The central plug, also known as the central transporter, within the NPC, is a relatively early development in the quest to resolve the inner structure of the NPC. First described by Unwin and Milligan in 1982 as a result of a cryo-EM study on Xenopus oocytes, the central plug was shown to be a large, spheroidal particle residing in the very center of some of the pores, possessing a diameter of approximately 35 nm (Unwin and Milligan, 1982) (Figure 2A). This structure has been both mysterious and controversial, leading to much debate about its composition and function. While the exact nature of the plug has been unclear, the consensus, however, has been that it is a dynamic structure that presents itself only under certain conditions (Reichelt et al., 1990; Beck et al., 2004; Schwartz, 2005; Lim and Fahrenkrog, 2006). As a result, more recent studies using cryo-EM or similar techniques tend to omit the plug and present the structure of the scaffold by itself (Eibauer et al., 2015; Von Appen et al., 2015; Zhang et al., 2020; Zila et al., 2021). Speculation has arisen suggesting that the presence of plugs could potentially stem from the entrapment of sizable cargo while in transit during imaging procedures (Stoffler et al., 2003). However, a study by Li et al. employed cryo-electron microscopy (cryo-EM) for visualizing pre-60S particles ensnared within yeast NPCs. Intriguingly, these substantial particles exhibit a predilection for tracing the periphery of the pore structure, veering away from a central trajectory. This observation holds the promise of tempering the plausibility of pre-60S particles being the probable origin of the observed plug phenomenon (Li et al., 2023).

A possible mechanism of transport utilizing the plug was described by Talcott and Moore (1999). According to their description, the central plug acts as a barrier that prevents passive diffusion from occurring through the middle of the central channel of the pore, forcing small particles instead to traverse peripherally. This passive diffusion occurs through eight symmetric channels that are each ∼10 nm in diameter and surround the center of the pore. Due to filling nearly the entire interior of the central channel, the plug also acts to control facilitated diffusion, blocking the transit of large particles and providing some form of substrate or channel with which signal-mediated import and export can occur. In this model, facilitated transport would cause reversible displacement or deformation of the plug, which may assist in explaining why the pore has been observed to vary between images of individual pores. However, the model does not provide a plausible explanation for whether the shape change requires the NPC itself to use energy or whether the NPC’s structure is flexible enough to allow sufficiently large molecules to pass through. Thus, additional research is needed to elucidate the mechanisms by which the NPC undergoes shape changes and facilitates the transport of larger molecules as well as to determine the energy requirements associated with these processes.

Beck et al. (2004), while studying the NPCs of Dictyostelium discoideum using cryo-ET, not only showed the variability of the central plug but also provided some new structural insights (Beck et al., 2004). Their findings indicate that the plug is not a single spheroidal density, but rather two overlapping densities. The smaller one is in the same plane as the cytoplasmic filaments, and the larger one is closer to the center, slightly biased toward the nuclear side of the pore. In a 2016 paper, Sakiyama et al. utilized HS-AFM to analyze the spatiotemporal dynamics of X. laevis NPCs (Sakiyama et al., 2016). It was found that within this population of NPCs, approximately 40% contained a “plug-like” feature, which they have stated is likely cargo trapped in the pore. In addition, they characterize a “central plug/transporter” (CP/T) which is also a centrally located object; however, it is distinct from the “plug-like” features. Analyzing their dynamic data by plotting the mean squared displacement (MSD) in the z-direction over time, they found that the FG-Nups inside the pore behave as tethered polypeptides, reaching MSD saturation. The z-diffusional limit was found to be approximately 0.9 nm for FG-Nups near the edge of the pore, while for those near the center, it was approximately 1.6 nm. The authors indicate that as a result of the increased mobility of FG-Nups closer to the center of the pore, the identity of the CP/T is likely to be the result of averaging this dynamic motion, creating what appears to be a static component. This conclusion was further explored in a 2020 paper by Mohamed et al. who also utilized HS-AFM to study the central plug, coming to a similar conclusion that the central plug is at least partially composed of FG-Nups (Mohamed et al., 2020). They described the behavior as FG-Nups forming transient “knots” among themselves in the center of the pore and found that the plugs in cancerous HCT116 cells had a greater extent of conformational dynamics than non-cancerous colon cells. The conception of the central plug architecture can be attributed to the inherent constraints of cryo-EM, primarily its necessity to arrest samples in a frozen state. The plug model was conceived as an initial endeavor to elucidate the inner workings of the NPC. Despite its vintage, the definitive nature of this putative plug element remains unknown. This model, although aging, has retained its utility as a foundational template, inspiring subsequent explorations that have spurred investigators to ponder profound queries concerning the NPC's composition, structure, and transport mechanisms. These inquiries have set the cornerstone for the evolution of more contemporary models, designed to unravel different facets of this intricate structure. As technology continues to progress, expanding our comprehension of the pore, the underpinning framework of the plug model has undergone refinement, harmonizing with fresh insights to engender novel questions, studies, conjectures, and models. The advent of dynamic, live-cell data holds the promise of potentially unveiling the elusive identity of the plug, finally unraveling a longstanding biological enigma.

The concept of the polymer brush model, also referred to as the virtual gating model, was initially introduced by Rout et al. in October 2003 (Rout et al., 2003). This model stands as an intuitive framework aimed at elucidating the mechanisms underlying the passive, swift, and discerning translocation of cargo facilitated by the FG-Nups. Rooted in biochemical and biophysical thermodynamics, the polymer brush model offers insights into the functionality of the NPC. According to the polymer brush model, the NPC’s central channel is envisaged as an entropic barrier. This barrier serves as the foundation for both mediated passive and facilitated transport through the pore (Shulga and Goldfarb, 2003). Broadly, the model seeks to expound upon the pore’s capacity to regulate the assortment of cargo that can traverse it, based on the principle of electrostatic repulsion. This inherent repulsion gives rise to the barrier-like feature, shaping the pore’s ability to selectively permit certain molecules while excluding others. In essence, the polymer brush model encapsulates the intricate dynamics of the NPC’s cargo transport, offering a cogent explanation rooted in fundamental biophysical principles. Entropy (S) refers to the number of ways the energetic motions of a macromolecule can be distributed. Within the cytosol, a macromolecule can move freely, resulting in high entropy. The central channel of the NPC restricts movement, leading to decreased entropy. Therefore, placing a macromolecule within the central tube comes with an entropic penalty, and a region densely populated with FG-Nups increases that cost by further restricting available diffusion space (Hoh, 1998). A macromolecule needs to enter a “transition state” to pass through the NPC. One way to achieve this is through affinity and binding to the FG-Nups in the NPC. Macromolecules need to bind NTRs to enter a “transition state” and pass through (Rout et al., 2003). Although the rope-like structure of FG-Nups may allow them to move aside to let macromolecules pass through, this also requires energy, so macromolecules that are not bound to NPCs have a very low probability of entering the “transition state,” and subsequently crossing the NE. In this model, cargo smaller than 30 nm can pass through the narrow channel protected by the FG-Nup barrier if they can afford to pay the entropic penalty. With the increase in size of a macromolecule, the entropic cost of passing through the central tube increases and the probability of passage decreases. Beyond a certain size, the probability of passage becomes negligible (Rout et al., 2003) (Figure 3C).

Furthermore, the polymer brush model explains translocation across the NPC using the Gibbs free energy (ΔG) of a system. Gibbs free energy is defined as the difference between the system’s change in enthalpy (ΔH) and the product of its temperature (T) and change in entropy (ΔS) (Eq. (1)). In a simplified consideration of the NPC system, the ΔS across the NE describes the entropic barrier of the NPC, while the ΔH describes the binding energy of macromolecules to the NPC. Passing through a physical barrier such as the pore temporarily reduces entropy, resulting in a positive ΔG and creating an energy barrier to activation, causing the process to be thermodynamically unfavorable (e.g., a nonbinding macromolecule attempting to cross the NPC). For translocation to occur, the ΔG must be decreased below the diffusion energy available to a macromolecule (kT). This can be achieved by using binding energy as compensation to decrease ΔH, flattening the energy landscape and reducing the activation energy of translocation across the NE. Binding also has an entropic term, but the sum of the binding and diffusion entropies can be canceled out by sufficient ΔH. In an optimal scenario, the binding and barrier energies are balanced, allowing a macromolecule to pass rapidly through the NPC with minimal hindrance. ∆ G = ∆ H − T ∆ S . (1)

In the kinetics of binding, the authors note that using binding sites to overcome an entropic barrier has its drawbacks. A binding macromolecule spends time attached to its binding sites, which slows down its overall translocation rate across the NPC. If the process takes too long, translocation becomes excessively slow. Effective binding sites surrounding the central tube must have low enough affinities and high enough off-rates to enable the rapid passage of transport factors through the channel. High-affinity binding sites tend to exhibit low off-rates, which can retain bound macromolecules for an extended period, impeding their passage or even trapping them at the NPC (Rout et al., 2003). The authors propose that the NPC possesses numerous low-affinity binding sites, allowing the central channel to provide sufficient binding energy to lower the entropic barrier without compromising transport speed (Rout et al., 2003). In conclusion, the virtual gating model suggests an entropic barrier constituted by a highly dynamic and non-cohesive polymer brush of FG-Nups with weak FG–FG associations. Large TR–cargo complexes require enthalpic gain from hydrophobic TR–FG associations to overcome the entropic penalty of the excluded volume effect. This model underscores the inherent lack of cohesion among FG-Nups, resulting in their mutual repulsion. This repulsion can be attributed, for instance, to electrostatic interactions occurring between the positively charged segments of FG-Nups (Peyro et al., 2021). This perspective diverges from that of alternative models scrutinized in the subsequent sections of this review, which accentuate the cohesive tendencies of FG-Nups arising from hydrophobic interactions between them.

The term “oily spaghetti” was first published by Ian Macara in 2001 and was used to describe the way that FG-Nups could potentially be arranged within the NPC (Macara, 2001). In the review article, Macara provided an overview of the molecular basis for nuclear transport, in particular the behaviors of NTRs and how they interact with NESs and NLSs to facilitate large cargo passage through the pore. Macara proposed that the FG-Nups form a sieve-like structure dominated by hydrophobic interactions that cargo carrier molecules can dissolve into, thus permitting the facilitated transport of cargo (Weis, 2007) (Figure 2B). This is an early example of a model that focuses on the hydrophobic interactions among FG-Nups that could allow for the formation of a cohesive, yet mobile structure within the pore, as opposed to the more repulsion-focused model of the polymer brush. Assuming the central channel of the pore which allows for the passive diffusion of small particles is approximately 10 nm in diameter, Macara approximated that the FG-Nups could form a spaghetti-like structure that lines the inside of the pore in a 7-nm-thick lining. It is important to note that this model does not present a clear explanation as to why the central channel would be left open for passive diffusion and, therefore, leaves the door open for a few possibilities. Perhaps molecules that are small enough could passively diffuse between the “spaghetti,” or the FG-Nups could reach into the center and block passive diffusion or force it to redirect from the center to occur more peripherally. From the perspective of Gibbs free energy, the idea behind this model is that the ∆ G required for conformational changes in the FG-Nups is low enough that they can move freely; however, due to transient interactions between the FG repeats, the “spaghetti” somewhat clumps together onto the inner wall. The balancing act between random motion and transient interaction is the impetus for facilitating transport through the pore because the cargo–carrier proteins will also transiently interact with the FG-Nups. Because of the random movement assumed by this model, the FG-Nups themselves do nothing to impart directionality to the transport of particles. Directionality is instead attributed to the assembling and disassembling of the cargo–carrier complexes by RanGTP, which produces a ratcheting mechanism where cargo becomes trapped in its destination after its carrier has been removed. This ratcheting mechanism has been displayed by simulation as well as in vivo using single-molecule fluorescence microscopy (Lowe et al., 2010; Mincer and Simon, 2011).

Utilizing super-resolution single-point edge-excitation sub-diffraction (SPEED) microscopy, a study focusing on single-molecule dynamics has unveiled compelling findings (Ma et al., 2012). These findings revealed discrete pathways undertaken by individual particles during passive and facilitated transport through the NPC. The small fluorescently tagged molecules (<40 kDa) including single dyes, dyes bound to dextran, insulin, α-lactalbumin, and GFP showed a very high density in the center of the pore. The smallest particle, single fluorescein dyes, occupied the largest diameter region of approximately 37 nm, while the largest sub-40 kDa construct of Alexa Fluor-tagged GFP occupied approximately 17 nm, and the candidates in between generally occupied a region with decreased diameter with increased molecular mass. The transit of a 102-kDa construct of impβ1 bound to four Alexa Fluor dyes, and a 217-kDa construct of two GFPs, an NLS, and four Alexa Fluor dyes were tracked to measure the spatial densities of TRs and large molecules under facilitated transport, respectively. It was found that in both cases, the particles largely remained peripheral to the center of the pore. The tagged cargo-free Impβ1 was absent in a 23-nm-diameter region in the center, and the complexes of cargo/Impβ1 left an 8-nm-diameter hole and had a higher density closer to the inner surface of the NPC scaffold than those of Impβ1 alone. Discerning the differentiation between passive diffusion zones within the pore’s core and facilitated diffusion along its periphery might align with the discrete pathways proposed by the oily spaghetti model.

Additionally, within the context of the oily spaghetti model, FG-Nups are envisioned as being randomly scattered, forming transient connections without a defined pattern. Subsequent models have expanded on this notion, introducing a more organized representation of hydrophobic interactions.

The hydrogel model of the NPC, also known as the selective phase model, is based primarily on hydrophobic interactions of both FG-Nups with themselves and with cargo passing through the NPC (Figure 2C). This model was first described by Ribbeck and Görlich (2001). In this paper, they investigated the properties of facilitated translocation through the NPC under the selective phase model name, which they later expanded upon and described in 2007 as the hydrogel model (Frey and Görlich, 2007). The general concept of this model is that FG-Nups combine to form a homogenous meshwork, where the hydrophobic FG repeats interact with each other, forming a series of small gaps between them. These gaps allow for small particles to passively diffuse through the NPC, while larger particles must interact with the hydrophobic regions to displace the FG-Nups and facilitate their transport through the pore (Frey et al., 2006; Mohr et al., 2009). The paradox present is that this large cargo that needs to interact with the hydrophobic regions must efficiently pass through the pores while also having sufficient interactions with the FG-Nups to break through the hydrogel barrier. The many hydrophobic interactions between both the FG-Nups and the cargo passing through must therefore be sufficiently strong, as well as weak enough to dissociate and allow for passage. The efficient transport of large cargo is believed to be assisted by the re-association of FG-Nups behind the transiting cargo as it is displaced through the pore.

The hydrogel-forming properties of the yeast FG-Nup known as Nsp1 were investigated both in vitro and in vivo by Frey et al. (2006). An aqueous solution with a millimolar concentration of wild-type FG-repeat domains from Nsp1 was placed in a silicon tube and naturally formed into a transparent gel that could remain stable up to 95 ℃ . The gels could, however, be dissolved using the chaotropic agent guanidinium chloride, which supports the primary cohesive force being a noncovalent interaction between polypeptide chains. Samples of these FG-repeat domains were mutated, with each of the 55 phenylalanine amino acids being replaced by serine, hence removing the hydrophobic contributors to the repeats and replacing them with polar, hydrophilic contributors. These mutated regions, when treated in the same manner as the gel-forming wild-type counterparts, remained dissolved in the solution even at concentrations ∼four times greater than that of the wild-type, which shows that the hydrogel properties of FG-Nups are made possible by interactions with hydrophobic phenylalanine. To test the same mutations in vivo, yeast with their copies of Nsp1 were deleted. The yeast could recover from this lethal mutation if treated with replacement by Nsp1, even if their FG repeats were removed. The mutated version of the protein with the serine substitutions, however, could not rescue the yeast, which shows that even the removal of FG-repeats can be tolerated in vivo, but conversion to more hydrophilic structures cannot be tolerated, displaying the potential cruciality of the hydrophobic meshwork formation. Interestingly, a separate mutation that replaced the phenylalanine with tyrosine residues was still able to form a gel in vitro, albeit a less homologous one, suggesting that hydrophobic interactions alongside potentially stacking interactions could be factors contributing to the behaviors of FG-Nups.

It has also been found that the FG domains of at least 10 Xenopus Nups are capable of forming hydrogels and that Nup98 creates a stricter sieve than other tested Nups (Labokha et al., 2013). Recently, the hydrogel model has been the basis for a study on the effects of dipeptide repeats (DPRs) in the NPC and found that a synthetic hydrogel mimicking FG interactions was disrupted by arginine-rich DPRs, preventing the entry of Impβ (Friedman et al., 2022). These DRPs are associated with C9orf72 repeat expansions, which have been found to be the leading cause of frontotemporal dementia and amyotrophic lateral sclerosis (Smeyers et al., 2021). Both the hydrogel model and the oily spaghetti model imply that the behavior of FG-Nups is primarily influenced by hydrophobic interactions. However, the hydrogel model inherently lacks distinct pathways for passive or facilitated diffusion, given that passive diffusion could potentially occur through any sufficiently large gaps within the FG-Nup structure. Moreover, it is essential to highlight that the hydrogel structure of FG-Nups has been exclusively synthesized under controlled laboratory conditions (in vitro) and has not been observed within a natural NPC environment.

The reduction of dimensionality (ROD) model was proposed by Peters (2005). This model suggests the presence of a dense layer of FG domains that coats the inner wall of the NPC as a continuation of the FG motifs of the cytoplasmic fibrils. Furthermore, the TRs required for facilitating transport are proposed to act as ferries that bind to cargo and stick to the dense layer of FG domains (Figure 3B). The name of this model refers to the idea of reducing the dimensions of transport from 3D to 2D, explaining the movement of TRs as a 2D random walk, through which they randomly traverse this dense FG surface until reaching their exit. The principle of ROD is a method of increasing the rate of reaction, i.e., between an enzyme and ligand. A 2D surface on which a ligand can walk across provides an increased likelihood of it locating and successfully interacting with its enzyme, given that the enzyme is hosted on the same surface. Conceptually, transport complexes in the NPC can be thought of as ligands, and the nuclear or cytoplasmic exit from the pore is the target enzyme (Adam and Delbrück, 1968). The ROD model portrays the structures and actions of the FG-Nups by proposing that they coat the inner surfaces of the pore, establishing a foundation for the erratic mobility of TRs. Additionally, the model suggests the creation of a narrow central pathway within the pore for transpiration of passive diffusion. As per this model, passive diffusion is facilitated through an 8–10-nm-diameter space at the core of the pore, encircled by a loosely interconnected network of hydrophilic polypeptide chains. This network functions as a selectivity barrier, impeding the passive diffusion of larger molecules. Different from previous models, the ROD model does provide a possible explanation as to how passive and facilitated diffusion would be spatially distinct based on the proposed collapse of the FG-Nups. However, it does not fully explain why the inner diameter that permits passive diffusion would have a size of 8–10 nm and not larger as a result of the FG-Nups further retreating out to the periphery.

While the ROD model was proposed as a hypothetical model, a study performed by Schleicher et al. evaluated the potential for the 2D walk behavior of traversing particles predicted by the model in vitro using optical trapping and single-particle tracking (Schleicher et al., 2014). The experimental design consisted of a layer of surface-tethered Nup153 being moved toward an optically trapped colloidal probe coated with Impβ and using a photonic force microscope to measure the probe’s movement with nanometer-microsecond spatiotemporal resolution. The experiment was performed with different concentrations of impβ solutions, ranging from 0.5 to 30 µM. It was found that the “jump-to-contact” forces gradually decreased as the environmental concentration of impβ increased, along with an increase in 2D mobility of the probe. At 30 μM, which is a physiologically reasonable concentration of importins (Paradise et al., 2007), the probe was shown to move across the X-Y plane nearly unhindered in a 2D walk while remaining bound. The assumed causes of these effects, and the general effect of TRs in the pore, are explored further in the next section of the paper. Ultimately, this study has shown that under certain cellular conditions, the ROD model could accurately predict the type of movement imparted upon receptor-dependent cargo in the NPC.

In 2010, Yamada et al. published a study performed on purified Saccharomyces cerevisiae Nups that described the heterogeneous nature of FG-Nups, specifically proposing that there are two distinct types of FG-Nups, each with different chemical properties and spatial arrangements reminiscent of a “forest” (Yamada et al., 2010). Some FG-Nups appear in the form of a collapsed coil with low charge content near the NPC anchor domain, while others contain extended coil regions of higher charge content with collapsed coil or folded globule on the end opposite to the anchor domain (Figure 3D). The first set of FG-Nups, with collapsed coils adjacent to the anchor domain, are reminiscent of “shrubs,” whereas those that possess extended coils appear as “trees,” hence the name “forest model.” Because of their charged, extended coil regions, the “trees” possess more flexibility than the “shrubs,” which are more constrained to the inner surface of the pore. This model represents a synthesis of some of the properties described in previous models. While the polymer brush emphasizes electrostatic repulsion and non-cohesiveness and other models such as the hydrogel model emphasize cohesive hydrophobic interactions, the forest model follows both models to produce a more complete picture of the NPC.

In general, the “shrubs” tended to be less than ∼10 nm in hydrodynamic diameter, while the “trees” were found to be around double that, with their collapsed coil regions making up a little more or less than half of that value depending on the particular Nup. It was determined that the best way to predict whether an FG-Nup is a “shrub” or “tree” is by the ratio of the charge to hydrophobic amino acids. Nearly all the collapsed coil FG-Nups analyzed had a charged/hydrophobic amino acid ratio of less than 0.2, whereas all FG-Nups classified as relaxed or extended had ratios ranging from 0.7 to 1.4. One of the most important aspects of this model is the spatial separation of the largely hydrophobic FG regions and the heavily charged regions. Because the collapsed coil regions tend to aggregate with each other as a result of the hydrophobic effect, and because the “shrubs” and “trees” have their collapsed regions near the periphery and near the center of the pore, respectively, a distinct new region is formed that has not been described by other models. This new region appears around the radial midpoint from the center of the pore and corresponds to the extended, highly charged regions created by the “trees.” This model, thus, proposes two distinct transport zones, one through the very center of the pore and the other in this newly described region, termed by Yamada et al. as “zone 1” and “zone 2,” respectively. It is important to note that although this was the first study to mention the forest model by name, it was not the first to suggest that the NPC possessed multiple gates formed by different properties of FG-Nups. Patel et al. (2007) found evidence to support a two-gate model of the NPC. One of the gates they proposed was formed by hydrophobic attractions that formed a selective phase barrier in the center of the pore, as described by the hydrogel model, while another region contains a non-cohesive virtual gate more peripherally localized (Patel et al., 2007). This description seems to point to the same behaviors of FG-Nups as described by the forest model.

Liashkovich et al.(2012) used AFM to investigate the binding patterns of two different particles in X. laevis oocyte NPCs to resolve whether or not the forest model may be accurate (Liashkovich et al., 2012). The first particle was wheat germ agglutinin (WGA), which interacts with N-acetylglucosamine-modified (GlcNAc) Nups such as Nup62, and the other was the mutated TR Impβ^45–462, which specifically binds to FG regions. While the WGA was found near the periphery of the pore, the Impβ^45–462 was found primarily in the center of the pore rather than the periphery, suggesting the presence of the “trees” as described by this model. In addition, this study found that despite the WGA binding primarily to the periphery of the pore, it provided a stronger barrier for passive diffusion than the Impβ^45–462, suggesting the presence of the alternate transport route described as “zone 2” in the forest model.

Eibauer et al. (2015) paired cryo-ET with sub-tomogram averaging on the NPC scaffold of the X. laevis oocyte and showed a structure with the proposed central and peripheral transport routes as described by the forest model (Eibauer et al., 2015). In addition, coarse-grained simulation work studying the NPCs of S. cerevisiae found that longer FG-Nups can simultaneously form a dense FG domain and charged disordered region, described as a diblock polymer reminiscent of “trees” (Ando et al., 2014). It was then found that additional low-charge, FG-rich single-block Nups in the simulation could conform into a “shrub” shape and help stabilize a more open conformation in the presence of large cargo in the pore’s center (Ando and Gopinathan, 2017).

The gradient model, also known as the affinity gradient model, was proposed by Ben-Efraim and Gerace (2001). According to this model, different Nups are localized at different positions within the NPC, forming an affinity gradient for TRs and cargo along the transport pathway. As cargo passes through the NPC, it encounters these different Nups with increasing binding affinity, which is the impetus for directional transport (Figure 3A). Other models have assumed that the FG-Nups impart no directionality on transport, but that directionality is created by concentration gradients between the nucleoplasm and cytoplasm and the RanGTP cycle. The gradient model represents a bold departure from this assumption, proposing that FG-Nups play a role in directionality beyond random, transient interactions. While it does seem to be generally accepted that an affinity gradient is not the driving force of directionality, there may be particular cases in which it does play a role, as is explored later in this section.

In their study, the researchers measured the affinity of import complexes for Nups that interact with Impβ, namely, Nup358, the Nup62 complex, and Nup153. These Nups were selected based on their localization within the NPC, representing early, intermediate, and late binding sites, respectively. The movement of Impβ through the NPC involves sequential transfer from Nup358 to the Nup62 complex and then to Nup153. This directional bias is attributed to the increasing affinity of the transport complex for the Nup binding sites encountered sequentially; although the transfer between Nups can occur in both forward and backward directions, the increasing affinity for more distal Nups promotes unidirectional transport (Ben-Efraim and Gerace, 2001). The release of the cargo complex from the terminal Nup binding site is mediated by RanGTP (Görlich et al., 1996).

It has been proposed that the transfer of transport complexes between two Nups may involve a cooperative reaction, where an input cargo complex bound to one Nup is induced to release from its binding site upon interaction with a second Nup (Ben-Efraim and Gerace, 2001). Researchers have found that different regions of Nup358 and different subunits of the Nup62 complex bind to Impβ with similar affinity. However, they were unable to provide experimental data to support this theory due to the inability to measure the individual interactions between Nup153, Nup62, Nup58, and Nup358 fragments.

Azimi et al. designed an agent-based modeling (ABM) simulation to analyze the interactions between Impβ and FG-Nups and tested whether or not the affinity gradient could significantly improve the transport rate (Azimi and Mofrad, 2013). An ABM is essentially a method of simulating interactions between individual objects known as “agents,” and it has been used in a wide variety of applications (Macal and North, 2009). This simulation is based on a simplification that interactions between Impβ and FG-Nups occur with a single probability based on bulk affinity values, and because it is assumed that this binding occurs rapidly and often enough, this simplification is valid. One of the findings of this paper is that in the models with no affinity gradient, the transport rate for Impβ peaks at an affinity of approximately 86.24 µM and decreases at higher or lower affinities. It was also found that transport rates of Impβ are most impacted by the affinity of the nuclear basket Nups, less so by that of central channel Nups, and negligibly by the affinities of cytoplasmic Nups. Ultimately, it was determined that import efficiency is more dependent on the concentration gradient of RanGTP, highlighted by the fact that even a complete reversal of the affinity gradient does not prevent import. However, the affinity gradient does marginally increase import efficiency and, therefore, could still be a regulatory mechanism for the NPC.

A recent paper by Shen et al. presents synthetic, programmable recreations of the NPC, designed by using DNA corrals that can be associated with different Nups (Shen et al., 2023). The study was performed to analyze the nuclear import behaviors of human immunodeficiency virus 1 (HIV-1), specifically the interactions between the capsid and various Nups, because it has recently been shown that these large capsids can pass through the pore intact (Zila et al., 2021). It was found that multiple copies of Nup153 or Nup358 are required for successful import and that the affinity between the capsid and Nup153 is greater than that of Nup358. It was also found that Nup62 tends to self-interact, creating a barrier for the capsid. Even when 32 copies of Nup62 were grafted to the outside of a 60-nm channel, they still managed to come together at the center of the pore. This paper proposes a three-step mechanism of capsid penetration involving an affinity gradient. The lower-affinity interactions between the capsid and Nup358 at the cytoplasmic region allow for docking, and then interactions with the stronger-affinity Nup153 in the nuclear basket of the pore orient the cone-shaped capsid lengthwise, overcoming the barrier provided by Nup62 in the central channel and pulling it inside the NPC. Vial et al. utilized a combination of SMLM and AFM to show that the nuclear basket is flexible enough to fold inward toward the center of the pore (Vial et al., 2023). This could help explain how Nup153 can cling on to the capsid and assist in its import. It may be the case that, even if the gradient model is not the driving force for smaller particles, it plays a critical role in transporting larger particles, especially if they require a particular orientation for successful transport.

The term “dilation model” is introduced here not to supplant existing models but to complement them, enriching our understanding of the NPC as a whole. It is firmly established that NPCs exhibit the ability to expand and contract (Feldherr and Akin, 1990; Kiseleva et al., 1998; Shulga and Goldfarb, 2003; Beck et al., 2004). However, the exact mechanisms behind this behavior and its implications for transport remain subjects of ongoing investigation. Despite this well-established pore behavior, the other models do not incorporate it into their explanations of NPC functionality. Thus, the dilation model fills an essential gap in the puzzle. In a sense, this model extends beyond the others previously described. Rather than concentrating solely on cargo–Nup interactions within the pore, the dilation model focuses on how external environmental factors induce conformational shifts in the NPC (Oberleithner et al., 2000; Mooren et al., 2004; Shahin et al., 2005; Kastrup et al., 2006; Lim et al., 2006; Zimmerli et al., 2021). Notably, the measured diameter of the NPC can even be altered by the techniques used for pore analysis, and purified NPCs often exhibit smaller diameters than those measured in situ (Schuller et al., 2021; Akey et al., 2022).

Calcium has been shown to play a role in regulating the size and shape of the NPC (Moore-Nichols et al., 2002; Erickson et al., 2006); however, it is not clear whether or not this action has a significant impact on transport through the pore (Sarma and Yang, 2011). Cho et al., investigating the regulation of the nucleocytoplasmic transport of truncated ataxin-3 (ATXN3) proteins in Drosophila neurons, found that decreasing intracellular calcium levels can reduce the nuclear accumulation of ATXN3; however, this is likely a result of calcium-mediated regulation of CBP rather than the constriction of the pore (Cho et al., 2022).

In general, it has been observed primarily through AFM and SEM in X. laevis oocytes that in the presence of calcium, the NPC has a wider diameter on either the cytoplasmic side (Jäggi et al., 2003), nuclear side (Stoffler et al., 1999), both (Erickson et al., 2006), and/or displaced central mass (Wang and Clapham, 1999; Moore-Nichols et al., 2002; Mooren et al., 2004). Calcium stores are located within the NE and can be released via inositol triphosphate (IP₃) receptors, and pore conformation can be regulated by IP₃ and other effectors of the receptor such as the agonist adenophostin A or inhibitor xestospongin C (Moore-Nichols et al., 2002; Erickson et al., 2006). For a more in-depth review of the effects of calcium on the NPC, see Sarma and Yang (2011). It has also been shown via AFM that the introduction of 1 mM of ATP has a transient contracting effect on the shape of the NPC, causing the height of the pores to increase alongside a decrease in the diameter (Rakowska et al., 1998). Another molecular effector of NPC shape is CO₂, and it has been shown by AFM that exposure to 5% CO₂ led to significant closure of the pore within minutes, an effect that was slowly reversible (Oberleithner et al., 2000).

Shahin et al. discovered that the steroid dexamethasone (dex), when injected into X. laevis oocytes, causes a two-step mechanism involving the dilation of the NPC (Shahin et al., 2005). In the first step, dex-initiated proteins (DIPs) bind to the NE, which causes dilation of NPCs from approximately 82–110 nm, and then, the DIPs enter the dilated pores. To explore the limits of this effect, the same laboratory reproduced the procedure, this time measuring the effect of dex exposure after longer periods of time (Kastrup et al., 2006). After a period of 5–11 min, dilated pores reached diameters of approximately 140 nm. Even more interestingly, this treatment produced a subpopulation of so-called “giant pores,” some of which reached diameters of approximately 300 nm. Using nuclear envelope electrical conductivity (NEEC) to measure the permeability of small ions through the NPC, it was found that the dex-injected cells had an increased in the NEEC by 125% compared to non-treated and control cells and that the treated cells returned to baseline levels after around half an hour. This extreme dilation behavior could lead to potential advancements in gene therapies because entrance into the nucleus is one of the major barriers to introducing genes into eukaryotic genomes. For example, researchers have attempted to use adeno-associated viruses as a way of passing genes into the nucleus because the large virus capsid can pass fully intact through the NPC (Kotin et al., 1990; Deyle and Russell, 2009); however, the transport efficiency through the pore has been found to be quite low (Kelich et al., 2015). Therefore, treatments that could dilate the pore may be useful for increasing the import efficiency of these capsids. It also could prove useful for treating diseases that have been shown to hinder pre-ribosomal subunit export such as Diamond Blackfan anemia (Léger-Silvestre et al., 2005; Choesmel et al., 2007). Because the pre-40s and pre-60s ribosomal subunits are among the largest cargo to export through the pore, a significant increase in diameter could increase their export efficiency (Zemp and Kutay, 2007; Gerhardy et al., 2014).

Zimmerli et al. utilized cryo-EM to examine the mechanosensitive properties of the NPC relating to cellular stress and nuclear volume (Zimmerli et al., 2021). Under conditions of hyperosmotic shock and cellular energy depletion, the nuclear volume is decreased, which results in loss of NE tension, causing the central channel diameter of the NPC to contract from approximately 70 nm to less than 50 nm, in both cases reducing passive transport as measured by freely diffusing GFP. This study was performed using Schizosaccharomyces pombe cells, and energy depletion was induced via exposure to nonhydrolyzable 2-deoxy-glucose and antimycin A, while osmotic stress was provided by sorbitol treatment. Another recent in vivo study by Pulupa et al. utilized polarized total internal reflection fluorescence microscopy (pol-TIRFM) to study conformational changes in individual NPCs within HeLa and Hap1 cells (Pulupa et al., 2020). Their developed technique utilized the shallow illumination depth of TIRF microscopy to illuminate the bottom of the nucleus, exciting Nup–mEGFP fusion proteins with either parallel (p) or perpendicular (s) polarized light relative to the central axis of the NPC. By measuring the p:s ratio, they determined the conformational change in the pore indicated by orientational changes in the selected Nup fusion protein, which included Nup133, Nup93, Nup58, and Nup54. It was found that Nup54 and Nup58 had orientational shifts after cellular starvation, whereas Nup133 and Nup93 had no significant changes. It was also found that altering karyopherin content in the pore changes the orientation of Nup54, which may indicate that changes in the transport state of the NPC induce changes in the geometry of the scaffold.

The exact molecular mechanisms of pore dilation and contraction are still under investigation; however, research by Melčák et al. has been unfolding the potential mechanisms for such behavior (Melcák et al., 2007; Solmaz et al., 2013; Sharma et al., 2015). Melčák et al. (2007) proposed a mechanism for NPC dilation based on circumferential sliding between two Nup58/45 dimers (Melcák et al., 2007). Each Nup58/45 protomer is folded into a hairpin structure with N and C α helices separated by a short loop; hydrophobic interactions between two protomers form a dimer, and tetramerization occurs via a continuous electrostatic surface with alternating positively and negatively charged amino acids between adjacent, parallel N-helices on opposite dimers. This electrostatic interface could allow for sliding along the helical axes approximately 1.1 nm and, given the eight-fold symmetric arrangement of these tetramers, could permit a channel diameter increase by approximately 3 nm. It was in their 2013 paper that this group introduced the “ring cycle” model, which was later expanded upon in 2015, and is used as an explanation to further explain how pore dilation could occur (Solmaz et al., 2013; Sharma et al., 2015). The central mechanism behind this model lies within interactions between Nup54 and Nup58. In the constricted state, it is believed that Nup58 forms an eight-fold homotetrameric ring around the midplane of the NPC, while Nup54 forms two similar rings above and below the midplane. In the dilated state, the two Nups interact with each other, forming eight Nup54–Nup58 dodecamers in the midplane that can slide against each other to modulate the diameter of the pore. The center of the Nup54 homotetramers contains highly conserved polar residues that can also interact with polar residues along Nup58, both of which provide a level of intermolecular instability that could permit low-energy interconversion between their homo-oligomeric and hetero-oligomeric states. Hence, the “ring cycle” name refers to the process of interconversion between the large Nup58–Nup54 and the three smaller rings of homotetramers (Figure 4B).

The transport receptor model aims to elucidate the influence of TRs on both the structure and behavior of FG-Nups. It is worth noting that the term “transport receptor model” is introduced here, although it primarily represents a synthesis of observations concerning TR behaviors within the NPC. The presence of TRs within the NPC brings about alterations in the architecture of FG-Nups, triggering conformational changes that can impact nucleocytoplasmic transport (Aramburu and Lemke, 2017; Tan et al., 2018). Heightening the concentrations of the TR Impβ has been demonstrated to not only enhance the efficiency of transporting signal-dependent and -independent cargo but can also substantially accelerate transit speeds by at least 10-fold (Yang and Musser, 2006) (Figure 4A).

One group investigating the mechanisms for this behavior is Lim et al., who have greatly advanced the current understanding of the effects of TRs in the NPC across the last two decades (Lim et al., 2006; Lim et al., 2007; Schoch et al., 2012; Kapinos et al., 2014; Wagner et al., 2015; Barbato et al., 2020; Kalita et al., 2022). In a 2006 study investigating the in vitro dynamics, it was demonstrated through AFM analysis of tethered Nup153 that treatment with a less polar solvent (5% 1,2-hexanediol) collapses the FG-Nups, which can be reversed by replacement with PBS (Lim et al., 2006). This collapsing mechanism of Nup153 induced by an increase in hydrophobicity became a basis for the karyopherin-centered model of the NPC. By covalently tethering Nup153 to gold nanodots, Lim et al. effectively produced a simplified NPC-like environment where the effects of karyopherins could be measured by AFM (Lim et al., 2007). They found that like the hexanediol treatments, increasing the concentration of Impβ1 reversibly decreased the reach of the FG-Nups. Concentrations of Impβ1 in PBS from 0 M, 115 fM, 2.5 pM, and 33 nM corresponded to a respective decrease in Nup lengths of 29.1, 17.9, 13.7, and 11.3 nm. Because RanGTP is the mechanism by which impβ is dissociated from FG-Nups in the termination of nuclear import in vivo, it was speculated that treatment with RanGTP could be an effective mechanism for reversing the collapse of these Nups (Kalab et al., 2002). It was found that after treatment with 0.33 nM Impβ, increasing the concentrations of RanGTP from 0 M, 0.1, 0.2, and 0.56 nM corresponded to an increased Nup length of approximately 7.4, 10.4, 16.2, and 34.5 nm, respectively.

A further in situ investigation of Nup62 utilizing surface plasmon resonance (SPR) measurements found that there are multiple phases of interaction between Impβ and these FG-Nups (Schoch et al., 2012). Based on their results, they proposed that Impβ binds to FG-Nups, initially causing a slight collapse, but as more Impβ proteins bind tightly to this layer, the steric repulsion induces the recovery of the Nups, and they extend outward again. In addition, they proposed that after sufficient binding of Impβ, a second layer can stack on top of the first layer with weaker interactions than the first layer, causing a “pile-up” effect. These effects were further investigated using SPR, and it was found that Nup214, Nup62, and Nup153 are all capable of extending to accommodate several layers of Impβ. At physiological Impβ levels, this mechanism can produce a closely packed Impβ layer against the inner surface of the pore, and a looser and more fast-moving layer of Impβ that appears after the first layer is saturated (Kapinos et al., 2014). The proposed mechanism for this is that the first layer of Impβ has many interactions with the FG regions, and the next layer now has fewer potential interaction sites and, therefore, moves more freely. Furthermore, NFT2 was found to have faster transport kinetics than Impβ, which is believed to be facilitated by the occupancy of Impβ already fulfilling many of the potential FG interactions, which would otherwise make transit slower (Wagner et al., 2015).

High-speed super-resolution light microscopy has corroborated the observation that the presence of Impβ1 within the pore can induce a collapse of FG-Nups, thereby modifying both the facilitated and passive transport routes. Notably, at a concentration of 15 μM Impβ1, the permeability for 70-kDa dextran molecules increased by enlarging the central passive diffusion channel. These molecules, which typically struggle to diffuse through the NPC at lower Impβ1 concentrations, experienced enhanced permeation (Ma et al., 2012). Additionally, this methodology has recently enabled the three-dimensional tomography of the FG–Nup barrier, unveiling its interactions with multiple TRs in native NPCs (Ma et al., 2016). The findings suggest that each TR occupies a distinct interaction zone within the FG–Nup barrier. Notably, two key TRs, Impβ1 and exportin 1 (CRM1), surpass other TRs in their binding to FG-Nups. Furthermore, the TRs are capable of modifying the tomography of the FG–Nup barrier, exerting influence on each other’s pathways, particularly in situations characterized by intense competition.

Kalita et al., using SPR to examine the behaviors of Impβ and CRM1 in relation to interactions with FG-Nups, showed that these TRs are required to bolster the inner structure of the NPC and that leakage through the pore was a symptom of low TR levels (Kalita et al., 2022). Silencing of Impβ produced a 16% increase in the nuclear/cytoplasmic ratio of 2xEGFP-NES, which signifies that leakage increased NPC permeability. Further increased levels of Impβ silencing induced cell death. It was also discovered that CRM1 is capable of compensating for a loss of Impβ to still maintain structure in periods of low Impβ content. Both Impβ and CRM1 were found to bind comparably to FG-Nups, and a silencing of Impβ measured by an immunofluorescent signal decrease of approximately 12%–18% in the NE corresponded with an increase in the CRM1 signal of up to 121%, suggesting that CRM1 is being recruited to counteract the lack of Impβ in the pore. Ultimately, the findings contributing to the transport receptor model implicate TRs as being more important to the structure and function of the NPC than previously understood, pointing to TR–FG interactions potentially being comparably critical to FG–FG interactions.