Prolines in the α-helix confer the structural flexibility and functional integrity of importin-β

The nuclear pore complex (NPC) serves as the sole gateway for communication between the eukaryotic nucleus and cytoplasm by exchanging functional molecules. Both inbound and outbound molecules are controlled by the selective barrier formed in the central channel of the NPC. Molecules smaller than 40 kDa can pass through the NPC by passive diffusion, while larger proteins cannot pass through without the aid of transport mediators (for review, see Gorlich and Kutay, 1999). Karyopherin-β family proteins, which include importin-β (KPNB1) and CRM1 (exportin 1 or XPO1), form the largest group of transport mediators (20 members), and mediate the transport of distinct sets of cargo proteins (for reviews, see Mosammaparast and Pemberton, 2004; Pemberton and Paschal, 2005).

Many approaches using biochemistry, structural biology and bioinformatics have revealed the structure of the NPC and the properties of its components, called nucleoporins (Nups) (Cronshaw et al., 2002; Rout et al., 2000). The entire complex is composed of more than 30 different Nups, each of which exists as multiple copies. The central channel of the NPC comprises a number of unstructured (disordered) polypeptides carrying phenylalanine-glycine (FG) motifs (FG-Nups). FG-Nups are similar to other intrinsically disordered proteins in that they are rich in polar residues, such as serine and threonine, but distinct from them by the frequent presence of hydrophobic residues (phenylalanine). This unique amino acid composition enables FG-Nups to form a hydrogel, where hydrophobic interactions between the phenylalanine residues crosslink the disordered polypeptides into a gel-like matrix (Frey and Gorlich, 2006; Mohr et al., 2009). In vivo, a strong hydrophobic crowding barrier has been shown to form in the NPC, which is especially enriched at both peripheries of the pore (Konishi et al., 2017).

Karyopherin-β family proteins are designed to pass through the amphiphilic matrix of the NPC. The crystal structures of karyopherin-β family proteins exhibit significant similarities in their overall molecular shape, although their amino acid sequence similarities are very low (15–20% identity) (O'Reilly et al., 2011; Xu et al., 2010). Karyopherins are composed of 19–21 tandem repeats of HEAT (Huntingtin, elongation factor 3, protein phosphatase 2A and PI3-kinase TOR1) motifs, each containing two amphiphilic α-helices (A-helix and B-helix) connected by a short linker region (Chook and Blobel, 1999; Cingolani et al., 1999). This characteristic helical composition makes the entire karyopherin molecule amphiphilic, with the hydrophobic sides facing each other toward the inside of the molecule and the hydrophilic sides facing outward to the solvent (Fig. 1A). A long array of HEAT motifs (HEAT repeat) confers the entire molecule with structural flexibility, which plays important roles in its interactions with cargo proteins and RanGTP (Chook and Blobel, 2001; Conti et al., 2006; Forwood et al., 2010; Lee et al., 2000; Stewart, 2007).

The structural flexibility of karyopherins is also important for their passage through the pore. Previous structural analyses have revealed flexible structural changes of importin-β upon binding to its cargo (Cingolani et al., 2000) or under different solvent conditions (Yoshimura et al., 2014), and its conformational flexibility has been shown to be important for protein–protein interactions and nuclear translocation (Lee et al., 2000). Karyopherins bind firmly to FG-Nups through hydrophobic pockets between adjacent A-helices (Bayliss et al., 2000, 2002; Liu and Stewart, 2005; Otsuka et al., 2008). Substitution of these residues with polar or charged residues reduces the affinity of karyopherins for FG-Nups. Hydrophobic residues within the NPC induce conformational changes on karyopherins to expose more hydrophobic residues on the molecular surface, allowing the molecule to diffuse through the pore. The affinity of karyopherins to FG-Nups is also known to be modulated by the structural changes induced by binding to their cargos (Lott et al., 2010). These findings explain how karyopherins can overcome the hydrophobic NPC barrier. Other HEAT-motif-rich proteins (such as PPP2R1A and CAND1) can also pass though the NPC by themselves (Yoshimura et al., 2014), indicating that the HEAT motif is generally suitable for NPC passage.

One of the striking features that distinguishes the HEAT motif from other helical repeat motifs (such as TPR, Arm and Ankyrin) is the presence of conserved proline residues in the A-helix (Chook and Blobel, 1999; Yoshimura and Hirano, 2016). Since the Cα–N bond of a proline residue in a peptide is incorporated in the Y-ring, the φ and ψ angles of proline are not suitable for α-helix formation. As a result of this, proline is known as the least likely of the 20 amino acids to be present in an α-helix and is widely regarded as an α-helical ‘breaker’ (O'Neil and DeGrado, 1990; Piela et al., 1987). Indeed, several mutational substitutions of an α-helix amino acid into proline are related to disease, including an alanine-to-proline mutation in a coiled-coil domain of keratin-10 (A158P) found in epidermolytic hyperkeratosis (Yang et al., 1997). Although it is known that a proline residue can form a kink in an A-helix (Cingolani et al., 1999), its functional significance has not been fully addressed. In this report, we focused on the proline residues in HEAT motifs and investigated their structural and functional roles. In particular, we examined how proline residues confer structural flexibility upon importin-β and achieve its efficient translocation through the NPC.

Through a global search of proline residues within karyopherins and other HEAT-rich proteins, many proline residues were found in the A-helix (outer helices) and loop regions that connect the A- and B-helices. Importin-β1 contains 15 proline residues in the A-helices, 21 in the loops, and 1 in the B-helices based on its crystal structure (Fig. 1B). Almost all of the proline residues in the A-helices are located at the molecular surface (convex face) and are in contact with solvent (Fig. 1C, Movie 1).

A series of mouse importin-β1 mutants were constructed in which all 15 prolines in the A-helices were mutated to alanine (ΔPro-all), together with strains containing a subset of the mutations (ΔPro-I, ΔPro-II, ΔPro-III and ΔPro-IV) (Fig. 2A). These mutants were expressed in HeLa cells in an enhanced green fluorescent protein (EGFP)-fused form. Anti-EGFP western blotting was performed against whole cell lysates obtained at 24 h post-transfection to confirm expression of the full-length molecules. A single band at the expected molecular mass (∼120 kDa) was obtained in all mutants, suggesting similar stability among them (Fig. S1A). The subcellular localization of the wild-type and mutant proteins was observed by confocal laser-scanning microscopy (Fig. 2B and Fig. S1B). Cells with similar EGFP expression levels were selected by quantifying cellular fluorescent intensity to analyze the subcellular localization of the mutants (Fig. S1C). Averaged signal intensities at the nuclear envelope, nucleoplasm and cytoplasm were quantified and their ratios were analyzed statistically for all of the mutants. The results clearly showed that ΔPro-IV and ΔPro-all accumulated more in the nuclear envelope and less in the nucleoplasm compared with the wild-type protein (Fig. 2B). ΔPro-III also showed significantly reduced nuclear accumulation than the wild-type protein, suggesting the importance of C-terminal prolines in the nuclear transport of importin-β. When EGFP–ΔPro-all was expressed in HeLa cells, significant deformation of the shape of the nucleus was observed, which accompanied the apparent increase in the cytoplasmic population of the NLS-cargo protein (Fig. 2C,D). This suggests a functional defect of the ΔPro-all mutant that causes an aberration in the importin-β-dependent nuclear transport pathway. This mutant also caused severe defects in mitotic progression. Time-lapse observations revealed that cells expressing ΔPro-all were arrested at the beginning of prophase, showing a defect in breakdown of the nuclear envelope (Fig. 2E, Movies 2 and 3). Cells expressing EGFP–ΔPro-IV exhibited similar mitotic defects to those expressing ΔPro-all (n=11 for ΔPro-IV and n=11 for ΔPro-all), and the aberrant cells eventually died after approximately 3 h arrest at the beginning of mitosis (Fig. S2). These results demonstrated a dominant-negative effect of the ΔPro-IV and ΔPro-all mutants in vivo that may be the result of a defect in their nuclear transport ability.

The passage of the mutants through the NPC was examined in vivo by fluorescence recovery after photobleaching (FRAP) analysis. After bleaching the fluorescence signal in the nucleus, including the nuclear envelope, signal recovery was monitored by time-lapse imaging (Fig. 3A). Fluorescence recovery was analyzed separately for the nuclear envelope and nucleoplasm (Fig. 3B,C). Compared with the fast recovery of the wild-type protein for both the nuclear envelope and nucleoplasm, the accumulation of the ΔPro-all mutant was significantly slower for both compartments. Final fluorescence recovery (I_final) and association rate constant (k_on) for the nuclear envelope were estimated by exponential association fitting (Sprague et al., 2006) (Fig. S3). Although I_final values may not be directly comparable owing to the different amounts of bleached populations between the analyses (Fig. 2B), the results indicate distinctively smaller recovery and attenuated association of the ΔPro-all mutant to the nuclear envelope (Fig. 3D). Nuclear import kinetics were also analyzed by fitting fluorescent recovery at the nucleoplasm as described previously (Kumeta et al., 2012). The import rate (k_in) for ΔPro-all was one order of magnitude smaller than that of the wild-type protein, clearly showing a defect in its translocation through the NPC (Fig. 3E). Among the partial mutants, ΔPro-III showed significantly slower nucleoplasmic recovery and ΔPro-IV exhibited significantly reduced recovery for both the nuclear envelope and nucleoplasm compared with the wild-type protein, suggesting the importance of the C-terminal prolines.

The detailed kinetic properties of the mutants were analyzed further using an in vitro transport assay. EGFP-fused mutant importin-β, as well as the wild-type protein, were expressed in bacteria, purified by affinity chromatography and incubated with digitonin-treated HeLa cells. Time-lapse observations showed that all mutants tended to accumulate in the nuclear envelope, resulting in slower nuclear accumulation (Fig. 4A,B). In good agreement with the FRAP analysis, the influx rate constant of the ΔPro-all mutant was one order of magnitude smaller than that of the wild-type protein (Fig. 4C). The same experiment was performed with cargo-loaded importin-β by using the importin-β binding domain (IBB) of importin-α. All of the IBB-bound mutants exhibited smaller influx rate constants than those measured for the wild-type protein (Fig. 4D).

A gel filtration assay was performed using purified wild-type and ΔPro-all importin-β to investigate their oligomeric status. Although the elution peak was slightly different for wild-type and ΔPro-all importin-β, both showed a single peak, indicating that they both exist as monomers (Fig. 5A). The role of proline in the interaction of importin-β with RanGTP and cargo proteins was then examined using a pull-down assay. All of the single mutants and multiple mutants exhibited similar affinity to RanGTP (Fig. 5B) and to the IBB (Fig. 5C) in the µM range, which is similar to that in the cytoplasm. These results indicate that proline-to-alanine mutations in the A-helix do not affect the interaction of importin-β with cargo or RanGTP. This is reasonable because the import cargo and Ran mainly interact with the concave surface (B-helix) of importin-β, whereas the proline residues are located at the convex surface (A-helix). The interaction of importin-β with FG-Nups [full-length Nup54, FG-rich regions of Nup62 (amino acid residues 1–178) and Nup153 (amino acid residues 896–1475)] was also analyzed using a pull-down assay. These nucleoporins play important roles in transport kinetics and have high affinity to their import cargos (Ben-Efraim and Gerace, 2001). Relative quantification of the amount of bound protein clearly showed enhanced binding of the ΔPro-all and ΔPro-IV mutants with Nup62 and Nup153, but not with Nup54 (Fig. 5D). Considering that the interaction of karyopherins and FG-Nups is hydrophobic in nature, these mutants may assume different molecular conformations that are more suitable for interacting with Nups than the wild-type protein.

To understand further the effect of the mutations on structural stability, molecular analyses were performed using recombinant proteins purified from bacteria. When analyzed by circular dichroism (CD) spectra measurements, all of the ΔPro mutants and wild-type importin-β exhibited a typical α-helix-rich spectrum with two characteristic negative peaks at 208 and 222 nm (Fig. S4). Proline-to-alanine substitutions slightly enhanced this signal, but did not significantly affect the spectra peaks, suggesting that the ΔPro mutants retain similar α-helical content to that seen in the wild-type protein. The tertiary structure of importin-β was evaluated by the fluorescence spectrum of tryptophan residues. Folding free energy (ΔG₀) was estimated from a shift of peak center wavelength with increasing concentrations of urea (0–9 M) (Fig. S5A). When compared with a globular protein [bovine serum albumin (BSA)], importin-β showed smaller ΔG₀ and cooperativity, indicating that helical repeat proteins have a weaker hydrophobic core than globular proteins (Fig. 6A). The addition of amphiphilic molecules, such as trifluoroethanol (TFE), which have been demonstrated to mimic the hydrophobic environment of the NPC (Yoshimura et al., 2014), altered the peak center shift and reduced the folding energy of importin-β (Fig. 6A,B, and Fig. S5B). These results imply that the hydrophobic core of importin-β, which consists of amphiphilic α-helices of HEAT repeats, is weakened in the NPC.

The effect of the proline mutations on the folding free energy of importin-β was then examined. Apparent folding free energy (ΔG₀) was estimated by observing the irreversible unfolding curve of the purified proteins. A comparison of ΔG₀ between the wild-type and ΔPro mutants revealed that the ΔPro-IV and ΔPro-All mutants showed significantly larger ΔG₀ values than the wild-type protein, whereas the other mutants showed comparable or even smaller values (Fig. 6A and Fig. S6). These results indicate that mutations in the C-terminal HEAT motifs reduce molecular flexibility, which resulted in the aberration of nuclear transport ability (Figs 2 and 3).

The role of proline residues in the structural flexibility of importin-β was also analyzed by measuring the hydrophobicity of the protein surface. The hydrophobic fluorescent probe bis-ANS binds to the hydrophobic surface of proteins and emits strong fluorescence. When using a low concentration of the probe (0.1–10 μM), the fluorescent intensity of the ΔPro-all mutant was slightly higher than that of wild-type importin-β (Fig. 6C). This is consistent with the results of the binding assays, indicating that the molecular surface of the mutant is slightly more hydrophobic than that of the wild-type protein in an aqueous solution. When the concentration of bis-ANS was increased, the wild-type protein showed a drastic increase in fluorescence intensity whereas ΔPro-all exhibited a considerably smaller change. A comparison of the fluorescence intensity at 50 μM bis-ANS with that at 5 μM clearly showed differences between the wild-type and mutant proteins (Fig. 6C). It is reasonable to conclude that wild-type importin-β underwent a structural change in response to the hydrophobic environment to expose more of its hydrophobic area to the molecular surface. The ΔPro-all mutant clearly lacks this molecular flexibility and was unable to adapt to the hydrophobic environment.

Molecular dynamics simulation was performed to understand the molecular mechanism of structural flexibility conferred by proline residues. The crystal structure of importin-β [Protein data bank (PDB) ID: 1ukl] was used as the initial structure for the simulation. The proline mutant (ΔPro-all) was constructed by replacing the proline residues in the A-helices by alanine in the PDB data. The structures of wild-type and ΔPro-all mutant importin-β in water did not show significant differences in their overall molecular shape (Fig. 7A,B). In good agreement with the results from the CD spectra and bis-ANS binding assays, detailed structural analyses revealed that the structure of individual α-helices was not significantly affected by the loss of proline residues; helix length, helix–helix distance and the kink angle of the helix were all similar between wild-type and ΔPro-all mutant importin-β (Fig. 7A,B, and Fig. S7).

The simulation was then performed in a 50% water:50% TFE mixture to reveal structural flexibility in a hydrophobic environment. We have demonstrated previously that a water/TFE mixture mimics the environment within the NPC because of its amphiphilic property (Yoshimura et al., 2014). When wild-type and ΔPro-all mutant importin-β were transferred to the water/TFE mixture, they started to change conformation and ended up with a partly altered structure (Fig. 7C,D,G). As demonstrated in our previous study, wild-type importin-β started to unfold its helical repeat partially, especially at the C-terminal HEAT repeats (HEAT 15–18). Surprisingly, structural fluctuations of these helices were increased in the presence of TFE, even though TFE is known to stabilize α-helical structures. This suggests that the α-helices in the HEAT repeats are stabilized mainly by hydrophobic interactions between helices, and not by hydrogen bonds within the helices. The effect of TFE was smaller on the ΔPro-all mutant than on the wild-type protein. The change of root mean square deviation (RMSD), as well as the fluctuation of individual A-helices, was significantly smaller in the ΔPro-all mutant than in wild-type protein only in the water/TFE mixture (Fig. 7E–G). It is interesting to note that such differences were found not only in proline-containing helices (HEAT 15A and 16A) but also in helices lacking proline (HEAT 17A). This implies that not only intra-helix interactions but also inter-helix interactions play an important role in the stability of HEAT repeats.

The reversibility of the TFE-induced structural change was analyzed by transferring the molecules at certain time points of simulation (100 or 200 ns) from 50% water:50% TFE to 100% water. The wild-type protein reconverted rapidly to its original structure, whereas the ΔPro-all mutant barely changed status (Fig. 7H). These results demonstrated that the structural flexibility and reversibility of importin-β is maintained cooperatively by multiple prolines in the α-helices that are facing the solvent.

The structural role of proline residues in the loop regions has been investigated in many proteins. They play a key role in the regulation of molecular conformation and function, as was revealed for the ion channel 5-hydroxytryptamine type-3, in which a proline residue in a loop region undergoes cis–trans isomerization and functions as an open–close switch (Lummis et al., 2005). Conversely, the structural and functional roles of prolines in helices have been studied less. Many examples of proline residues in α-helices were found in the protein structural database. Many of them formed a kink in the helix with fixed φ and ψ angles of −60° and +20° to +30°, respectively (Barlow and Thornton, 1988). These were mostly found in the transmembrane α-helices of channels and receptors, such as bacteriorhodopsin and transmembrane conductance regulator, to create a suitable channel space within a membrane (Henderson et al., 1990; Luecke et al., 1999; Mitsuoka et al., 1999; Wigley et al., 2002).

Prolines in α-helices are one of the unique characteristics of karyopherins and some other HEAT-repeat-containing proteins (Yoshimura and Hirano, 2016). Although these prolines locally break or kink each α-helix, the entire molecule retains a helical solenoid configuration when analyzed in water (Fig. 7A). This might be because the destabilizing effect of prolines is compensated by a stabilizing effect exerted by adjacent interacting α-helices in the helical repeat. Indeed, the substitution of proline to alanine largely affects the structure of isolated peptides, but not the entire molecule of importin-β (Fig. 7B). Analysis of the molecular trajectory showed no clear relation between proline residues and structural fluctuations of α-helices (Fig. 7E,F, and Fig. S7); proline-containing helices showed a similar level of fluctuation as proline-free helices in water. These results demonstrate that the destabilizing effect of prolines in each helix is compensated by interactions with adjacent helices in a long array of helices.

In contrast to the hydrophilic environments of the cytoplasm and nucleoplasm, the destabilizing effect of proline seems to be prominent in hydrophobic environments, such as the central channel of the NPC, in which a protein matrix made by FG-Nups prevents soluble proteins and other macromolecules from traveling through the pore. Phenylalanine residues and other hydrophobic groups interact with the outer helices of HEAT repeats and loosen helix–helix interactions. So far, several strong binding pockets for the FG motif have been identified in the cleft between two adjacent A-helices (Otsuka et al., 2008). In such a protein-rich environment, the destabilization effect of proline dominates the stabilizing effect from helix–helix interactions, and loosens the overall folding of the helical repeat, which further accelerates the interaction with FG-Nups. This conformational change is crucial for passing through the protein-rich environment of the NPC. Our results demonstrated that the ΔPro-all mutant was more stable and less flexible than the wild-type protein (Figs 6 and 7). It is reasonable to conclude that the substitution of proline to alanine deprives the α-helix of its destabilizing force, and prevents flexible conformational changes in an FG-rich protein crowding environment.

It should be noted that the effect of proline-to-alanine substitutions was more prominent in the C-terminal HEAT repeats than in the N-terminus. In comparison with the ΔPro-all mutant, the ΔPro-IV mutant exhibited similar subcellular localization (Fig. 2B) and FRAP at the nucleoplasm (Fig. 3C), but significantly faster FRAP at the nuclear envelope (Fig. 3B). Taken together with their similar folding stability (Fig. 6A), C-terminal HEAT repeats are likely to play a central role in maintenance of the flexibility of the entire molecule, which is important mainly in the release of importin-β into the nucleoplasm. In addition to the well-characterized N-terminal Nup-binding region around HEAT 5–7 (Bayliss et al., 1999; Chi and Adam, 1997; Kose et al., 1997), the C-terminal region corresponding to HEAT 14–16 was also shown to interact with Nups (Bednenko et al., 2003). Both the N- and C-terminal Nup-binding regions were shown to contribute to nuclear transport to a similar extent. Nups may have a preference for the N- or C-terminal binding sites of importin-β, as a C-terminal proline-to-alanine mutation affected interactions with only a subset of FG-Nups (Fig. 5D). As the effect of proline-to-alanine substitutions was observed not only in the proline-containing helices but also in the proline-free ones (Fig. 7E,F), it is reasonable to suggest that multiple prolines work cooperatively to ensure the global molecular structure of importin-β.

It is reasonable to assume that the ΔPro-all mutant of importin-β possesses less association efficiency and far less dissociation efficiency to the NPC than that seen in the wild-type protein. This can explain its reduced localization to the nucleus in vivo (Fig. 2B), slower accumulation in the nucleoplasm and nuclear envelope in vivo (Fig. 3) and smaller influx rate in vitro (Fig. 4). Therefore, the mutant has a longer retention time at the pore, blocks the endogenous transport factors, and might lead to severe cellular defects. Our findings provide direct evidence of a structure–function correlation in the nuclear transport mechanism: loss of adaptive molecular flexibility leads to an aberration in nuclear transport.

Expression of the ΔPro-all mutant in HeLa cells caused a severe defect in mitotic progression; the cells were arrested at prophase with an aberration in breaking down the nuclear envelope (Fig. 2). Importin-β is known to play an important role in mitosis, especially in the mid-to-late stage to organize the spindle and reassemble a functional nuclear envelope. Importin-β generally plays negative roles in these steps through its interaction with Ran (reviewed by Harel and Forbes, 2004; Hetzer et al., 2002), by chelating spindle-organizing factors and Nups (Ciciarello et al., 2004; Harel et al., 2003; Nachury et al., 2001). In contrast, little is known about the role of importin-β in nuclear envelope breakdown. A study using an in vitro nuclear disassembly system with Xenopus egg extracts demonstrated that excess RanGTP as well as the IBB and BSA-NLS inhibited nuclear envelope breakdown (Muhlhausser and Kutay, 2007). This finding suggests a role for importin-β in sequestering nuclear envelope proteins such as lamins or other inner nuclear membrane proteins, leading to the rupture of the nuclear envelope during prophase. Since the ΔPro-all mutant had reduced cargo transport activity (Fig. 2), its overexpression could disturb the subcellular localization of factors needed for nuclear membrane breakdown.

Depletion of Nup153 was shown to lead to a severe cellular defect at the early stage of mitosis, which resulted in a characteristic multilobed nucleus (Mackay et al., 2009). Since the mitotic defect induced by overexpressing the ΔPro-all mutant was observed at a similar stage of mitosis, there may be functional crosstalk between importin-β and Nup153 to regulate the early stage of mitosis. Analysis of the binding proteins that differentially associate with wild-type and mutant proteins will clarify the responsible factor for the observed mitotic arrest, and will reveal novel mechanisms of how importin-β controls nuclear envelope breakdown.

Karyopherin family proteins are found in a diverse range of eukaryotic species, and homologs of importin-β are found in most eukaryotes (O'Reilly et al., 2011). The amino acid sequences of importin-β1 from a variety of eukaryotic species were aligned against that of human importin-β1 to compare the number and position of proline residues in the HEAT repeats. As summarized in Fig. 8A, the positions of the prolines in importin-β1 are not well conserved at identical positions in distant species (Fig. 8A and Fig. S8, indicated by red boxes). However, judging from the human A-helix regions in the sequence alignment (Fig. S8, indicated by gray highlights), many prolines were found in potential A-helix regions of other species (Fig. S8, indicated by green boxes). When all of these potential A-helix prolines are taken into account, the importins of all species are predicted to contain a similar number of prolines in their A-helices, ranging from ∼12 to 15 (Fig. 8A). Although the degree of effect on molecular flexibility may be different for each proline residue, this suggests that the requirement for prolines in the structural maintenance of karyopherins follows a ‘generous rule’, in which their presence, but not exact position, is crucial.

It is intriguing that HEAT-rich proteins (such as PPP2R1A and CAND1) that do not belong to the karyopherin family also contain multiple proline residues in the A-helix (Fig. 8B). These proteins have been demonstrated to pass though the NPC and function as transport mediators (Fagotto et al., 1998; Koike et al., 2004; Wiechens and Fagotto, 2001; Yoshimura et al., 2014), suggesting that proline-containing HEAT repeats are involved in nuclear transport. Conversely, a number of HEAT repeats that do not contain proline in the A-helix (cohesion etc.) were found in the structural database. Although the consensus function of HEAT repeats is not understood fully (Yoshimura and Hirano, 2016), the fact that the bacterial protein YibA contains HEAT-like repetitive helices with multiple prolines in its outer α-helices (PDB: 1oyz) implies that the ancestral type of HEAT repeats might have contained proline residues in the A-helix. During evolution, some HEAT-repeat-containing proteins kept those prolines and started to become involved in nuclear transport. Other proteins might have lost their prolines and became involved in other cellular functions, such as mitotic chromosome segregation.

In addition to HEAT-repeat-containing proteins, proteins containing several amphiphilic helical motifs are able to pass through the nuclear pore spontaneously, including armadillo (ARM) repeats and spectrin repeats (Fagotto et al., 1998; Koike et al., 2004; Kumeta et al., 2010). Structural analyses of the typical ARM-containing proteins importin-α (PDB: 1ial) and β-catenin (PDB: 2bct) revealed the similar localization of multiple prolines in their outer helix (Fig. 8B). The total number of prolines in the outer helix is 8 and 10 for importin-α and β-catenin, respectively, which may be sufficient to maintain molecular flexibility and reversibility in a hydrophobic environment. Considering the nature of the NPC barrier, it is tempting to suggest that the proline-based strategy evolved a long time ago to achieve efficient nuclear transport.

The presence of proline residues in an α-helix has been regarded as problematic because of their ability to break its structure, and indeed there are several reports of proline mutations as a pathogenic mechanism (Yang et al., 1997). Here, we have reported the first example of the positive effect of proline residues in α-helices in the maintenance of a flexible molecular structure. It is interesting to note that prolines in protein molecules, which restrict polypeptide flexibility by possessing fixed φ and ψ angles, lead to the local destabilization of an α-helical structure, and finally contribute to the overall flexible molecular conformation of karyopherins (Fig. 8C). Thus, proline does not always serve as a ‘breaker’ of α-helical proteins, but rather behaves as a ‘protector’ of flexible molecular structure in some situations. Further studies focusing on the relations between molecular flexibility and functionality will uncover the cellular strategy used to orchestrate the proper function of molecules working in a variety of different subcellular environments.

The plasmid encoding GST-HA-tagged mouse importin-β was described previously (Otsuka et al., 2008). A series of importin-β mutation fragments were amplified by overlap extension PCR from GST-HA-tagged mouse importin-β (ΔPro-I, ΔPro-II, ΔPro-III, ΔPro-IV and ΔPro-all) and cloned into the pGEX-2T vector (GE Healthcare). The importin-β genes were excised and cloned into pEGFP-C1 vector (Clontech-Takara) for EGFP fusion expression. cDNA fragments encoding full-length human Nup54, FG-rich regions of rat Nup62 (amino acid residues 1–178), human Nup153 (amino acid residues 896–1475), and the IBB of rat importin-α (amino acid residues 1–69) were amplified by PCR from full-length cDNA and cloned into the pGEX6P-2 vector (GE Healthcare). cDNAs encoding NLS of SV40 large T antigen (MPPKKKRKVED) and human histone H3 were cloned into the pmCherry–C1 and pmPlum–C1 vector (Clontech).

A series of GST-fused wild-type and mutant importin-β proteins were expressed in Escherichia coli strain BL21 codon Plus (Agilent Technologies) and protein expression was induced by adding IPTG (0.5 mM at 18°C for 6 h). Finally, the proteins were dialyzed with 100 mM KPO₄ after the GST tag was cleaved by biotinylated thrombin (Novagen). GST-importin-α IBB expression was induced by 0.5 mM IPTG at 18°C for 3 h and dialyzed with transport buffer consisting of 20 mM HEPES-KOH (pH 7.3), 110 mM CH₃COOK, 2 mM (CH₃COO)₂Mg, 5 mM CH₃COONa, 0.5 mM EGTA and 1 mM DTT after purification. GST-Nups were expressed in the same way and immobilized on glutathione Sepharose (GE Healthcare) for the binding assay. For the gel filtration assay, purified importin-β (wild-type and ΔPro-all, ∼10 µg) was subjected to gel filtration chromatography (Superdex 200; GE Healthcare) in 50 mM HEPES-KOH (pH 7.4), 100 mM NaCl, and 1 mM β-mercaptoethanol. The elution profiles were recorded and the molecular masses were estimated by the elution pattern of the markers.

The CD spectra of purified proteins were measured using a JHL850 (Nihon-bunko) with a 0.1 cm cuvette at 20°C. Proteins were dissolved in 50 mM KPO₄ (pH 7.4).

The fluorescent measurements of bis-ANS (4,4′-dianilino-1,1′-binaphthyl-5,5′-disulfonic acid, dipotassium salt; Sigma-Aldrich) were performed with ARVO (Perkin Elmer) using emission and excitation wavelengths of 405 and 460 nm, respectively. Importin-β (final concentration, 0.5 µM) and bis-ANS (0.1–100 µM) were incubated in 50 µl transport buffer for 15 min at room temperature.

All cellular assays were performed using HeLa cells. For live cell observation, HeLa cells co-transfected with EGFP–importin-β and mPlum–Histone-H3 were observed in time-lapse under fluorescence microscopy (FV1200; Olympus) equipped with a stage-top heating chamber (Tokai Hit). For the nuclear transport assay, HeLa cells were washed with transport buffer, incubated with 0.05% digitonin at 0°C for 15 min, washed twice with transport buffer, and incubated at 37°C for 15 min. Purified GFP-fused proteins (1–5 µM) were added, and nuclear accumulation was observed in time-lapse under fluorescence microscopy (PASCAL 5; Zeiss). To check the integrity of the nuclear envelope, 10 µg/ml Alexa Fluor 568-labeled IgG was added to the sample. All analyses were performed by MetaMorph software (Molecular Imaging). Curve fitting and other kinetic analyses of the obtained data were performed using Origin software (LightStone).

Molecular dynamics simulations were conducted as described previously (Yoshimura et al., 2014), with the Amber11 package and an ff99SB force field. A GPU version of the particle mesh Ewald method was used with an NVIDIA Tesla M2090. The time step was set to 2 fs. Temperature was maintained at 310 K using Langevin dynamics with a collision frequency of 1.0 ps⁻¹; pressure was kept at 1 atm with a relaxation time of 2.0 ps. Bonds involving hydrogens were constrained with the SHAKE algorithm, and the long-range interaction cut-off was set to 10.0 Å. Before starting the production runs, the system underwent 100–500 ps equilibration with weak position restraint on the protein. The obtained datasets are available from the corresponding author upon reasonable request.

EGFP-fused importin-β (wild-type and mutants) was expressed in HeLa cells. FRAP analysis was performed with a laser scanning confocal microscope (FV-1200; Olympus) equipped with a 37°C heating stage and 5% CO₂ supply. During the time-lapse observations (1.7 s/frame), the nucleus was irradiated by a bleach laser (488 nm) for 1 s. After bleaching, the time-lapse observations were continued for 100 s. Fluorescent recovery for the nuclear membrane region and nucleoplasm was quantified and analyzed statistically. Curve fitting was performed using data analysis software (Origin; LightStone) with the following fitting functions: for the nuclear envelope, y=I_final×(1−(k_on/(k_on+k_off))×exp(−k_off×x)); for the nucleoplasm: y=exp(−x×k_out)×Y₀+(k_in/k_out)×(1−exp(−x×k_out)).

Purified wild-type and mutant importin-β were mixed with various concentrations of urea (0–9 M) (final concentration of importin-β was 0.1 mg/ml in 50 mM KPO₄, pH 7.4) and/or TFE (0–50%). The fluorescence spectrum of the protein solution was measured with an excitation wavelength of 320 nm by a fluorescence spectrometer (FP8300; JASCO). The center wavelength of the emission peak was plotted against the concentration of urea. The m-value and folding free energy (ΔG₀) were obtained as described previously (Eftink, 1994).

We thank S. Iwasaka, K. Ogawa, Y. Takashima, S. Asai and R. Watanabe for technical assistance.