Interaction of nucleoporins with nuclear transport receptors: a structural perspective

1 Introduction

Transport of proteins and RNAs between the cytoplasmic and nuclear compartment is an essential and highly regulated cellular activity facilitated by permeable gates within the nuclear pore complexes (NPCs). NPCs are large multi-protein complexes that perforate the nuclear envelope (NE). They are built from more than 30 different proteins termed nucleoporins or Nups. Small cargoes like metabolites and small proteins may diffuse passively through the NPC, whereas larger molecules require dedicated soluble nuclear transport receptors (NTRs). NTRs mediate a facilitated diffusion, allowing passage of the transport complex with translocation rates of ∼1000 events per second per NPC (Ribbeck and Görlich 2001). This type of transport is often described as “active” in the sense that the cargo can accumulate in one compartment, against a concentration gradient. In nuclear protein import and export, the energy that is required for this accumulation ultimately derives from GTP-hydrolysis on Ran, a small GTP-binding protein that interacts with a subset of NTRs (see below). The interplay of NTRs with their cargo and with certain nucleoporins thus enables a controlled transition through the pore. A detailed understanding of the relevant interactions is key for a mechanistic description of nucleocytoplasmic transport. In this review, we will focus on the interaction of nucleoporins with NTRs. For a recent review on NTR-cargo-interactions and nucleocytoplasmic transport in general see (Wing et al. 2022).

Recent structural analyses using a combination of X-ray crystallography, cryo-electron microscopy (cryo-EM), electron tomography (ET) and structure prediction approaches led to a detailed understanding of the more rigid components of the NPCs in Xenopus laevis, yeast, human, and other species (Bley et al. 2022; Huang et al. 2022; Mosalaganti et al. 2022; Petrovic et al. 2022; Zhu et al. 2022). Despite the overall architectural resemblance of NPCs from different species, they reveal significant variance in the number of constituent proteins and resulting spatial density, most likely due to variations in nucleoporin sequence and localization (Lin and Hoelz 2019). NPCs exhibit an eightfold rotational symmetry with a copy number of at least eight for each nucleoporin per NPC and a total of 500 to 1000 individual proteins (yeast versus human; (Dickmanns et al. 2015; Lin and Hoelz 2019)). The symmetry axis runs through the middle of the transport channel and is perpendicular to a symmetry plane paralleling the nuclear envelope. This second level of symmetry is not absolute, as there are a number of nucleoporins that occur only on the nuclear or the cytoplasmic side of the NPC. These asymmetric nucleoporins are linked to nuclear or cytoplasmic rings and form the nuclear basket and the cytoplasmic filaments, respectively, structures that were suggested to serve as binding sites for nuclear transport complexes before or after NPC-passage (Wälde and Kehlenbach 2010). A central framework, also named spoke complex, spoke-ring complex or scaffold-ring complex consists of eight spokes that on one side attach to a set of proteins that anchor the NPC into the membrane and on the other side enclose the central pore through which nucleocytoplasmic trafficking occurs (Lin and Hoelz 2019).

2 FG-nucleoporins, FG-repeats and their interactions

Only a subset of nucleoporins seems to be directly involved in the formation of the selective gate in the center of the NPC that controls the passage of macromolecules, the so-called FG (phenylalanine-glycine)-Nups. In the most recently determined structural models of the NPC (Bley et al. 2022; Huang et al. 2022; Mosalaganti et al. 2022; Petrovic et al. 2022; Zhu et al. 2022), these Nups (or at least parts of them) are not structurally resolved to the same extent as the more rigid building blocks of the NPC. Typically, they exhibit structural characteristics of “natively unfolded” or intrinsically disordered proteins (IDPs), as they harbor highly flexible stretches of amino acids that lack ordered secondary structure elements, termed intrinsically disordered regions (IDRs) (Denning et al. 2003; Oldfield and Dunker 2014; van der Lee et al. 2014). Hence, these regions are expected to occur in a continuum of possible conformations. This high flexibility of the binding surface of nucleoporins with multiple interaction sites may explain the high on-rates for NTR-binding, close to the theoretical diffusion limit (Milles et al. 2015). One characteristic feature of these IDRs in nucleoporins is the occurrence of FG (phenylalanine-glycine)-motifs. Repeats of such motifs (FG-repeats) were originally identified in the first set of yeast nucleoporins that were described at the molecular level (Davis and Fink 1990; Nehrbass et al. 1990; Wente et al. 1992) and also in one mammalian protein, Nup62 (Starr et al. 1990). Importantly, they mediate the interaction of NTRs with the respective Nups (Moroianu et al. 1995; Radu et al. 1995; Rexach and Blobel 1995). In total, ten FG-Nups have been described in human cells: Nup358, Nup214 and Nup42 are asymmetric nucleoporins located at the cytoplasmic side of the NPC, Nup153 and Nup50 are nucleoporins of the nuclear basket, Pom121 is a nucleoporin containing a transmembrane domain and Nup98, Nup62, Nup58 and Nup54 are nucleoporins of the central transport channel. Being disordered, their FG-regions may expose hydrophobic amino acid residues (phenylalanines in particular), inviting other hydrophobic regions of other Nups or of soluble NTRs for interactions. Three types of FG-motifs can be distinguished by sequence analysis (Rout and Wente 1994): the simple FG-, and the more elaborate GLFG-and FxFG-motifs. Furthermore, FG-repeats are characterized by the spacers between the FG-residues. An analysis of all nucleoporin FG-domains in Saccharomyces cerevisiae indicated that the GLFG-rich domains typically lack acidic residues in the spacer region and are enriched in serine, threonine, asparagine, and glutamine. In contrast, the spacers between FxFGs are enriched in basic and acidic amino acid residues. Simple FG-repeats can have either spacer type (Rout and Wente 1994). Another defining criteria of FG-motifs is their propensity to engage in FG-FG-interactions: cohesive FG-domains (typically depleted of charges) can transiently interact with each other (Patel and Rexach 2008; Patel et al. 2007). As isolated protein fragments, such regions are known to phase-separate, forming a gel-like structure that in situ (i.e. in the context of an intact NPC) may be part of a dynamic meshwork that controls the diffusion and/or transport of macromolecules (Frey and Görlich 2007; Frey et al. 2006). Hence, these FG-domains are at the basis of the permeability barrier of the NPC, with Nup98 being a very important component in vertebrates (Hülsmann et al. 2012). For detailed discussions on the different models that try to explain the biophysical nature of the permeability barrier see (Hoogenboom et al. 2021; Schmidt and Görlich 2016; Wälde and Kehlenbach 2010). Of note, cohesive nucleoporins containing GFLG-domains may also interact with nucleoporins completely lacking FG-motifs (Patel et al. 2007; Schrader et al. 2008) and may stabilize the overall structure of the NPC (Onischenko et al. 2017).

Recent structural analyses of FG-domains of nucleoporins by NMR-spectroscopy have led to a deeper understanding of the biophysical underpinnings of cohesive FG-nucleoporins. Using artificial FG-repeat sequences based on Nup98, FG-FG-interactions could be monitored in solution as well as in phase-separated samples, with a high rotational mobility of the FG-domains (Najbauer et al. 2022). A combination of NMR-spectroscopy and cryo-EM allowed a detailed analysis of reversible FG-FG interactions, again using Nup98 as a paradigm, and revealed insight into the nature of stable and less stable interactions (Ibanez de Opakua et al. 2022). Although Nup98 with its characteristic GLFG-repeats served as a model nucleoporin in these and other studies (Ng et al. 2023), the FG-motif as such does not seem to be essential for phase separation and for the formation of an assembly that allows selective entry of NTR-cargo complexes (Ng et al. 2023). Apparently, phenylalanine residues in the context of other amino acid sequences can promote the formation of a selective barrier. This observation supports the assumption that hydrophobic regions in nucleoporins (i.e. others than the conventional FG-motifs) can also engage in interactions with soluble transport receptors. Other FG-domains seem to be non-cohesive (Patel et al. 2007). They do not interact with each other or with FG-motifs of other Nups, yet they may be available for hydrophobic interactions with NTRs or other molecules that pass the NPC. The different biochemical and structural properties of FG-Nups are thought to result in differences in the overall arrangement of the FG-repeat regions. While cohesive FG-Nups tend to exhibit collapsed-coil configurations characterized by a low charge content, non-cohesive Nups are often highly charged and can adopt a more dynamic, extended-coil conformation (Yamada et al. 2010).

In general, local hydrophobicity seems to be an important feature of proteins or protein complexes that exhibit a facilitated diffusion across the NPC. Indeed, chemically increasing the hydrophobicity of the artificial cargo protein bovine serum albumin allowed its passage through nuclear pores, remarkably without the requirement for a NTR (Naim et al. 2009). Similarly, introducing hydrophobic amino acid residues that would be exposed on the surface of an evolved GFP-molecule as a cargo strongly increased its NPC-passage rates, again without the need for a NTR (Frey et al. 2018). NTRs, on the other hand, can be almost quantitatively removed from a cytosolic lysate using phenyl-sepharose as a hydrophobic affinity matrix (Ribbeck and Görlich 2002), suggesting that they have a greater surface hydrophobicity compared to most soluble cellular proteins. This property explains the early observations that certain NTRs interact with the peptide repeat regions of various nucleoporins (Iovine et al. 1995; Moroianu et al. 1995; Radu et al. 1995; Rexach and Blobel 1995). Interestingly, NTRs can also interact with FG-containing proteins, other than nucleoporins. TDP-43, for example, is a nucleic acid binding protein that is linked to neurological disorders like amyotrophic lateral sclerosis. It contains eight FG-and six FG-like motifs and is prone to aggregation. Aggregates may also include FG-nucleoporins, possibly leading to a disruption of the NPC (Chou et al. 2018; Doll and Cingolani 2022). NTRs like the importin α/β dimer, on the other hand, can bind to TDP-43 and prevent aggregation, suggesting that similar mechanisms govern the interaction of NTRs with FG-Nups and TDP-43.

3 Interaction of nucleoporins with NTRs

Their natively unfolded nature makes the FG-domains of nucleoporins inapplicable for crystallographic analysis, unless they are stabilized by an NTR. Hence, most of our knowledge about FG-domain structures derives from protein complexes containing NTRs. NTRs fall into three classes, all of which have been analyzed at a structural level in complexes with short FG-containing peptides, revealing multiple “classic” or “conventional” interactions as depicted in Figure 1. Some individual sites are described below and depicted in detail in Figure 2.

Figure 1:

FG-binding sites on nuclear transport receptors. (A) Homodimer of NTF2 (PDBid 1gyb; Bayliss et al. 2002a), depicted in different gray tones with six binding sites per monomer. (B) Importin β (PDBid 1qgk; Cingolani et al. 1999) with the superhelical shape of its individual HEAT repeats indicated by a gray line and eleven FG-binding sites indicated in color. (C) Cse1 (PDBid 1z3h; Cook et al. 2005), depicted as above with 14 FG-binding regions. Circles indicate alternative binding pockets (Isgro and Schulten 2007b). (A–C) FG-binding sites as determined by X-ray crystallography (orange), MD-simulation (pink), biochemical experiments (cyan) or by both of the latter two methods (green). See also inset below panel (A) and text for details. (D) Superposition of three CRM1 structures (using PDBid 5dis (human), which also contains a fragment of Nup214 (orange lines and FG sites indicated by aromatic rings) (Port et al. 2015) with a total of 12 FG-binding pockets. RanGTP (green) is localized in the center of the CRM1 ring in all structures. Yrb2 (blue lines and rings) and Nup42 (brown lines and rings) are derived from two alternative structures of yeast complexes (PDBid 3wyf, Koyama et al. 2014, and PDBid 5xoj, Koyama et al. 2017, respectively). (A, B, D) The boxes around some of the crystallographically identified binding sites and their coloring refer to the interactions in the following two figures (NTF2-dimer, see Figure 2A; Importin β, see Figures 2D–F and 3BD–FBFigures 2D–F and 3BFigures 2D–F and 3BFigures 2D–F and 3B; CRM1, see Figure 3A, C, D).

Figure 2:

NTR-binding sites for conventional FG-and FxFG-motifs. (A) Yeast NTF2 interacting with a 9 aa FxFG-fragment (DSGFSFGSK; (PDBid: 1gyb; (Bayliss et al. 2002a)). (B) Human Tap-p15 interacting with a 12 aa FG-fragment (GQSPGFGQGGSV) of Nup214 (PDBid: 1jn5; (Fribourg et al. 2001)). (C) Human Tap (UBA-domain) interacting with a 9 aa FxFG-fragment (DSGFSFGSK; PDBid: 1oai; (Grant et al. 2003)). (D) Kap95p interacting with a 9 aa FxFG-fragment of Nsp1 (PDBid: 1o6p; (Bayliss et al. 2002b)). (E) Kap95p interacting with a 28 aa FxFG-fragment of Nsp1, binding site 1 (PDBid: 1f59; (Bayliss et al. 2000)). (F) Kap95p interacting with a 28 aa FxFG-fragment of Nsp1, binding site 2 (PDBid: 1f59; (Bayliss et al. 2000)). (A–F) The visualization of interatomic interactions was performed in Arpeggio (Jubb et al. 2017) using the standard settings. Ring centers are indicated as black dots, ring interactions are shown in yellow. The other interactions are carbon-π (green), donor-π (blue), methyl sulfur-π (cyan) amide-ring (raspberry), amide-amide (blue), hydrophobic van der Waals (forest green) and undefined van der Waals (magenta). The thickness of the individual dashes refers to the distance of the interactions (the thicker the closer). Only distances within the respective van der Waals radii are depicted. Proximal interactions of any kind that are within a 5 Å radius are colored black. Undefined interactions have been omitted. Hydrophobic surface regions in NTRs are indicated in yellow. See text for details.

(i) NTF2 (Paschal and Gerace 1995) is the characteristic member of the first class of NTRs, a dedicated import receptor for RanGDP (Ribbeck et al. 1998). In the nucleus, GDP on Ran is exchanged for GTP, and RanGTP can then function as a cofactor in nuclear protein import or export, as it binds to all members of the third class of NTRs (see below). Nuclear RanGTP dissociates incoming import complexes, leaving the import cargo in the nucleus. Export complexes, on the other hand, typically form only in the nucleus in the presence of RanGTP. NTF2 forms homo-dimers bearing distinct binding sites for RanGDP and FG-nucleoporins (Chaillan-Huntington et al. 2000; Clarkson et al. 1996; Cushman et al. 2004). A mutant form of yeast NTF2 (N77Y) together with an FG-segment of the yeast nucleoporin Nsp1p has been crystallized. This structure revealed the dimer interface formed by two paralleling β-sheets, one from each subunit (Bayliss et al. 2002a). The phenylalanines are positioned in different pockets of the resulting cleft between the monomers and are bound due to predominantly hydrophobic and carbon-π interactions (Figures 1A and 2A). Only the side chains have been described to closely interact with NTF2, whereas the corresponding main chain atoms and the remaining residues revealed a much higher flexibility, indicating the importance of the phenylalanine side chains for complex formation. This structure confirmed biochemical findings, describing an N-terminally located phenylalanine (Phe5, corresponding to Trp7 in human) as a crucial residue for binding (Bayliss et al. 1999; Quimby et al. 2001; Ribbeck and Görlich 2001). Interestingly, the individual phenylalanine side chains reveal little direct interaction and the only significant one is a methylsulfur-π interaction mediated by Met36 (Figure 2A). In close vicinity of these distinct binding regions within one pocket, an additional one was determined by molecular dynamics (MD)-simulations (Isgro and Schulten 2007a). Another binding site was originally identified to interact with Phe72 from canine RanGDP (Stewart et al. 1998), and subsequently also found to bind to FG repeats in MD-simulations (Isgro and Schulten 2007a). Of the remaining four binding sites on NTF2 that have been identified in the MD-simulations, two had been suggested upon NMR-and biochemical analyses (Cushman et al. 2004; Morrison et al. 2003). The remaining two sites are in in a conformationally flexible region of the NTF2 surface, with one being unique with respect to its surface properties. It is composed of mainly hydrophilic residues and binding of the phenylalanine residue is achieved by the aliphatic part of the side chains, whereas all five others are composed of predominantly hydrophobic residues of the NTR (Isgro and Schulten 2007a). Together, these 12 specific hydrophobic binding spots for FG-repeats are thought to delicately tune the ability of the NTF2-dimer to interact with nucleoporins. As for other NTRs, this large number of potential binding sites implies a rather low affinity of individual interaction, allowing a fast transition of the protein/the protein complex through the transport channel of the NPC. Of note, only peptides containing FG-residues have so far been analyzed in the crystallographic studies and the MD-simulations addressing NTF2. As for other NTRs, most sites identified by these simulations have not been verified experimentally (Figure 2A–C). Hence, the predictive value of MD-simulations for Nup-NTR-interactions still remains unclear.

(ii) Members of the second class of NTRs, such as Mex67p in yeast or TAP (tip-associated protein; also known as NXF1, nuclear RNA export factor 1) in humans (Bachi et al. 2000; Bear et al. 1999; Herold et al. 2000; Segref et al. 1997) are involved in mRNA export. In contrast to protein transport mediated by NTRs of the third class (see below), Ran is not involved in bulk cellular RNA export. TAP is a multidomain protein comprising a nuclear localization signal (NLS), an RNA binding domain (RNA recognition motif: RRM), and a leucine-rich repeat (LRR) domain in the N-terminal half (Liker et al. 2000; Teplova et al. 2011). The C-terminal half of TAP consists of two domains, an UBA (ubiquitin associated) domain (Hofmann and Bucher 1996) and one resembling NTF2 (NTF2-like, NTF2L). The latter domain forms a heterodimer with NXT1 (NTF2-related export protein 1; also known as p15 or Mtr2p in yeast (Katahira et al. 1999; Santos-Rosa et al. 1998)), which also resembles NTF2. The complex formation of TAP/NXT1 has been shown to reduce the affinity of TAP to FG-repeat nucleoporins from nM to µM into a range required for transient interactions, thus enabling efficient and fast diffusion of the complex (Katahira et al. 2002). The bulk of nuclear mRNA is exported into the cytoplasm by this TAP/NXT1 heterodimer, together with a series of associated RNA-binding proteins (Ashkenazy-Titelman et al. 2020).

Structural analyses of the yeast Mex67p/Mtr2 complex revealed that the individual domains form a stable overall fold with an extensive and continuous positively charged surface area for RNA binding on one side of the complex (Aibara et al. 2015). The RRM and LRR of TAP have been crystallized in complex with a retroviral constitutive transport element (CTE)-RNA (Teplova et al. 2011). Furthermore, in yeast NTF2L and Mtr2p have been connected to binding ribosomal subunits in concert with the RRM and LRR domains of Mex67p (Yao et al. 2007). The opposing side of the TAP/NXT1 complex harbors a negatively charged region (Aibara et al. 2015), which comprises the NTF2L-and UBA-domains and has been shown to interact with multiple Nups (Bachi et al. 2000; Blevins et al. 2003; Lévesque et al. 2001; Lévesque et al. 2006; Matzat et al. 2008; Santos-Rosa et al. 1998). The crystal structure of a short FG-peptide derived from Nup214 together with TAP/NXT1 revealed a specific hydrophobic pocket on TAP, with a single exposed phenylalanine side chain of the Nup-fragment protruding into this hydrophobic pocket (Figure 2B) (Fribourg et al. 2001). The adjacent glycine residue was suggested to provide conformational flexibility, allowing the phenylalanine residue to access the hydrophobic pocket of TAP. Interestingly, this pocket is located in the vicinity of the TAP/NXT1 dimer interface in a region similar to the one found on NTF2, where an additional binding pocket was identified by MD-simulation (1A).

Recently, two proteins were shown to use the same binding site in the TAP-NTF2L-domain for modulation of TAP function. First, Sac3, which functions as a scaffold for formation of the TREX-2-complex, carries degenerated FG-repeats (Fischer et al. 2002) that interact with the nucleoporin binding region in TAP and not with ctCRM1 or ctKap104 (Dimitrova et al. 2015). Second, the influenza virus NS1 protein was shown to bind to this FG-binding site of TAP with an exposed phenylalanine residue (F103) located in a loop of its effector domain, thereby inhibiting binding to nucleoporins and also mRNA export (Zhang et al. 2019).

Using the UBA domain of TAP and an FXFG-type peptide for co-crystallization, hydrophobic interactions between both phenylalanines and a distinct hydrophobic pocket in TAP were observed (Grant et al. 2003). In contrast to the binding mode of the identical peptide used in the structure obtained for NTF2 (Figure 2A), the two phenylalanine residues exhibited a direct interaction, stabilizing each other in their orientation (Figure 2C). A highly homologous interaction pattern has been observed for the yeast orthologues (Hobeika et al. 2009). These observations support the notion that phenylalanines in FxFG repeats may either bind in individual pockets or, upon close interaction, in a single pocket.

(iii) The third and largest class of NTRs, the importin β superfamily, encompasses around 20 members in humans, all of which interact with the small GTPase Ran in its GTP bound form and also with nucleoporins (Fried and Kutay 2003). They are generally made up of HEAT repeats (Andrade et al. 2001; Friedrich et al. 2022), which were initially identified in Huntingtin, elongation factor 3 (EF3), protein phosphatase 2A (PP2A) and the yeast kinase TOR1 (Target Of Rapamycin 1). HEAT repeats are arrangements of two nearly anti-parallel α-helices (termed A-and B-helices), usually connected by a short linker (Cook et al. 2007). In NTRs, the stacked arrangement of the 19–24 HEAT repeats results in an overall superhelical shape, with the A-helices located at the outer convex surface of the molecule and the B-helices forming the inner concave side (Monecke et al. 2014). Structural investigations of a multitude of transport complexes showed that cargo and Ran bind at the inside of the superhelix of the NTRs, whereas nucleoporins bind via individual FG-motifs at the convex outer surface of the NTRs, thus not interfering with cargo binding (Bayliss et al. 2000; Bayliss et al. 2002a; Bayliss et al. 2002b; Koyama et al. 2017; Koyama et al. 2014; Liu and Stewart 2005; Port et al. 2015). Proteins of the importin β superfamily can either function as nuclear import receptors, interacting directly with an import cargo or indirectly via an adapter protein, or as nuclear export receptors. A few members of the family may function in both directions. The best-characterized import receptors are importin β (Kap95p in yeast), which functions together with its adapter protein importin α, and transportin. Both NTRs recognize specific sequences (nuclear localization signals, NLS) on their cargo proteins. Similarly, nuclear export sequences (NES) as they occur on proteins that are transported from the nucleus to the cytoplasm, are recognized by CRM1 (also known as Xpo1p in yeast), the best-characterized nuclear export receptor. For a detailed inventory of these NTRs see (Fried and Kutay 2003; Pemberton and Paschal 2005; Wing et al. 2022). Importin α/β, transportin and CRM1 all have a large number of transport cargos, whereas other members of the importin β superfamily tend to have a more restricted set of clients (Kimura et al. 2017; Kirli et al. 2015; Mackmull et al. 2017; Thakar et al. 2013). With respect to interactions with nucleoporins, structural data is so far only available for importin β and CRM1. Furthermore, MD-simulations were performed for Cse1p, the yeast homologue of CAS, a dedicated export receptor for importin α (Figure 1C) (Isgro and Schulten 2007b). Therefore, we will focus on these NTRs in this review.

For importin β, multiple FG-binding sites have been identified over the years, first experimentally (Bayliss et al. 2000, 2002b; Bednenko et al. 2003; Liu and Stewart 2005) and later also in MD-simulations (Isgro and Schulten 2005) (Figure 1B). The latter approach used a variety of FXFG-containing peptides and suggested nine binding sites on importin β, some of which had been revealed before, and were later further analyzed by atomic force microscopy (Otsuka et al. 2008). Interestingly, during the course of the MD-simulation, the FG-peptides either contacted one of those binding sites on importin β or interacted with each other, as expected for a cohesive domain. As shown in Figure 1B, these sites mostly occur on the convex outer surface of importin β. In order to exploit X-ray crystallography as a method to localize and visualize binding of simple FG repeats on importin β, experiments were carried out with a short peptide (GLFG; (Bayliss et al. 2002b);) or a longer one containing four FXFG-and one IXFG-motif derived from the yeast nucleoporin Nsp1 (Bayliss et al. 2000), defining binding sites for simple FGs on importin β/Kap95p (Bayliss et al. 2002b) (Figure 2D). The regions involved in FG-binding form hydrophobic depressions or pockets, usually located between two A-helices of two neighboring HEAT-repeats. In two identified regions in Kap95p, the second phenylalanine of the respective FXFG-motif contributed significantly to the hydrophobic interaction. Interestingly, at the first site the first phenylalanine residue engaged in aromatic stacking with the second phenylalanine as well as with residues in the hydrophobic pocket of importin β (Figure 2E). At the second site (Figure 2F), the first phenylalanine did not interact with the second one but rather with a different patch of the binding site. These structures reveal weak or no interaction of the glycine residues, which are sometimes not even resolved in the electron density map, suggesting that a glycine residue C-terminal to the phenylalanine is not absolutely required for hydrophobic interactions with NTRs (Figure 2E and F) (Bayliss et al. 2000), but adds to the flexibility of the peptide mainchain to “leave” the pocket.

Similar to importin β, MD-analyses were performed for Cse1p, revealing a total of 14 binding spots for FXFG-peptides (Isgro and Schulten 2007b). Again, the interaction sites on Cse1p were mostly found on the convex outer surface of the protein, many of them in close proximity to each other (Figure 1C). Interestingly, the simulation also suggested five FG-binding sites on Kap60p (the yeast homologue of importin α), supporting the notion that a transport cargo can engage in fruitful interactions with the FG-permeability barrier as well (Frey et al. 2018; Naim et al. 2009).

For CRM1, it was shown that several FG-motifs in Nup214 are required for binding (Roloff et al. 2013). More recent analyses of NTR-Nup interactions by X-ray crystallography used longer stretches of FG-repeat regions (Koyama et al. 2014, 2017; Monecke et al. 2015; Port et al. 2015), leading to the identification of binding regions in CRM1/Xpo1p with a total of 15 F-binding sites (Figure 1D). Analysis of a protein complex comprising human CRM1, RanGTP, the export cargo Snurportin 1 (SPN1) and a 231-amino acid sequence derived from the FG-rich C-terminal end of Nup214, for example, showed that FG-motifs served as prominent anchor points in three regions on CRM1. The spacer regions between the FG-motifs, on the other hand, were more loosely attached to the export receptor (Port et al. 2015). A similar arrangement of phenylalanine binding sites was observed for two of the three binding sites of a complex containing yeast Xpo1p, RanGTP (Gsp1p-GTP), an NES-cargo and a 35-amino acid long stretch of Nup42p containing four conventional FG-motifs (Figure 1D). Here, the phenylalanine side chains contacted hydrophobic binding sites on yeast Xpo1p that are conserved in human CRM1 (Koyama et al. 2017). Strikingly, besides the conventional FG-binding sites, phenylalanines with different neighboring residues were found in Yrb2 contacting Xpo1p and in Nup214, contacting CRM1. One example is the sequence TFS (Figure 3A), where the phenylalanine contacts residues within the hydrophobic pocket, whereas the side chains of the flanking residues are facing away from it. A careful analysis of all the available crystal structures of NTRs in complexes with nucleoporin fragments revealed that such non-conventional interactions are not uncommon. The swapped arrangement of the flanking residues of the TFS-fragment found on CRM1 has been structurally defined in Kap95p, with the amino acid sequence SFT (Figure 3B). As seen for the TFS-peptide bound to CRM1, the phenylalanine interacts with a coordinated aromatic side chain within the pocket. The flanking side chains are facing away from the pocket and some of the main chain atoms interact with residues in the pocket. Interestingly, a nearby proline residue (n + 2) interacts with the phenylalanine residue (n) and also with an aromatic side chain (Tyr) in Kap95p (Figure 3B). This arrangement resembles that of the conventional FxFG repeat (n, x, n + 2, G), suggesting that proline can substitute for a phenylalanine as seen in FxFG repeats. Both peptides share a common arrangement of the n-and n + 2-residues, respectively. SFTP resembles the inverse sequence of the conventional FxFG repeats, with the n + 2-residue in xFxP and the n-residue in FxFG stabilizing the arrangement of the phenylalanine located deeper in the binding pocket. This arrangement adds to the interaction with the NTR, as suggested previously for PxFG (Bayliss et al. 2002b).

Figure 3:

Non-conventional binding sites for phenylalanines in NTRs. (A) Human CRM1 interacting with a 117 aa fragment of Nup214 (res 1916–2033) (PDBid: 5dis; (Port et al. 2015)). (B) Yeast importin β interacting with a 28 aa fragment of Nsp1 (PDBid: 5owu; (Liu and Stewart 2005)). (C) Xpo1p interacting with a 15 aa region of Yrb2p lacking the N-terminal 89 residues (PDBid: 3wyf; (Koyama et al. 2014)). (D) Yeast CRM1 interacting with a 14 amino acid long resolve FG-fragment of Yrb2p lacking the N-terminal 89 residues (PDBid: 3wyf; (Koyama et al. 2014)). Here, the KFV residues are followed by an FG-motif (KFVFG), but the first phenylalanine side chain interacts with its own set of residues on CRM1. (A–D) Coloring as in Figure 2. The colors of the rectangles around the individual panels refer to the mayor binding areas on the individual proteins as depicted in Figure 1D.

In the complex structure of Xpo1p-Gsp1GTP-Yrb2, two additional non-conventional sequences are present. A GFN peptide, i.e. a sequence inversion of the conventional FG-repeat (Figure 3C), and a KFVFG motif (Figure 3D). The latter formally represents a conventional xFxFG motif, however the X-ray structure suggests that KFV alone could potentially bind to NTRs as well.

4 Potential non-FG-interaction spots on nucleoporins

Retrospectively, the available structures are probably biased, because “high-affinity”-FXFG-peptides lacking other potential interaction spots were widely used for the initial crystallization trials (Table 1) and also for MD-simulations. In the context of larger nucleoporin fragments, however, other hydrophobic regions containing phenylalanine residues may engage in short-term interactions with NTRs, as seen in Figure 3. Such interactions may well contribute to overall binding and to the efficiency of nucleocytoplasmic transport. We therefore analyzed the sequences of all known human nucleoporins for potential hydrophobic interaction hotspots, based on the available structural data (Table 1). The following approach was used for the analysis: 33 human proteins were scanned for phenylalanines. Secondary structure elements were positioned based on the available structures and/or available models obtained from AlphaFold calculations (Jumper et al. 2021; Varadi et al. 2022). Phenylalanines within secondary structure elements were considered inaccessible and omitted in the analysis. Table 2 (see also Supplementary Figure S1) lists the conventional FG-motifs of all nucleoporins and also a multitude of novel “non-conventional” F-containing sites (F-sites) located in intrinsically disordered regions or longer loop regions. Such sites could be accessible for protein-protein interactions, substantially increasing the number of possible binding sites on nucleoporins for NTRs. Strikingly, the number of novel F-sites is particular high in FG-Nups. Nup358, Nup214 and Nup153, for example, present 53, 29 and 32 alternative F-sites, respectively (marked in magenta and yellow in Figure S1). For non-FG-Nups, on the other hand, we identified much lower occurrences, e.g. six sites for Nup210 and three sites for Nup205. Two major exceptions here are ELYS and Tpr, with 23 and ten “novel” accessible F-sites, respectively (see below). Together, we identified 193 novel F-sites in all 10 FG-Nups (corresponding to 1.7 sites per 100 amino acid residues) and 68 sites in 21 non-FG-Nups (corresponding to 0.3 sites per 100 amino acid residues; ELYS was omitted from this calculation). Some of the phenylalanines found in accessible regions of non-FG-Nups may be hidden in Nup-Nup contacts, for example in Nup205, Nup155 and Nup43.

Table 1:

X-ray crystallography structures of NTR-nucleoporin complexes.

PDBid	NTR:Nup	Motif	Reference
1f59	Kap95:Nsp1	FSFG	Bayliss et al. (2000)
1jn5	Hs NTF2-like domain of TAP-p15:Nup214	GFG	Fribourg et al. (2001)
1o6o	Sc Importin β:Nsp1	FSFG	Bayliss et al. (2002b)
1o6p		LFG
1gyb	Sc NTF2: Nsp1	FSFG	Bayliss et al. (2002a)
1oai	Hs TAP-UBA domain: Nsp1	FSFG	Grant et al. (2003)
5owu	Kap95:Nup1	SFT, NFS, IFG	Liu and Stewart (2005)
2khh	Sc Mex67-UBA domain	FSFG	Hobeika et al. (2009)
3wyf	Sc Xpo1p-Yrb2p-Gsp1p-GTP	KFVFG	Koyama et al. (2014)
		KFG, AFG, SFG, GFN
5dis	Hs CRM1, Spn1, RanGTP:Nup214	LFG, TFG, SFG,	Port et al. (2015)
		TFS, GFG, FGFG, VFG
5xoj	Sc Xpo1p, PKI-NES, Nup42p, Gsp1p-GTP	FG, PAF, PAFG,	Koyama et al. (2017)
		SAFG

1. Summary of structures revealing interaction pattern of “FG”-motifs bound to nuclear transport factors. Residues not structurally defined are boxed in gray.

Together, the number of phenylalanines located predominantly within IDRs and longer loop regions connecting secondary structure elements is more than doubled for the conventional FG-Nups. Many of the novel motifs are inversions of the FG-motif, i.e. GF-motifs. Additionally, xFS-and SFx-motifs are common in FG-Nups. Facilitated diffusion of all three classes of NTRs, alone or in complex with transport cargoes, could be affected by these novel F-sites.

5 Conclusions

A transport complex that passages through the transport channel of the NPC has to overcome the permeability barrier that is established by a subset of the FG-nucleoporins. In all nuclear transport models, the interaction of NTRs with nucleoporins on their way through the NPC is critical. In situ, these interactions can be extremely complicated, because (i) several nucleoporins compete for NTR-binding, (ii) different NTRs compete for individual nucleoporins, (iii) nucleoporins contain multiple FG-or-FG-like motifs which are the primary binding sites for NTRs and (iv) NTRs contain multiple binding sites for nucleoporins. Hence, multivalent low-affinity interactions will dominate the process of translocation. Large molecules without the ability for such interactions face a free energy barrier that prevents successful translocation. NTRs, with or without cargos, on the other hand, profit from a small energy gain upon binding to individual interaction sites on the surface of a nucleoporin, allowing a stepwise movement within the FG-meshwork. The conventional FG-motifs seem to be the primary (and perhaps preferred) binding sites for NTRs. Nevertheless, alternative, non-FG motifs may contribute to the overall reduction of the free energy barrier and, thus, to the efficiency of nuclear transport. Many such non-conventional sites were identified in our analysis, foremost in FG-Nups that are prone to interact with NTRs anyway. Of particular interest here is Nup358 with >50 novel F-sites, in addition to the 16 previously known FG-motifs. There are probably 40 copies of Nup358 on the cytoplasmic side of each NPC (Bley et al. 2022). Together, ∼2500 FG-or other F-motifs could serve as potential binding sites for NTRs on the Nup358-filaments that emanate into the cytoplasm. Functionally, this cloud of binding sites could help to concentrate NTRs in the vicinity of NPCs, prior to translocation into the nucleus. Indeed, an accumulation of fluorescent importin β was observed on the cytoplasmic side of the NPC (Lowe et al. 2015). During translocation, other nucleoporins like Nup98 and components of the Nup62 complex (Nup62, Nup58 and Nup54), which are thought to contribute to the permeability barrier of the NPC, become important. Together, these nucleoporins present 61 FG-motifs and 38 novel F-sites that might be accessible to NTRs. Nup98 in particular contains many cohesive FG-repeats (Hülsmann et al. 2012) and additional binding sites for NTRs in close vicinity may help to facilitate the transition of the transport complex through the permeability barrier. On the nuclear side of the NPC, ELYS and Tpr stand out in our analysis, as they are considered non-FG-Nups, yet they contain a rather high number of novel F-sites, 24 and 9, respectively. Interestingly, Tpr is involved in several nuclear transport pathways (Aksenova et al. 2019; Bangs et al. 1998; Frosst et al. 2002; Snow et al. 2013) and also serves as a binding site for importin β (Shah et al. 1998). It remains to be investigated if ELYS also affects transport, perhaps in a cargo-dependent manner. Interestingly, importin β has also shown to interact with ELYS, probably in the context of assembly of novel nuclear pores (Rotem et al. 2009).

Obviously, there is no a priori directionality of transport imposed by a random assembly of FG-and non-FG-motifs within the transport channel. Ultimately, directionality of “active” transport (i.e. accumulation of a cargo against a concentration gradient) requires the Ran-system. While GTP-hydrolysis on Ran provides the energy needed for active transport, other factors may affect the directionality as well. The affinity-gradient model postulated that distal nucleoporins have a higher affinity for import receptors than the more central nucleoporins. In protein import, these would be nucleoporins of the nuclear basket. This could lead to preferred binding of the import complex to these terminal sites, where complex disassembly by RanGTP can then terminate the transport event. Indeed, such a gradient of affinities could be detected biochemically, with high affinities of importin β and its yeast homologue Kap95p to Nup153 and Nup1p, respectively (Ben-Efraim and Gerace 2001; Pyhtila and Rexach 2003). Remarkably, nuclear RanGTP not only dissociates import cargoes from NTRs, but also import receptors from nucleoporins (Rexach and Blobel 1995). This does not result from a simple competition for binding sites. RanGTP-binding to importin β rather seems to induce a conformational change, occluding some of the known FG-binding sites of the import receptor (Bayliss et al. 2000). How RanGTP affects additional binding sites in the complex milieu of the transport channel of a functional NPC remains to be investigated. On the other hand, a CRM1-and RanGTP-containing export complex has a high affinity for the cytoplasmic nucleoporin Nup214 (Kehlenbach et al. 1999). Interactions and affinities at individual spots are probably governed by the conventional FG-motifs. Ultimately, the multivalent interactions (i.e. the avidity) between nucleoporins and NTRs, however, could be modulated by novel F-motifs as well. Additional options for low affinity interactions could favor a smooth sliding of the transport complex through the NPC instead of a stepwise hopping-mode of translocation. A detailed understanding of these interactions and their dynamics will be required to fully understand the mechanism of nucleocytoplasmic transport.