1997). As it turns out, there are many proteins that contain variants of the prototypical sequence and are exported by Xpo1p. These include the proteins of the 40S and 60S preribosomal subunits (J. H. Ho et al. 2000; Stage-Zimmermann et al. 2000; Moy and Silver 2002); numerous transcriptional or signaling proteins (Ferrigno et al. 1998; Jensen et al. 2001; Menezes et al. 2004; Chang et al. 2006; Martin et al. 2006; Azevedo et al. 2007; Pelaez et al. 2009); key regulators of the cell cycle [Cdc14p (Bembenek et al. 2005)], which control exit from mitosis; and certain small RNAs (Gallardo et al. 2008; Thomson and Tollervey 2010). Xpo1p/Crm1p is also, at least indirectly, required for normal mRNA production and export (Feng et al. 1999; Strasser et al. 2000; Hammell et al. 2002; Dong et al. 2007).

Msn5p has also been shown to act as a nuclear export factor, exporting phosphorylated nuclear transcription factors (Kaffman et al. 1998a; DeVit and Johnston 1999; Gorner et al. 2002; Queralt and Igual 2003; Durchschlag et al. 2004; Ueta et al. 2007), the HO endonuclease (Bakhrat et al. 2008), and Whi5p, the yeast ortholog of Rb (Taberner et al. 2009). A consensus nuclear export signal (NES) for Msn5p has been elusive, but its preference for phosphorylated proteins suggests a role for regulated export. Indeed, regulation of transport provides an exquisite mechanism to control gene expression. Perhaps the best-characterized example of such regulation in yeast comes from studies of Pho4p. Pho4p is a transcription factor that induces the transcription of phosphate-responsive genes. When cells lack phosphate, Pho4p is imported into the nucleus by Kap121p. However, in the presence of excess phosphate, Pho4p is phosphorylated adjacent to its NLS, inhibiting Kap121p binding and consequently its import. In addition, phosphorylation at two different sites promotes the factor’s nuclear export(Kaffman et al. 1998b; Komeili and O’Shea 1999). A similar shuttling activity has been described for Tor1p (a target of the immunosuppressant rapamycin), which is a protein kinase that controls growth in response to nutrients through the regulation of diverse cellular processes in the cytoplasm and nucleus. Interestingly, PolIII transcription is also regulated by transport; Maf1p, a global inhibitor of PolIII, is exported by Msn5p when cells are shifted to conditions that favor growth and high PolIII activity (Towpik et al. 2008).

Surprisingly, mRNA uses a Kap- and Ran-independent mechanism for export (Santos-Rosa et al. 1998; Katahira et al. 1999). Moreover, mRNP assembly and export involve strict surveillance mechanisms to ensure that only fully mature and functional RNPs are transported to the cytoplasm (Palancade et al. 2005; Schmid and Jensen 2008; Skruzný et al. 2009). In addition, there are many different species of mRNA, each potentially with its own particular maturation pathway. This is a topic that has been both extensively researched and comprehensively reviewed, so we will only summarize these findings briefly and refer the reader to these reviews (Rondon et al. 2010; Stewart 2010; Rodriguez-Navarro and Hurt 2011).

Upon being processed and packaged into mRNP particles in the nucleus, the non-Kap transport factors Mex67p and Mtr2p associate with and chaperone the mRNP through the NPC (Figure 6). These factors are co-transcriptionally recruited to the maturing mRNP by a highly coordinated process that couples transcription, post-transcriptional processing, mRNP assembly, and docking to the nuclear basket, beginning in many cases even before the nascent transcript has left the gene; and once again, research in yeast has pioneered much of our understanding of these processes. Thus, co-transcriptional recruitment of the THO/TREX complex to the nascent mRNA of intron-containing transcripts ultimately leads both to correct 3′-end processing and to recruitment of (among other proteins) Yra1p, which in turn recruits the Mex67p-Mtr2p heterodimer (reviewed in Rodriguez-Navarro and Hurt 2011) (Figure 6). Another complex, TREX-2, associates with the mRNP at the NPC, apparently assuring that it is correctly packaged for its journey across the NPC. The Mex67p-Mtr2p heterodimer, by binding both the mRNP and the FG Nups, mediates the actual mRNP transport event, which can be surprisingly rapid; recent work in vertebrates has suggested that the actual translocation event, even for a multi-megadalton mRNP complex, lasts only milliseconds (Grunwald and Singer 2010; Mor et al. 2010). After transiting the NPC’s central channel, the mRNP encounters Dbp5p, an ATP-driven RNA helicase tethered to the cytoplasmic filament protein Nup159p. Regulated by Nup42p-tethered Gle1p and the small molecule IP6 (inositol hexaphosphate), Dbp5p’s action on the exiting mRNP serves to release the transport factors Mex67p and Mtr2p as well as mRNP proteins such as Nab2p, both actions preventing re-import of the mRNP (and thereby helping to confer directionality to export) and preparing the mRNA for translation (Figure 6) (reviewed in Carmody and Wente 2009; Rodriguez-Navarro and Hurt 2011).

First, the NPC defines a tube of defined width and height that connects between the nucleoplasm and cytoplasm. These dimensions delimit the upper size of the transport cargos, defined in vertebrate NPCs as ~35 nm, and, on the basis of morphological maps, are likely to be similar in yeast and other eukaryotes (reviewed in Strambio-de-Castillia et al. 2010). Second, the tube is lined with FG-repeat regions contributed by the ~160 copies of the different types of FG Nups anchored in and around this tube; work in yeast indicated that no ATPases or GTPases are needed as components of the NPC, such that the NPC does not appear to open and shut as a physical gate, but rather behaves as a “virtual” one (Figures 1 and ​and2)2) (Rout et al. 2000, 2003; Peters 2009). Thus, the power for transport is generated in the nucleoplasm and cytoplasm, and the NPC is chiefly responsible for selectivity. On the basis of mapping and deletion mutagenesis experiments, an affinity gradient of FG-binding sites between the nuclear and cytoplasmic faces of the NPC also does not seem to be essential for nuclear transport in yeast (Rout et al. 2000; Strawn et al. 2004). As the general architecture, distribution, and composition of the FG-repeat regions are similar throughout the eukaryote (see above), the mechanism of gating is likely conserved. The FG-repeat regions do not appear to fold into permanent secondary or tertiary structures, and indeed it is likely that they never form such structures. Rather, they appear highly flexible, allowing them both to assume many possible conformations and to dynamically switch between those conformations. Because they are unfolded, the FG-repeat regions fill a volume many times that of a folded protein of the same size. This means that they can extend tens of nanometers from their anchor point, such that the central tube is flanked by, and filled with, filamentous FG repeats, accounting for the “cloud” of filaments seen to surround the yeast NPC by electron microscopy (Fahrenkrog et al. 2000a; Kiseleva et al. 2004). Another advantage of disordered filaments as binding sites is that only a little protein is needed to fill a lot of volume—a very economical way of having a small amount of protein generate a huge binding site. As stated above, transport factors bind FG-repeat regions, and it is through this binding that they are allowed selective passage through the central channel. Regardless of their differing atomic structures, it seems that all transport factors carry numerous copies of surface-accessible hydrophobic pockets into which several of the F residues of an FG-repeat region can bind (see above). These appear to have low affinity and rapid exchange rates, although as there are several such interactions per transport factor (at least 14 in the case of the karyopherin transport factor Cse1), the avidity of transport factors for FG Nups is expected to be high (Isgro and Schulten 2005, 2007). In a sense, then, the FG repeats can be thought of as antennae, reaching out in a cloud of binding sites many tens of nanometers from the nuclear and cytoplasmic faces of NPCs to efficiently funnel transport factors and their associated cargoes into the NPC, while generating a zone of exclusion for nonspecific materials around the NPC (Figure 1) (Rout and Aitchison 2000, 2001; Rout et al. 2000, 2003; Macara 2001).

How does it actually work? We still do not know, but attempts have been made to describe the basic physical principles of NPC-mediated gating although this has been done without a detailed description of FG Nup behavior, considering only the consensus features of the NPC and making some basic physical assumptions, thus treating the NPC as a narrow hole lined with binding sites and allowing that molecules access and transit this hole through normal diffusion. It has been shown that a narrow channel filled with FG-repeat regions presents a significant barrier to passage across the NPC, such that the probability of a macromolecule translocating through the channel is low. However, transient trapping by a macromolecule that can bind to the FG repeats (such as a transport factor) increases the probability of that molecule remaining in the central channel and thus enhances its transport through the channel (Zilman et al. 2007, 2010). Such explanations are similar to those applied successfully to account for the transport properties of other channels (e.g., Berezhkovskii and Szabo 2005; Berezhkovskii and Bezrukov 2005). More elaborate analyses consider some of the proposed biophysical properties of the FG-repeat regions or invoke others (e.g., Bickel and Bruinsma 2002; Kustanovich and Rabin 2004). Molecular dynamics simulations are also beginning to shed considerable light on the likely behaviors of FG-repeat regions in the NPC (e.g., Miao and Schulten 2009, 2010), but the sheer complexity of computationally simulating this system remains a significant challenge. However, the fact that a narrow hole filled with a selective polymer is, in principle, all that is needed at the NPC for gating has been demonstrated by chemical analogs (Caspi et al. 2008) and, importantly, by a nanochannel filled with FG-repeat regions from yeast that exhibited selective passage of transport factors over control proteins and even transport of a cargo-carrying karyopherin (Jovanovic-Talisman et al. 2009). In this system, gating was exhibited without any other proteins, including an energy-regenerating system; thus gating in principle requires only the FG-repeat regions.

Although perhaps not needed to understand many of the basic principles of NPC-mediated gating, a full understanding of transport will ultimately require an understanding of how the FG Nups behave at the molecular level. Indeed, in the absence of any folded structure, the question of the physical form and behavior of the FG repeats in and around the NPC’s central channel comes down to what is the precise balance of the FG-repeat regions’ intramolecular and intermolecular forces. Cohesive forces have been measured between at least some types of FG-repeat regions and even within individual repeat regions (Krishnan et al. 2008). Conversely, repulsive forces resulting from entropic exclusion—the tendency of the Brownian motion of a disordered polymer to sweep away other molecules from its vicinity—has also been measured for FG-repeat regions (Lim et al. 2006b). The distribution of FG-repeat types has recently been cataloged and characterized extensively, showing that the FxFG (and similar) repeat regions are characterized by being highly charged and in vitro adopt dynamic, extended-coil conformations whereas the low-charge-content GLFG regions have been reported to form more globular, collapsed coil configurations in vitro (Yamada et al. 2010). Nsp1p, Nup159p, Nup1p, Nup60p, and Nup2p carry mainly charged (FxFG) regions while Nup100p, Nup116p, Nup145Np, Nup57p, Nup49p, and Nup42p carry mainly uncharged regions; however, many of these FG nucleoporins, although predominantly featuring one type, actually have both types of regions next to each other, such as Nsp1p, which has an ~190-amino-acid GLFG-like amino-terminus followed by an ~430-amino-acid region that is canonically FxFG-like (Yamada et al. 2010). The extended regions would tend to push away from each other and are predicted to be more mobile, while the more compact regions would tend to be cohesive and less mobile (Ader et al. 2010; Yamada et al. 2010). Thus, if intermolecular attraction forces dominate—even rigid, amyloid-like interactions as measured in vitro for these proteins—then the FG repeats would form a gel; such a “hydrogel” barrier has been proposed, with the central channel filled with a gel of FG Nups cross-linked by intermolecular cohesion, excluding nonspecific molecules by sieving but through which the transport factors dissolve a tunnel by binding to the FG repeats and thereby unzipping them (Ribbeck and Gorlich 2002; Frey et al. 2006; Frey and Gorlich 2007; Mohr et al. 2009; Ader et al. 2010). If repulsive forces such as entropic exclusion dominate, then the central channel is filled with a polymer brush. The repulsive forces exclude nonspecific molecules but can be nullified by the binding forces of transport factors (Lim et al. 2006b, 2007a,b, 2008; Rout et al. 2000, 2003). If intra- and intermolecular cohesions are modulated by entropic exclusion effects, then various “hybrid” models are possible. One such model is the “reduction in dimensionality” (“oily spaghetti”) model that suggests that the FG-repeat regions are more compact, such that they form a layer around the inner walls of the central channel (Macara 2001; Peters 2005). Because they bind, transport factors can enter this layer, allowing them access to the entire central channel, while nonbinding molecules can access only the narrow middle region of the channel, which is devoid of FG repeats. In addition, in vitro measurements led to the suggestion that the reversible binding of transport factors causes the FG repeats to collapse and open a path (Lim et al. 2006b, 2007b, 2008). Similarly, it has been proposed that the more compact and more elongated FG-repeat regions are arranged in such a way as to form an organized but highly dynamic structure, forming a tubular density within the central region of the NPC (Yamada et al. 2010), and perhaps accounting for the “central transporter” as seen in electron micrographs and tomographic reconstructions of the yeast NPC (Yang et al. 1998; Kiseleva et al. 2004). Different transport factors would have access to different portions of this transporter, allowing them to pass by each other to cross the NPC, while entropic exclusion effects prevent nonspecific molecules from crossing (Akey 2010; Yamada et al. 2010). Certainly, there is significant evidence that the different classes of FG repeats are spatially segregated in the yeast NPC (Rout et al. 2000; Alber et al. 2007a,b).

Interestingly, by taking advantage of the molecular genetic techniques available to yeast, a set of targeted deletions were made of the FG regions of FG Nups, either singly or in numerous combinations. These experiments showed that certain of these deletion combinations influenced the transport of only particular transport factors (Strawn et al. 2004; Terry and Wente 2007), and preferences of transport factors for certain Nups has been indicated in living yeast (Marelli et al. 1998; Makhnevych et al. 2003). This preference might manifest as spatially distinguished pathways because recent EM studies in yeast have suggested that mRNPs traffic mainly through the center of the NPC’s channel whereas karyopherins pass through the channel’s periphery (Fiserova et al. 2010). Thus, the different types of FG repeats might mediate multiple but physically and functionally segregated transport pathways through the NPC, perhaps so that the passage of one transport factor through the central channel does not adversely affect another (Fiserova et al. 2010; Yamada et al. 2010). Similarly, the observed asymmetric distribution of certain FG Nups along the nucleocytoplasmic axis of the NPC (Rout et al. 2000; Alber et al. 2007b), although not required for transport (Strawn et al. 2004), may aid in biasing the directionality of transport by providing a high-affinity binding site at the far end of a transport factor’s route through the NPC, trapping a transport factor and its cargo and preventing it from returning through the NPC until the transport reaction is terminated (below) (Rout et al. 2000; Rout and Aitchison 2001; Gilchrist et al. 2002; Gilchrist and Rexach 2003; Pyhtila and Rexach 2003; Strawn et al. 2004; Zilman et al. 2007). Such an idea agrees with affinity measurements made in vitro, which suggested a correlation of the position of asymmetrically located FG Nups with their affinity for certain transport factors (Gilchrist et al. 2002; Gilchrist and Rexach 2003; Pyhtila and Rexach 2003).

A major factor in the efficiency of cellular processes is the rate at which macromolecules in a given process can interact. As the cell size goes up, so the rate of diffusion of a macromolecule across the cell exponentially decreases (Fick’s law), potentially decreasing the efficiency of cellular processes. This effect can be offset by increasing the local concentration of macromolecules that are part of the same process either by compartmentalizing them (as in eukaryotic internal membranes) or, as here, by providing anchored binding sites. The binding sites for accessory transport factors found on the yeast NPC seem less numerous and diverse than those on vertebrate NPCs; the vertebrate Nup358, carrying up to half a dozen such sites (reviewed in Wente and Rout 2010), is absent in yeast. This may be because vertebrate cells are usually significantly larger than yeast cells.

The nucleus is a complex, highly organized subcellular compartment, and the chromatin contained within the organelle is not uniformly distributed. In yeast and metazoans alike, heterochromatin often appears concentrated at the nuclear periphery under the nuclear envelope. In many eukaryotes, heterochromatin is readily observed as electron-dense material in electron micrographs. In yeast, regions of silent chromatin associated with the nuclear periphery include telomeres and the mating-type loci (Cockell and Gasser 1999), and silencing of these domains requires peripheral localization. Indeed, repositioning normally repressed telomere proximal genes to distal positions along chromosomes leads to their inappropriate expression (Maillet et al. 1996), and physically tethering transcriptionally active genes to the nuclear periphery induces gene silencing (Andrulis et al. 1998). It is proposed that the peripheral domain of the nucleus is rich in Sir proteins—silencing factors that promote deacetylation of histones and alter chromatin structure (Gotta et al. 1996; Maillet et al. 1996). Thus, genes that are positioned near telomeres, which themselves are generally tethered to the periphery, are exposed to high concentrations of histone deacetylases and are silenced, while genes distal to telomeres are potentially active. How chromosomal regions come into association with the periphery has been the subject of some debate, but NPC proteins are certainly central to the picture (Figure 6); the basket proteins such as Nup60p, Mlp1p, and Mlp2p, and the mobile nucleoporin, Nup2p, function in the repression of the HMR locus and other subtelomeric genes (Galy et al. 2000; Feuerbach et al. 2002; Dilworth et al. 2005).

In addition to their roles in the formation and maintenance of the peripheral silencing apparatus, NPC components are also linked to the activation of gene expression (Figure 6). For example, while ChIP-chip experiments have revealed that Nups associate with silent mating-type loci and subtelomeric genes, Nups have also been detected in association with active genes (Brickner and Walter 2004; Casolari et al. 2004; Cabal et al. 2006; Taddei et al. 2006; Sarma et al. 2007). NPC-associated genes have been shown to be enriched with the binding site for the transcription factor Rap1p, which, together with its coactivators Gcr1p and Gcr2p, is proposed to link transcriptional machinery to the Nup84 complex (Menon et al. 2005) in a process termed “reverse recruitment.” Furthermore, highly expressed genes appear to be recruited to the periphery when activated (Brickner and Walter 2004; Casolari et al. 2004; Cabal et al. 2006; Taddei et al. 2006; Sarma et al. 2007). Once localized to the periphery, this positioning is maintained to ensure the potential for rapid reactivation of the gene, providing an important advantage to cells exposed to changing environmental conditions. This transcriptional memory, which can persist through several yeast cell divisions, involves the incorporation of the histone variant, Htz1p, into nucleosomes. Htz1p is known to be required for the rapid and robust activation of many environmentally responsive genes (Guillemette et al. 2005; Zhang et al. 2005; Wan et al. 2009) and, in the context of the NPC, functions to retain genes at the periphery and to promote their reactivation (Brickner et al. 2007; Brickner 2009).

Further obscuring the distinction between the gene-activating and gene-repressing functions of the NPC, some NPC components, for example, Nup2p, have been shown to harbor “boundary activity.” This activity is defined by the ability of a protein, when bound to DNA, to prevent the spread of transcriptional repression into adjacent genes by a mechanism that, at least in the case of Nup2p, involves chromosome association with the NPC (Ishii et al. 2002). In this case, the NPC appears to both prevent and promote gene expression, and the nuclear periphery appears to define different territories—both gene activating and repressing. Reconciling these apparent contradictions, it is attractive to think that the association of chromatin with NPCs represents a transient interaction, perhaps reflecting a role for the NPC in transitioning between states, and that these transcriptional states are further defined by chromosome movement between subnuclear territories, a model supported by observations associated with Nup2p (Dilworth et al. 2005). Nup2p has been shown to function in establishing boundaries between active and inactive genes, in transcriptional memory, and in transitioning genes between active and inactive states (Ishii et al. 2002; Dilworth et al. 2005; Brickner et al. 2007) but, unlike typical nucleoporins, Nup2p is not stably associated with NPCs. Rather, this protein transits between the nuclear interior and NPCs, suggesting that Nup2p-dependent recruitment and/or maintenance of genes at the periphery does not dictate stable NPC association. This emerging view is exemplified by recent work in higher eukaryotes, where non-NPC-associated pools of certain nuclear pore complex proteins were shown to associate with, and aid in, the expression of transcriptionally active genes deep within the interior of the nucleus (Capelson et al. 2010; Kalverda et al. 2010). Thus, a direct association with the NPC at the nuclear periphery is not a requisite element in mechanisms of gene regulation by NPC components.

The function of NPCs and nuclear basket-associated structures in chromatin organization is not limited to transcriptional regulation. The myriad of physical and genetic interactions observed between DNA repair factors, components of the Nup84 complex of the NPC and the Mlp proteins in yeast indicate that this nuclear subdomain also acts in the maintenance of telomere length and repair of DNA double-stranded breaks (Galy et al. 2000; Zhao et al. 2004; Zhao and Blobel 2005; Therizols et al. 2006; Palancade et al. 2007; Nagai et al. 2008). Taken together, this convergence of functions at NPCs—in the regulation of transcription, in the stabilization and repair of DNA ends, and in transcript quality control through mRNA surveillance—suggests that the NPC serves as a nexus to inextricably link these processes (Figure 6) (reviewed in Strambio-de-Castillia et al. 2010).

We now have an impression of the NPC as a large multimeric structure that not only plays a pivotal role in transport but also operates to facilitate nuclear organization, gene expression, and chromosome maintenance. In many cases, these functions are positionally linked to NPCs and the extended structures emanating from the nuclear basket, but it is becoming increasingly clear that NPC components exert control far beyond this nuclear microenvironment. A well-studied means by which NPCs exert far-reaching control is through regulated sequestration. The most well-understood examples of this mechanism come from the study of cell cycle regulation in yeast. The overarching theme in these mechanisms is that NPCs and their associated structures provide a scaffold at the nuclear periphery capable of acting as a switchable molecular reservoir, or sink, for proteins temporally regulated through the cell cycle. One example is the cell cycle regulation of septin sumoylation at the bud neck in yeast. Septins are sumoylated by Siz1p prior to anaphase and desumoylated by Ulp1p at cytokinesis (Johnson and Blobel 1999; Makhnevych et al. 2007). Septin sumoylation by Siz1p occurs earlier in mitosis, preceding desumoylation by Ulp1p, and is also controlled by sequestration; however, in this case, Siz1p is sequestered in the nucleus by NES masking. Phosphorylation of Siz1p during mitosis renders the protein competent for nuclear export by Msn5p/Kap142p and, as a result, Siz1p relocalizes to the cytoplasm where it mediates septin sumoylation (Makhnevych et al. 2007). Ulp1p is normally sequestered at NPCs, but is specifically released during mitosis, allowing temporally regulated desumoylation of septins at the bud neck (Takahashi et al. 2000; Makhnevych et al. 2007).