Nucleoporin Nup50 Stabilizes Closed Conformation of Armadillo repeat 10 in Importin α5*

Background: The isoform importin α5 recognizes a subset of cargos such as phosphorylated STAT1 and the influenza virus polymerase subunit PB2. Results: Nucleoporin Nup50 stabilizes the closed conformation of armadillo repeat 10 in importin α5. Conclusion: We suggest that Nup50 primary function is to prevent cargo rebinding in preparation for recycling. Significance: Nup50 modulates the import reaction by directly altering the three-dimensional structure of import α5. The human genome encodes six isoforms of importin α that show greater than 60% sequence similarity and remarkable substrate specificity. The isoform importin α5 can bind phosphorylated cargos such as STAT1 and Epstein-Barr Virus Nuclear Antigen 1, as well as the influenza virus polymerase subunit PB2. In this work, we have studied the interaction of the nucleoporin Nup50 with importin α5. We show that the first 47 residues of Nup50 bind to the C terminus of importin α5 like a “clip,” stabilizing the closed conformation of ARM 10. In vitro, Nup50 binds with high affinity either to empty importin α5 or to a preassembled complex of importin α5 bound to the C-terminal domain of the import cargo PB2, resulting in a trimeric complex. By contrast, PB2 can only bind with high affinity to importin α5 in the absence of Nup50. This suggests that Nup50 primary function may not be to actively displace the import cargo from importin α5 but rather to prevent cargo rebinding in preparation for recycling. This is the first evidence for a nucleoporin modulating the import reaction by directly altering the three-dimensional structure of an import adaptor.

The human genome encodes six isoforms of importin ␣ that show greater than 60% sequence similarity and remarkable substrate specificity. The isoform importin ␣5 can bind phosphorylated cargos such as STAT1 and Epstein-Barr Virus Nuclear Antigen 1, as well as the influenza virus polymerase subunit PB2. In this work, we have studied the interaction of the nucleoporin Nup50 with importin ␣5. We show that the first 47 residues of Nup50 bind to the C terminus of importin ␣5 like a "clip," stabilizing the closed conformation of ARM 10. In vitro, Nup50 binds with high affinity either to empty importin ␣5 or to a preassembled complex of importin ␣5 bound to the C-terminal domain of the import cargo PB2, resulting in a trimeric complex. By contrast, PB2 can only bind with high affinity to importin ␣5 in the absence of Nup50. This suggests that Nup50 primary function may not be to actively displace the import cargo from importin ␣5 but rather to prevent cargo rebinding in preparation for recycling. This is the first evidence for a nucleoporin modulating the import reaction by directly altering the threedimensional structure of an import adaptor.
Trafficking of most cellular proteins between the nucleus and the cytoplasm is mediated by soluble transport factors known as karyopherins that shuttle constantly through the nuclear pore complex (NPC) 2 (1). The human genome encodes at least 20 karyopherins related to importin ␤ (also named ␤-karyopherins), which are classified as importins and exportins depending on their involvement in nuclear import or export (2), respectively. Cargos flagged for nuclear import usually expose a highly basic patch of residues known as the nuclear localization signal (NLS), which is exemplified by the SV40 large T-antigen NLS ( 126 PKKKRKV 132 ) (3). In the classical import pathway, the NLS is recognized and bound by the adaptor importin ␣, which heterodimerizes with the receptor importin ␤ via an N-terminal importin ␤ binding (IBB) domain (4). The import complex translocates through the NPC rapidly in a process that involves importin ␤ binding to phenylalanine/ glycine-containing nucleoporins lining the NPC (5). Binding of the small GTPase RanGTP to importin ␤ dissociates it from phenylalanine/glycine-containing nucleoporins and breaks up the importin ␣/␤ heterodimer. Importin ␤ complexed to RanGTP is then recycled back to the cytoplasm, where GTP is hydrolyzed to GDP. Consequently, Ran plays a pivotal role both in energizing and conferring directionality to nuclear transport (1).
␤-Karyopherins are extraordinarily flexible molecules composed of 19 -21 tandem HEAT repeats (6 -8). Each HEAT repeat is comprised of two ␣-helices connected by a turn. Similarly, the adaptor importin ␣ presents a repetitive structure of 10 consecutive armadillo (ARM) motifs (9). Each ARM consists of three ␣-helices (named H1-H2-H3) that stack together to form an extended NLS-binding surface (the "ARM core"), flexibly connected to the N-terminal IBB domain (4). The concave surface of the core of importin ␣ contains two NLS-binding sites: a major site between ARM repeats 2-4 and a minor site between repeats 7 and 8 (see Fig. 1A). In all crystal structures of importin ␣1 bound to classical monopartite NLSs determined to date, the NLS is bound at the major NLS-binding site, and a partially occupied peptide is also seen at the minor binding site (10,11). In contrast, the bipartite NLS of nucleoplasmin spans the entire NLS-binding groove, occupying both major and minor NLS-binding sites (11,12). Other nonclassical NLSs, such as those found in the mitotic regulator TPX2 (13) or phospholipid scramblase 4 (14), bind largely or exclusively to the minor NLS-binding site on importin ␣1. The IBB domain of importin ␣ can also bind the NLS-binding groove to function as an internal autoinhibitory signal (4). This ensures that only fully assembled import complexes (consisting of NLS-importin ␣/␤) are imported and aids in their disassembly once in the nucleus (4,9).
The human genome encodes six isoforms of importin ␣ that fall into three phylogenetically distinct groups: ␣1s, ␣2s, and ␣3s (9). All importin ␣ isoforms contain 10 stacked ARMs and share greater than 60% sequence similarity (9). At the functional level, different importin ␣ isoforms exhibit different substrate specificity while maintaining the ability to bind and import classical NLS substrates (15). A well characterized isoform is importin ␣5, which is involved in the nuclear import of the phosphorylated transcription factor STAT1 (16), the Epstein-Barr virus nuclear antigen 1 (17), and influenza virus RNA polymerase subunit PB2 (18,19). A crystal structure of importin ␣5 in complex with the C-terminal core of influenza PB2 (residues 678 -759; herein referred to as PB2) (18) revealed that PB2 occupies the entire binding groove of importin ␣5, including the major and minor NLS recognition sites. In this structure, the C-terminal helix of importin ␣5 ARM 10 is unpacked and mediates domain swap dimer formation in the crystal (18). This conformation of ARM 10 reflects the unusual flexibility of this isoform, which represents an important determinant both for cargo specificity and high affinity binding (20).
Nup50 (also known as Npap60 or in yeast Nup2p) is a mobile (21) nucleoporin localized at steady state primarily in the nucleoplasmic fibrils of the NPC (22,23). It functions as a cofactor for the importin ␣1-␤ nuclear import complex and was hypothesized to play a role in cargo disassembly (24 -26). The N terminus of Nup50 (residues 1-109) binds the C terminus of importin ␣1 at two distinct sites, corresponding to the minor NLSbinding site and ARM 10 (27). Despite the high level of specify for importin ␣1, disagreement exists in the literature as to the functional role of this interaction. Lindsay et al. (28) proposed that Nup50 functions as a tri-stable switch that stimulates nuclear protein import through the classical import pathway. This function was shown to be dependent on Nup50 simultaneous binding to importin ␣1 and an NLS cargo during nuclear import (28). In contrast, Matsuura and Stewart (27) found that binding of Nup50 and an NLS (both monopartite and bipartite) is mutually exclusive, and thus the main role of Nup50 may be related to displacing import complexes and recycling importin ␣ to the cytoplasm. This idea is corroborated by the evidence that some of the sites bound by Nup50 in importin ␣1 are also critical for the assembly of CAS (27,29), the export receptor that, in concert with RanGTP, exports importin ␣ back to the cytoplasm. However, the putative role of Nup50 in dissociating the import complex was recently challenged by single-molecule observations in permeabilized cells that revealed Nup50 is insufficient to dissociate an importin ␣-NLS cargo complex on the time scale of nuclear transport (30). Finally, Nup50 was shown to be expressed in two splicing isoforms that play opposite roles in regulating the import reaction (31). Although the longer isoform (residues 1-469) promotes release of NLS cargo from importin ␣, the shorter isoform (residues 29 -469), found only in humans, accelerates nuclear import of classical NLS cargos. In the present study, we have determined the structure of the isoform importin ␣5 in complex with the Nup50 and characterized how this nucleoporin modulates importin ␣5 binding to the representative import cargo PB2.

EXPERIMENTAL PROCEDURES
Molecular Cloning and Recombinant Proteins-Cloning of ⌬IBB-importin ␣5 (residues 66 -538) and PB2 (residues 678 -759) were previously described (18). A FLAG-tagged version of ⌬IBB-importin ␣5 was constructed by introducing the FLAG epitope (DYKDDDDK) upstream of residue 66 of importin ␣5 using long PCR (construct FLAG-⌬IBB-importin ␣5). His-MBP-PB2 was constructed by inserting the fragment of the PB2 gene encoding residues 678 -759 into an engineered pET28a vector that also contains the MBP gene downstream of the His 6 tag. A synthetic gene encoding residues 1-109 of human Nup50 was inserted between the BamHI and NotI sites of a pGEX-6p-1 vector (GE Healthcare). All of the plasmids were verified by DNA sequencing. Expression and purification of ⌬IBB-importin ␣5 (20), PB2, MBP-PB2 (32), and a complex of ⌬IBB-importin ␣5 bound to PB2 (18) were performed as previously described. GST-Nup50 was expressed overnight at 20°C in the Escherichia coli strain BL21; lysed in 100 mM sodium phosphate, pH 7.4, 0.6 M NaCl, and 5 mM ␤-mercaptoethanol; and purified on glutathione-agarose beads (Pierce). Either the GST-Nup50 was eluted with 15 mM reduced glutathione, or PreScission protease was added to cleave Nup50 from the GST. Both were further purified by gel filtration chromatography on a Superdex 75 column (GE Healthcare) equilibrated in GF buffer, consisting of 1ϫ PBS (3.2 mM Na 2 HPO4, 0.5 mM KH 2 PO4, 135 mM NaCl, pH 7.4) and 5 mM ␤-mercaptoethanol. To form homogeneous complexes of ⌬IBB-importin ␣5 bound to Nup50, two E. coli lysates expressing GST-Nup50 and His-⌬IBB-importin ␣5 were mixed in ratio 1:2 and immobilized on glutathione-agarose beads (Pierce) as described (27). After washing the beads in GF buffer plus 0.5% Tween 20, PreScission protease was added to the beads to cleave the importin ␣5-Nup50 complex from GST. The complex was further purified by gel filtration chromatography on a Superdex 75 column (GE Healthcare) pre-equilibrated in GF buffer.
Crystallization and Structure Determination-The ⌬IBBimportin ␣5-Nup50 (residues 1-109) complex was concentrated by ultrafiltration to 7.5 mg/ml and crystallized in the presence of 24% PEG 3350, 100 mM BisTris, pH 6.5, using the hanging drop vapor diffusion technique. The drop was equilibrated at 20°C, and crystals appeared within 7 days. 27.5% ethylene glycol was added as cryo-protectant before flash-freezing in a nitrogen stream at Ϫ170°C. X-ray data were collected at NSLS Beamline X25 and X6A on ADSC Quanta CCD detectors. The best crystal resulted in a complete data set to 2.7 Å resolution (supplemental Table SI). All data were indexed, integrated, and scaled using HKL-2000 (33). Crystals of ⌬IBB-importin ␣5-Nup50 complex belong to space group P2 1 2 1 2 1 with two complexes in the asymmetric unit. The structure was determined by molecular replacement using PHASER (34) and the structure ⌬IBB-importin ␣5 as a search model (Protein Data Bank code 2JDQ). The initial solution was refined by rigid body refinement, simulated annealing, and isotropic B-factor in PHENIX (35). Nup50 was built in F o Ϫ F c electron density difference maps using the program Coot (36) followed by positional refinement with PHENIX (37). Further cycles of positional and isotropic B-factor refinement with distinct TLS domains reduced the R work /R free to ϳ21.5/25.5% (supplemental Table SI). The final model includes residues 82-508 for ⌬IBBimportin ␣5 in chain A and B, residues 1-47 of Nup50 in chain C, residues 1-46 in chain D, and 130 water molecules. Four residues from the expression vector (Pro-Leu-Gly-Ser) upstream of Nup50 residue 1 were also included in the final model. The average refined B-factors for ⌬IBB-importin ␣5, Nup50, and solvent are ϳ67.7, 102.8, and 41.6 Å 2 , respectively. The final model has r.m.s.d. for bond and angles of 0.005 Å and 0.795°, respectively, with 92.1% of residues in the most favored regions of the Ramachandran plot and no outliers in disallowed regions (supplemental Table SI). The structure was analyzed using PDBsum (38) and PDBePISA (39); structural figures were made using PyMOL (Delano Scientific). Coordinates for the ⌬IBB-importin ␣5-Nup50 complex have been deposited in the Protein Data Bank with accession code 3TJ3.
Biochemical Techniques-EMSA on agarose was performed as described (20,41). In the assay, 20 g of ⌬IBB-importin ␣5, 24 g of a gel filtration purified complex of ⌬IBB-importin ␣5 bound to either PB2 or Nup50 were separated on a 1.5% agarose gel at room temperature for 1 h. For the Nup50 titration, 24 g of a gel filtration-purified ⌬IBB-importin ␣5-PB2 complex was incubated with a 0.25-10-fold molar excess of Nup50, for 30 min on ice prior to electrophoretic separation on agarose gel. The pulldown assay was performed on glutathione-agarose beads (Pierce) coupled to GST-Nup50 (residues1-109) as previously described (41). 100 g of ⌬IBB-importin ␣5 (37 M in 50 l) was incubated with 25 l of beads for 20 min at room temperature. For PB2 displacement, a 4-fold molar excess of PB2 (151 M in 50 l) was added to the beads after the addition of ⌬IBB-importin ␣5. Alternately, a molar equivalent (120 g) of importin ␣5 precomplexed to PB2 was added to beads and incubated 20 min at room temperature. The beads were washed three times at 20°C in PBS. After washing, all of the samples were dissolved in SDS loading buffer and analyzed on a 12.5% SDS-PAGE.
Affinity Measurements by Flow Fluorimetry-Solution phase equilibrium dissociation constants (K D ) were determined using a KinExA 3000 flow fluorimeter (Sapidyne Instruments). The instrument was configured to sample the free (unbound) concentration of ⌬IBB-importin ␣5 in equilibrated solutions of ⌬IBB-importin ␣5, Nup50, and/or PB2. Sampling occurred in fluorescence flow cells containing a small volume of azlactone beads (Pierce) covalently coupled to either GST-Nup50 or MBP-PB2. The beads were prepared by incubating 200 g of GST-Nup50 or MBP-PB2 in 500 mM sodium bicarbonate, pH 9.5, at 4°C overnight; unreacted groups were subsequently blocked using BSA (1 mg/ml) in 1 M Tris, pH 8.0, at 20°C for 30 min. Equilibrated reaction mixtures were passed through the flow cell, and the amount of FLAG-tagged ⌬IBB-importin ␣5 captured on the beads was detected using a FITC-conjugated anti-FLAG (DYKDDDDK) mouse monoclonal antibody (Genscript, catalogue number A01632). The change in fluorescence signal at 518 nm was linearly proportional to the unbound ⌬IBB-importin ␣5 concentration in solution. All of the reaction mixtures were prepared in a buffer of 10 mM Tris, pH 8.0, 150 mM NaCl, 2 mM DTT, 0.1 mg/ml BSA, and 1 mM PMSF and allowed to come to equilibrium (empirically determined) at 20°C. To determine the K D for the interaction between ⌬IBBimportin ␣5 and Nup50, Nup50 was titrated into solutions of constant ⌬IBB-importin ␣5 concentration (10, 1, or 0.1 nM), and the reaction mixtures were passed over GST-Nup50-conjugated beads. To determine the K D for the interaction between ⌬IBB-importin ␣5 and PB2, PB2 was titrated into solutions of constant ⌬IBB-importin ␣5 concentration (10, 1, or 0.2 nM), and the reaction mixtures were passed over MBP-PB2-conjugated beads. To determine the effect of PB2 on the affinity of ⌬IBB-importin ␣5 for Nup50, Nup50 was titrated into premixed solutions of 10 nM PB2 and a constant ⌬IBB-importin ␣5 concentration (either 10 or 2 nM), and the reaction mixtures were passed over GST-Nup50-conjugated beads. The complex of ⌬IBB-importin ␣5 and Nup50 failed to bind appreciably to MBP-PB2-conjugated beads even at 80 nM; consequently, the effect of Nup50 on the affinity of ⌬IBB-importin ␣5 for PB2 could not be directly measured. The samples were tested in duplicate, and all of the titrations were performed at least twice. The fluorescence signals were globally fit to a general bimolecular interaction model (manufacturer's software).

RESULTS
Structure of Importin ␣5 Bound to Nup50-We used x-ray crystallography to determine the molecular basis of the interaction between human importin ␣5 and Nup50. Crystals of ⌬IBB-importin ␣5 bound to Nup50 (residues 1-109) diffracted to 2.7 Å resolution and have two complexes in the asymmetric unit. Inspection of F o Ϫ F c electron density difference maps revealed an extended tubular density for Nup50 spanning the C terminus of importin ␣5 (Fig. 1). Residues 1-47 of Nup50 were modeled in this density, and the C terminus of importin ␣5 (ARM 10) was entirely rebuilt as compared with the search model (importin ␣5 solved in complex with PB2 (18)). The final model of ⌬IBB-importin ␣5 bound to Nup50 has been refined to R work /R free of 21.5 and 25.5% (supplemental Table SI). In the crystal structure, importin ␣5 reveals a characteristic banana shape formed by 10 ARM repeats stacked against one another (10, 42) (Fig. 1, A and B). Importin ␣5 structure superimposes well to importin ␣2 (the mouse homologue of importin ␣1) previously crystallized with an equivalent fragment of the mouse Nup50 (27) and to yeast Kap60p (10) (r.m.s.d. of 1.85 and 1.22 Å, respectively) but shows dramatic deviations in the C-terminal ARM 10 as compared with the crystallographically observed cargo-bound conformation of importin ␣5 (18) (r.m.s.d., 2.7 Å). Furthermore, the backbone of Nup50 adopts a fully extended conformation, which binds along the surface of importin ␣5 like a "clip," burying ϳ4,320 Å 2 of solvent-accessible surface area (Fig. 1B). Nup50 lacks any tertiary structure, and a two-turn ␣-helix between residues 31-37 is the only secondary structure element present in this fragment of nucleoporin. Consistently, the B-factor for Nup50 is higher than in importin ␣5 (102 versus 67 Å 2 ) and varies throughout its structure with its highest value in the acidic patch between residues 18 and 25, which is poorly resolved.
Structural Determinants for Binding to Nup50-Three molecular determinants contribute to the binding of Nup50 to ARMs 4 -10 of importin ␣5 ( Fig. 2A) and account for a total of 22 hydrogen bonds and 273 nonbonded contacts. First, residues 1-15 bind along the minor NLS-binding site, in a direction antiparallel to that of importin ␣5. The first portion of this moiety, between residues 1 and 6 ( 1 MAKRNA 6 ) is structurally superimposable to the smaller NLS box of the nucleoplasmin bipartite NLS (11) and to the hPLSCR4-NLS (Fig. 2B), which binds exclusively to the minor NLS site (14). This region of Nup50 also largely overlaps with residues 31-37 of Kap60p IBB domain visualized in the context of an export complex, with Cse1p and RanGTP (29). Furthermore, residues 11-15 of Nup50 ( 11 TDRNW 15 ) form a ␤-turn that packs against the side of importin ␣5 and inverts the direction of the Nup50 main chain, from being antiparallel to parallel with respect to importin ␣5. Second, residues 16 -30 of Nup50 extend toward the importin ␣5 C terminus lining on the outer surface of ARM 7-10 ( Fig. 1B). This region contains a cluster of five acidic residues ( 18 EDEAEE 23 ), of which only two, Glu 20 and Glu 22 , directly contact importin ␣5. The other residues are mainly solvent-exposed and have poor electron density. In addition, Met 24 and Phe 27 of Nup50 make hydrophobic contacts with importin ␣5 on the outer surface of ARM 9. Third, residues 31-46 of Nup50 account for most of the interactions with ARM 10 of importin ␣5. Notably, the first part of this moiety, residues 31 SEEVMK 36 , forms two turns of an ␣-helix tightly packed against the intrarepeat loop connecting ARM 10 helix H2-H3 (Figs. 1B and 2A). Downstream of this short helix, the Nup50 main chain continues in a basic cluster (residues 36 KNRAIK-KAKRRN 47 ) that stabilizes the acidic surface of ARM 10 helix H3. Finally, residues 48 -109 were not visible in the crystal structure, although they were present in the crystal. This region contains 15 glycines and is likely highly flexible. Overall, the binding contacts for Nup50 seen in complex with importin ␣5 are very similar to those reported for Nup50 bound to mouse importin ␣2 (27), and the r.m.s.d. between the Nup50s determined in the two crystal structures is ϳ0.91 Å.  Nup50 Stabilizes Closed Conformation of ARM 10 in Importin ␣5-Perhaps the most striking feature of the importin ␣5 structure described in this paper lies in the conformation of ARM 10 (Fig. 1B). In our structure, this repeat adopts a canonical, "closed" conformation, with the three ARM helices H1, H2, and H3 stacked on top of each other to generate a hydrophobic core. In contrast, the structure of importin ␣5 visualized crystallographically in complex with PB2 shows that ARM 10 unfolds as a result of cargo binding (18). Here helices H2 and H3 project by ϳ40°and 90°, respectively, from the corresponding position seen in our structure (Fig. 3A). Unfolding of ARM 10 tertiary structure likely stabilizes cargo binding and results in several specific contacts between helix H3 and PB2 (supplemental Fig. S1) (18). Although the overall r.m.s.d. of importin ␣5 in complex with PB2 and Nup50 is only ϳ1.3 Å, the confor-mation of ARM 10 is drastically different in the two complexes: residue 507 of importin ␣5 bound to Nup50 is over 40 Å away from its corresponding position in the PB2-bound state (Fig.  3A). Previous biochemical studies have revealed that Tyr 476 is critical for ARM 10 swinging (20). This residue is located at the end of the first helix of ARM 10 (helix H1) where it functions like a "Tyr finger" (Fig. 3B). The bulky aromatic side chain projects at the interface of ARM 10 helices H1-H2-H3 to destabilize the interhelical stacking of the three ARM helices, thereby allowing the unfolding of ARM 10 and the extension of helix H3. In our structure, Nup50 stabilizes a closed conformation of ARM 10 that is identical to that seen in importin ␣1/Kap60p (42,43), where helices H1-H2-H3 stack onto each other to generate a flat surface. To accommodate the closed conformation of ARM 10, Tyr 476 swings 180°in response to Nup50 binding,

Nup50 Binding to Importin ␣5
thus allowing stacking of helix H1-H3 as seen in importin ␣1, where this residue is replaced by a Gly (20) (Fig. 3B). Thus, the extended conformation of Nup50 that wraps around the C terminus of importin ␣5 is likely an important structural determinant to stabilize the closed conformation of ARM 10.
Importin ␣5 Can Form a Trimeric Complex with PB2 and Nup50-Both Nup50 and PB2 (18) make wide surface contacts with importin ␣5, likely essential for nuclear import. The partial overlap between PB2 and Nup50 binding sites in importin ␣5 (Fig. 3A), as well as the distinct conformation of importin ␣5 ARM 10 in the two bound conformations led us to investigate whether binding of these two proteins to importin ␣5 is mutually exclusive. This hypothesis was tested using three independent binding techniques. First, we carried out a pulldown assay on glutathione-agarose beads coupled to GST-Nup50 using ϳ40 M ⌬IBB-importin ␣5 and 150 M PB2. As shown in Fig.  4A (lanes 5 and 6), GST-Nup50 beads were able to bind ⌬IBBimportin ␣5, whereas no interaction was seen for free PB2. Adding either purified ⌬IBB-importin ␣5 and then PB2 or a gel filtration-purified complex of these two proteins (as in Fig. 4A, lane 4) resulted in three species bound to beads, corresponding to ⌬IBB-importin ␣5, GST-Nup50, and PB2 (Fig. 4A, lanes 7  and 8, respectively). Thus, binding of Nup50 and PB2 to importin ␣5 is not mutually exclusive, and ⌬IBB-importin ␣5 can form a trimeric complex with PB2 and Nup50. To further explore this hypothesis, we used a native EMSA on agarose gel; in this experiment, increasing concentrations of purified Nup50 were titrated into a preformed importin ␣5-PB2 complex (Fig. 4B, lanes 3-9). This yielded a trimeric importin ␣5-Nup50-PB2 complex that migrated more slowly than the importin ␣5-Nup50 complex (Fig. 4B, lanes 2) and was significantly slowed compared with the importin ␣5-PB2 complex (Fig. 4B, lane 3). The putative trimeric complex assembled when 1-10 molar equivalents of Nup50 were added to the preformed importin ␣5-PB2 complex (at ϳ20 M concentration) (Fig. 4B, lanes 6 -9). Finally, we characterized the trimeric ⌬IBB-importin ␣5-Nup50-PB2 complex in solution, using analytical size exclusion chromatography on a Superose 12 column. At physiological ionic strength, unliganded ⌬IBB-importin ␣5 migrated as an ϳ85-kDa species, slightly larger than the true molecular mass (ϳ55 kDa), but consistent with the elongated monomer seen in the crystal structure (supplemental Fig.  S2) (a similar, larger than expected migration was also observed for importin ␣1 (44)). A trimeric complex formed using an excess of Nup50 or PB2 added to a gel filtration-purified complex of ⌬IBB-importin ␣5 bound to Nup50 or PB2 contained all three components, although not in exact equimolar ratio (supplemental Fig. S2). Thus, on beads, in agarose and in solution, at concentrations in the low micromolar range ⌬IBB-importin ␣5 can form a tertiary complex with Nup50 and PB2.
PB2 Reduces the Binding Affinity of Nup50 for Importin ␣5-Next, we sought to determine how the binding affinity of importin ␣5 for Nup50 and PB2 is altered by either factor. To address this question, we used a KinExA 3000 flow fluorimeter. This binding technique uses a solid phase immobilized molecule to probe for free concentration of one interacting partner after allowing sufficient time to reach equilibrium (45). In our experimental set-up (illustrated in supplemental Fig. S3A), equilibrated mixtures of ⌬IBB-importin ␣5 (that also included a short N-terminal epitope FLAG epitope) and Nup50 (residues 1-109) were passed through a fluorescence flow cell containing a small quantity (ϳ4 l) of GST-Nup50-conjugated beads. As reaction mixtures were passed through the optical flow cell of the instrument, a small fraction of the total unbound ⌬IBBimportin ␣5 was captured by beads coated with GST-Nup50. The exact amount of captured ⌬IBB-importin ␣5 was quantified using an anti-FLAG-FITC-conjugated mouse monoclonal antibody, and the resulting increase in bead fluorescence (⌬f) was directly proportional to the concentration of unbound ⌬IBB-importin ␣5 in the reaction mixture (supplemental Fig.  S3B). From full titrations of Nup50 at several ⌬IBB-importin ␣5 B, EMSA on native agarose gel. Free ⌬IBB-importin ␣5 is in lane 1, and preformed complexes of ⌬IBB-importin ␣5 bound to Nup50 or PB2 are in lanes 2 and 3, respectively. The addition of increasing quantities of Nup50 (from 0.25to 10-fold molar excess) to a preformed ⌬IBB-importin ␣5-PB2 complex leads to a trimeric species that is retarded as compared with ⌬IBB-importin ␣5 bound to Nup50 (lanes 2) or PB2 (lane 3). Single and double asterisks indicate the position of the dimeric ⌬IBB-importin ␣5-Nup50 and trimeric ⌬IBB-importin ␣5-Nup50-PB2 complex, respectively. The difference in migration between these two species is also emphasized by dashed horizontal lines. M.W., molecular mass. concentrations in the low nanomolar to high picomolar range, we have determined the K D between ⌬IBB-importin ␣5 and Nup50 to be Ͻ2 pM (Fig. 5A).
Interestingly, when the same experiment was repeated using ⌬IBB-importin ␣5 precomplexed with 10 nM PB2, ⌬IBB-importin ␣5 could still be efficiently captured by the Nup50-conjugated beads in the KinExA flow cell. At this concentration of PB2 (Ͼ100-fold above its K D for ⌬IBB-importin ␣5, see below and Fig. 5B), ⌬IBB-importin ␣5 was more than 99% bound to PB2. By titrating Nup50 into pre-equilibrated solutions of the ⌬IBB-importin ␣5-PB2 complex, we determined the apparent K D for Nup50 to be 0.815 Ϯ 0.029 nM (Fig. 5A). Assuming that importin ␣5 can simultaneously bind Nup50 and PB2 at the concentration used in the KinExA assay (and as also suggested by pulldown, native gel and size exclusion chromatography assays), these data suggest that the presence of bound cargo reduces the binding affinity of importin ␣5 for Nup50 by Ͼ400fold. Although this affinity is lower than that measured in the absence of PB2, the importin ␣5-PB2 complex is still captured by Nup50 with robust, nanomolar binding affinity.
In a parallel experiment, we immobilized MBP-PB2 on beads and performed full titrations of PB2 with ⌬IBB-importin ␣5 at concentrations in the low nanomolar to high picomolar range. The results indicated that ⌬IBB-importin ␣5 binds MBP-PB2 beads with a K D ϭ 53 Ϯ 2 pM (Fig. 5B). Interestingly, when a gel filtration-purified complex of ⌬IBB-importin ␣5 prebound to Nup50 (between 2 and 80 nM) was probed for binding of PB2coated beads, it failed to produce sufficient signal when passed over beads. This lack of capture implies that prebound Nup50 efficiently prevents binding of PB2 to ⌬IBB-importin ␣5. In contrast to the ⌬IBB-importin ␣5-PB2 complex that retains an efficient interaction with Nup50-beads, the ⌬IBB-importin ␣5-Nup50 complex appears to adopt a structure that prevents association with PB2. This argues that at the concentration used in the KinExA assay, the tripartite interaction behaves in a simple competitive manner.

DISCUSSION
Structural Plasticity of ARM 10-␣-Karyopherins are composed of ARM repeats and are thought to be predominately rigid. The ARM core generates a cargo-binding surface ideal to function as a static NLS-binding groove (46). This is very different from the HEAT-repeated architecture of ␤-karyopherins that form superhelical solenoids of remarkable structural flexibility (7,8,44). The results presented in this paper, together with previous structural (18) and biochemical (20) reports on importin ␣5 (Fig. 6), provide compelling evidence that ARM 10 of import ␣5 functions as a reversible mobile arm. Helix H3 in ARM 10 swings upon binding to the import cargo PB2 and folds back on itself in response to Nup50 binding (Fig. 6). Tyr 476 is likely the key residue in importin ␣5 controlling this switch (20). In the presence of cargo, this Tyr finger folds inward inside the hydrophobic core of ARM 10 (Fig. 3B), destabilizing the interhelical stacking of ARM 10 helices and hence promoting unfolding of ARM 10 (18). Binding of importin ␣5 to Nup50 causes Tyr 476 to swing 180°outward to promote a folded conformation of ARM 10, which is identical to that seen in importin ␣1/2 (42), where this residue is replaced by a Gly. What is the function of this structural switch in ARM 10 of importin ␣5? Reanalysis of the importin ␣5-PB2 crystal structure (18) suggests that the extended conformation of ARM 10 helix H3 stabilizes PB2 binding by forming at least six specific electrostatic contacts between the two proteins, namely, PB2-Lys 699 with ␣5-Gln 494 , PB2-Arg 703 with ␣5-Thr 506 /␣5-Glu 507 /␣5-Glu 509 , PB2-Tyr 704 with ␣5-Glu 509 , and PB2-Lys 721 with ␣5-Glu 509 (supplemental Fig. S1). Conversely, the closed conformation of ARM 10 would result in complete loss of these contacts, which likely weakens (but does not fully disrupt) the import complex. Interestingly, Gabriel et al. (47) recently determined that efficient replication of mammalian influenza virus requires the isoform importin ␣7. This isoform is 81% identical to importin ␣5 and also presents a Tyr at position 476 (48), supporting the hypothesis that this Tyr finger plays a role in cargo binding by modulating the interconversion of ARM 10 between a closed to an open conformation (Fig. 6). However, PB2 may not be the most representative cargo for importin ␣5, because it also binds (and can be imported) by other importin ␣ isoforms (32,47). The nonclassical import cargo STAT1, instead, is strictly dependent on the extended conformation of ARM 10. Phos- phorylation at Tyr 701 triggers a conformational change in STAT1 that exposes a nonclassical NLS, which binds specifically to import ␣5 but has negligible affinity for importin ␣1 (20). Mutagenesis studies revealed that replacing Tyr 476 in importin ␣5 with Gly (as in importin ␣1) or swapping the C terminus of importin ␣5 with equivalent residues of importin ␣1 dramatically disrupts binding to STAT1 (20). Thus, Tyr 476 and the structural switch of ARM 10 in importin ␣5 (as well as in other importin ␣5-like isoforms like importin ␣6 and ␣7 (48)) likely serve the function of stabilizing cargo binding and promoting specificity toward certain specialized cargos.
Role of Nup50 in Cargo Displacement and Importin ␣5 Recycling-Several reports have pointed out a role of Nup50 (or its yeast homologue Nup2p) in destabilizing the importin ␣1/NLS cargo interaction (25,27,31,43,49,50). In this study, we have determined that Nup50 binds importin ␣5 and stabilizes the closed conformation of ARM 10. Association of Nup50 with importin ␣5 is compatible with binding to PB2. At concentrations in the range of 5-40 M (as used in all experiments except for KinExA), a trimeric species formed by importin ␣5, Nup50, and PB2 can be stably assembled in solution and is sufficiently long-lived to be visualized using conventional biochemical techniques such pulldown assay, native gel electrophoresis, and size exclusion chromatography ( Fig. 4 and supplemental Fig. S2). However, the physiological concentration for these proteins is likely to be lower in vivo. For instance, the cellular concentration of all importin ␣ isoforms is estimated at 1 M (51), whereas the concentration of Nup50 in the nucleus is ϳ130 nM (based on 32 copies/NPC (21) and ϳ2,770 NPC/cell (52) and a nuclear volume of 1,130 M 3 (52)), and PB2 concentration is likely in the nanomolar/picomolar range (and stoichiometric to that of importin ␣5 inside the NPC). This range of concentration was explored in the KinExA assay, where we determined that the presence of 10 nM PB2 decreases the K D of importin ␣5 for Nup50 by Ͼ400-fold, likely because of formation of a trimeric complex. Although decreased, the binding affinity of the importin ␣5-PB2 complex for Nup50 remains very high (K D ϭ ϳ0.85 nM), which is likely incompatible with free dissociation of cargo. Furthermore, negligible interaction with the PB2 immobilized on beads was observed when Nup50 was prebound to importin ␣5, although this may be due to a kinetic effect during the attempted capture of the importin ␣5-Nup50 complex rather than a lack of trimeric complex formation. These results contrast with the report that, in vitro, Nup50 can actively displace a chimeric cargo consisting of GST-NLS from importin ␣2 (27). Instead, our observations are consistent with previous reports that a trimeric complex of importin ␣1, NLS cargo, and Nup50, or its yeast homologue Nup2p, can be isolated from both mammalian (28) and yeast (40,49) extracts. Furthermore, Nup50 does not bind to the major NLS-binding site of importin ␣5 or ␣2 (27), which would be the logical target for a factor with cargo disassembling activity. Together, these direct and indirect observations argue against the idea that the primary function of Nup50 is to actively displace import cargos from importin ␣5 (Fig. 6, function a). In agreement with this idea, single molecule analysis of importin ␣/cargo complex dissociation at the NPC found that Nup50 alone is insufficient to promote importin ␣1/cargo complex dissociation on the time scale of transport (30). Instead, in vivo cargo displacement is mainly exerted by the combined action of CAS and RanGTP (30). RanGTP dissociates the importin ␣/␤ complex, thereby freeing up the IBB domain that functions as a potent, intramolecular binder of major (42) and minor (29) NLS-binding sites. In turn, CAS in complex with RanGTP loads empty importin ␣ to recycle the adaptor into the cytoplasm. Given the exceedingly high binding affinity of Nup50 for importin ␣5 (K D Ͻ 2 pM), we propose that Nup50 may have a preferential role in binding empty, as opposed to cargo-bound importin ␣5. In this scenario, Nup50 would "capture" importin ␣5 after RanGTP/IBB-mediated displacement of cargo to prevent its rebinding to importin ␣5 (Fig. 6, function  b). Similarly, Nup50 could bind directly to empty importin ␣5 that has escaped the cytoplasm in the absence of cargo and avoid futile import of unliganded adaptors (Fig. 6, function c).
Nup50 was proposed to serve as a scaffold that binds both the importin ␣1-NLS cargo and CAS-RanGTP complex, possibly facilitating cargo dissociation by tethering the two complexes together (27). A similar function can be hypothesized for importin ␣5, with the noticeable difference that in this isoform the conformations competent for binding to cargo (␣5 O ) and CAS (␣5 C ) are likely distinct (Fig. 6). Like Nup50, in fact, CAS binds ARM 10 of Kap60p in a closed conformation (29), suggesting that the primary role of Nup50 may be to function in concert with CAS to fold back ARM 10 in preparation for recycling (Fig. 6). However, if the affinity of Nup50 for importin ␣5 is so high, how can the CAS-importin ␣5-RanGTP export complex assemble? The answer to this question likely lies in the modular nature of Nup50 interaction complexes. Nup50 could "transfer" importin ␣5 in a closed conformation to CAS in a two-step mechanisms. First, the IBB domain (which is stabilized by CAS in an autoinhibited conformation (29)) could compete off the first portion of Nup50 (residues 1-11), weakening the overall affinity for importin ␣5; this is plausible, because the IBB presents more binding determinants in the minor binding site (29) than Nup50. Second, CAS may bind the C terminus ARM 10 of importin ␣5, releasing the Nup50 and allowing for re-entry into the cytoplasm. In conclusion, the results presented in this paper expand our understanding of the isoform importin ␣5 and suggest a role for Nup50 in promoting the disassembly of the importin ␣5 import complex indirectly by capturing empty importin ␣5 and preventing cargo rebinding.