Crystal Structure of Human Importin-α1 (Rch1), Revealing a Potential Autoinhibition Mode Involving Homodimerization

In this study, we determined the crystal structure of N-terminal importin-β-binding domain (IBB)-truncated human importin-α1 (ΔIBB-h-importin-α1) at 2.63 Å resolution. The crystal structure of ΔIBB-h-importin-α1 reveals a novel closed homodimer. The homodimer exists in an autoinhibited state in which both the major and minor nuclear localization signal (NLS) binding sites are completely buried in the homodimerization interface, an arrangement that restricts NLS binding. Analytical ultracentrifugation studies revealed that ΔIBB-h-importin-α1 is in equilibrium between monomers and dimers and that NLS peptides shifted the equilibrium toward the monomer side. This finding suggests that the NLS binding sites are also involved in the dimer interface in solution. These results show that when the IBB domain dissociates from the internal NLS binding sites, e.g., by binding to importin-β, homodimerization possibly occurs as an autoinhibition state.

Importin-α works as an adaptor molecule to connect importin-β and the NLS cargo, which makes it possible for the limited variety of importin-βs to bind to a large variety of NLS cargoes. The hetero-trimer is then transported into the nucleus through the NPC, with the importin-β working as an active engine. Such a transportation system ensures that larger cargoes (Mw > 40 kDa) can travel through the NPC into the nucleus [9][10][11][12]. There seems to be some mechanism in the isoforms of human importin-αs that is responsible for the diversity of binding to a variety of NLS cargoes, even though the whole structures of the subfamilies closely resemble each other. On the other hand, some importin-αs are known to form homodimers. The crystal structure of yeast importin-α was first reported as a homodimer [13]. In the homodimer, major and minor NLS binding sites are still exposed to solution; thus, NLS peptides can bind to the sites. In addition, Xenopus importin-α seems to form homodimers or multimers during purification [14]. However, the functional aspects of these multimerization properties of importin-αs have not been resolved.
Here, we report homodimeric human importin-α1 (h-importin-α1) by the use of X-ray crystallography and shed light on the solution state of the homodimerization in relation to NLS binding by using analytical ultracentrifugation (AUC) and isothermal titration calorimetry (ITC).

Protein preparation
h-importin-α1 (1-526) and ΔIBB-h-importin-α1 were overexpressed by E. coli strain BL21 (DE3) Codonplus-RIL (Stratagene) and purified as follows: Frozen cells (10 g) obtained from a 1 L culture were suspended in 100 mL of sonication buffer (50 mM tris(hydroxymethyl)aminomethane (Tris)-HCl, 500 mM NaCl, 1 mM dithiothreitol (DTT), 1 mM ethylenediaminetetraacetic acid (EDTA), pH 8.0) and lysed by sonication using an Ultrasonic Processor VCX 500 (Sonics & Materials, Inc.). The lysate was collected and centrifuged at 15,000 rpm for 1 h. The supernatant was filtered through a 0.45 μm HT Tuffryn membrane syringe filter (Pall Corporation). A GSTrap FF column (GE Healthcare) packed with Glutathione Sepharose 4 Fast Flow (GE Healthcare) was pre-equilibrated with phosphate-buffered saline (PBS) buffer. This lysate was then loaded onto a column at a flow rate of 1 mL/min. After washing with 50 mL wash buffer (50 mM Tris-HCl, 150 mM NaCl, 1 mM DTT, 10% glycerol, 1 mM EDTA, pH 8.0), bound GST-tagged h-importin-α1 was eluted with 30 mL elution buffer (50 mM Tris-HCl, 150 mM NaCl, 1 mM DTT, 10% glycerol, 1 mM EDTA, 5 mM reduced glutathione, pH 8.2). Elution fractions containing GST-h-importin-α1 and GST-ΔIBB-h-importin-α1 were pooled. The PreScission protease (GE Healthcare) was incubated with the pooled fractions according to the manufacturer's instructions, and the solution was dialyzed against 2 L of digestion buffer (50 mM Tris-HCl, 150 mM NaCl, 1 mM DTT, 1 mM EDTA, 0.01% Triton X-100, pH 7.5) for 12 h at 277 K. The digested protein was filtered through a 0.45 μm HT Tuffryn membrane syringe filter and loaded onto a GSTrap FF column (GE Healthcare). The cleaved h-importin-α1 and ΔIBB-h-importin-α1 were collected in the flow-through and washed fractions. The protein samples were diluted to obtain an NaCl concentration of 75 mM and a pH of 8.0 with 20 mM Tris-HCl, 1 mM DTT, 1 mM EDTA, and 10% glycerol. The samples were loaded onto a Resource-Q column (GE Healthcare) packed with SOURCE 15Q (GE Healthcare) (S1 Fig.). For elution, the column was developed with a linear gradient from 0 to 500 mM NaCl in 20 mM Tris-HCl, 1 mM DTT, 1 mM EDTA, 10% glycerol, pH 8.0, at a flow rate of 0.2 mL/min. The eluted fractions of interest were pooled and concentrated with a Centriprep YM-10 centrifugal device (Amicon/Millipore) to a concentration of 10 mg/mL and used for the ITC and AUC-SV experiments. For the crystallization experiments, these fractions were filtered and loaded in a pre-equilibrated HiLoad 16/60 Superdex 75 size exclusion column (GE Healthcare) with 20 mM Tris-HCl, pH 8.0, 200 mM NaCl, and 5 mM DTT. The eluted fractions were collected and concentrated with the Centriprep YM-10 to 10 mg/mL.

Analytical ultracentrifugation (AUC)
AUC sedimentation velocity (AUC-SV) experiments were carried out in AUC buffer (0.1 M Tris-HCl, pH 8.0, 0.2 M NaCl, 1 mM tris(2-carboxyethyl)phosphine (TCEP)), using an Optima XL-I analytical ultracentrifuge equipped with two optical systems, the Rayleigh interference and absorbance systems (Beckman Coulter). For AUC-SV measurements, 10 mg/mL h-importin-α1 and ΔIBB-h-importin-α1 samples were diluted with the AUC buffer. Then, 4 mM SV40 and nucleoplasmin NLS solutions, dissolved in the AUC buffer, were added to the sample solution. The AUC-SV measurements were conducted at 50,000 rpm at a temperature of 293 K, using an An-50 Ti rotor featuring cells with a standard 12-mm charcoal-epon double sector centerpiece and sapphire windows. During the runs, changes in the concentration gradient were monitored with a Rayleigh interference optical system or absorbance at 280 nm. All of the AUC-SV raw data were analyzed by the program SEDFIT14.1, with the continuous C(s) distribution model [16]. The SEPHAT 10.58d program was used for analysis of the isotherm of weight-average s-values. A monomer-dimer self-association model was applied [17].

Isothermal titration calorimetry (ITC)
We investigated the binding of SV40 NLS and nucleoplasmin NLS to h-importin-α1 and ΔIBB-h-importin-α1. ITC measurements were carried out using a MicroCal iTC 200 (GE Healthcare) in 0.1 M Tris-HCl, pH 8.0, 0.2 M NaCl, and 5 μM DTT at 293 K. The cell chamber was loaded with 200 μL of 15 μM h-importin-α1 and titrated 20 times with 2 mM of SV40 NLS solution or 2 mM nucleoplasmin NLS solution, while stirring at 1,000 rpm. The data analysis was performed with Origin software equipped with the ITC (Fig. 3A). ΔIBB-h-importin-α1 (17 μM) was used for the ITC experiment with 40 titrations for SV40 NLS binding and 30 titrations for nucleoplasmin NLS (Fig. 3B, C).

Diffraction data collection, structure determination and validation
All diffraction data were collected at the BL26B2 station of SPring-8 using the mail-in data collection system [18], [19]. The crystals were soaked in the solution containing 30% (w/v) PEG 8000 as a cryoprotectant and flash-frozen at 100 K. The diffraction data were processed using the HKL2000 package [20]. The molecular replacement technique was executed using the MrBump package [21] using the truncated molecular model (84-476) of importin-α5 (PDB ID: 2JDQ) as a template, which gave two solutions in the asymmetric unit. The initial model was improved using RESOLVE [22] and further refined using LAFIRE [23] and ARP/wARP packages [24]. The model was manually rebuilt with the help of XtalView [25] and Coot [26], and refined with CNS [27] and PHENIX [28]. The current model of the ΔIBB-importin-α1 involves the residues of 75-496 (chain A) and 75-497 (chain B) with an R/R free factor of 0.193/ 0.221. The structural validation was performed with the Procheck [29] and MolProbity [30] programs. The coordinate and structure factors of ΔIBB-h-importin-α1 have been deposited at the Protein Data Bank with PDB ID: 3WPT.

AUC-SV experiments
shows that it is in a concentration-dependent monomer-dimer equilibrium (Fig. 1A, B). The C(s) distributions are characterized by two peaks, one at 3.1-3.3 S and one at 3.9 S; the relative sizes of the peaks vary with protein concentration, and they are interpreted as corresponding to monomer and dimer, respectively. The C(s) distributions were used to create a weight average S value (sw) isotherm. We calculated the sw by integration of the C(s) curves between 2.5 S and 5.0 S for each protein concentration, and then plotted the sw vs protein concentration using SEDPHAT (Fig. 1C). The monomer-dimer self-association model was applied for the curve fitting. The monomerdimer KD is estimated to be 8 ± 3 μM. The NLS ligands, SV40 and NP, shift the equilibrium to the monomer side (Fig. 1D, E). These results show that the NLS binding sites are involved in the dimer interface in solution.
h-importin-α1. However, h-importin-α1 remains monomeric in a variety of protein concentration (Fig. 1F). Neither of the SV40 nor NP NLS peptides changed the monomeric state (Fig. 1G, H). These results show that the self-bound IBB domain autoinhibits not only NLS binding but also homodimerization.

X-ray crystallography of ΔIBB-h-importin-α1
Only the diffractive crystals of ΔIBB-h-importin-α1 were obtained, not those of h-importin-α1. The crystals diffracted X-rays to 2.6 Å resolution in SPring-8. The molecular replacement technique was used to solve the structure ( Table 1). The crystal structure of ΔIBB-importin-α1 is shown in Fig. 2A. The pseudo 2-fold axis lies in close proximity to the N239 of chains A and Table 1. Crystallographic data collection and refinement statistics of ΔIBB-h-importin-α1.

Space group
Tetragonal P4  B, which are in the disallowed region in the Ramachandran plot. Structural stress is probably put on them owing to the dimerization. The K108 residue is extensively involved in the dimer formation and is represented in the close-up views. The side chain of K108 enters the P1 0 site of the minor NLS binding site [31], making hydrogen bonds with D325, T328, and Q369 of another protomer and one water molecule (Fig. 2B). The P1 0 site comprises V321, T322, D325, T328, N361, I362, G365, and Q369, which are conserved in the minor NLS binding site of importin-α proteins (Fig. 3). The major (W142, N146, W184, N188, W231, N235, W273,  Y277) and minor (W357, N361, W399, N403) NLS binding sites are buried in the dimer interface (Fig. 2C), thus NLS peptides cannot access the NLS binding sites. Because the major NLS binding site involves typically P1-P7 binding pockets [32], it has higher NLS binding affinity than the minor NLS binding site with P1 0 -P2 0 . The P1 0 -binding pocket in the minor NLS binding site plays a common role in a variety of importin-αs; the P1 0 accepts the K/R residues of the typical NLS signal motif KRXX [31][32][33], (Fig. 4). In the present crystal structure of ΔIBB-h-importin-α1, each of the K108 residues enters into the P1 0 -binding pocket of another protomer so that the minor NLS binding sites are in a state of autoinhibition (Fig. 2B), and, together with the major NLS binding sites, are buried in the dimer interface (Fig. 2C). As a result, the closed homodimer is stabilized owing to the limited access of NLS peptides to the major and minor NLS binding sites.

Isothermal titration calorimetry (ITC) assay for binding of NLS peptides to h-importin-α1 and ΔIBB-h-importin-α1
h-importin-α1. We carried out a binding assay of h-importin-α1 and ΔIBB-h-importin-α1 to study autoinhibition resulting from homodimerization. SV40 NLS bound to h-importin-α1 with a KD value of~100 μM (Table 2). While the fitting to the ITC data, one-site binding model was used because two-site model did not converge to adequate values. The AUC-SV experiments reveal that h-importin-α1 is monomeric during SV40 NLS binding (Fig. 1G). This result suggests that both the IBB domain and the NLS peptides bind to the internal NLS binding sites and prevent homodimerization.
On the other hand, the bipartite nucleoplasmin NLS peptide did not bind to h-importin-α1 (Fig. 3D). This result suggests that NLS binding is restricted by the IBB domain. ΔIBB-h-importin-α1. ΔIBB-h-importin-α1 bound to SV40 NLS with KD =~6 μM and to nucleoplasmin NLS peptides with KD =~0.14 μM (Table 2). During the NLS binding, the monomer-dimer equilibrium shifted to the monomer side (Fig. 1D, E), which suggests that the NLS binding sites are involved in the dimer interface even in solution, as in the crystal structure.

Homodimerization found in importin-αs
It has been shown or suggested that the importin-αs from yeast and Xenopus eggs form dimers. Yeast importin-α (PDB IB: 1BK6) provided the first crystal structure of an importin-α that was shown to dimerize not only in the crystal but also in solution, as revealed by dynamic light scattering (DLS) [13]. The SV40 NLS peptides are, however, able to bind to the NLS binding sites because the homodimer is an "open" structure where the NLS binding sites are exposed to the solution (Fig. 4A). The open homodimerization probably attributes to H119 at the top of the loop; this loop corresponds to the K108 loop of ΔIBB-importin-α1. H119 cannot enter the P1 0binding pocket in the minor NLS binding site, and this inability causes the protomers to slide along the long axis of the molecules. The NLS binding parameters of yeast importin-α were analyzed by fluorescence depolarization assay and it was found that yeast importin-α (1-542) binds to neither SV40 NLS-GFP nor Myc NLS-GFP [34]. On the other hand, yeast ΔIBBimportin-α (89-530) binds to both SV40 and Myc NLS, with KDs of 9 ± 4 nM and 6 ± 3 nM, respectively [35]. In addition, an enzyme-linked immunosorbent assay (ELISA)-based assay revealed that yeast importin-α (1-542) does not bind to SV40 NLS, whereas yeast ΔIBB-importin-α (88-530) binds very tightly [35]. These results suggest that the IBB domain is mainly responsible for the autoinhibition activity of yeast importin-α, while the "open homodimerization" does not restrict the NLS binding. Thus, the NLS binding affinity of yeast importin-α changes dramatically from 0% (off), when the IBB autoinhibits itself, to 100% (on), when it does not, that is keenly contrast to the closed-homodimer of h-importin-α1.
Moreover, it has been reported that wild importin-α from Xenopus eggs is homodimerized during purification [14], and a variety of studies have revealed that purified Xenopus importinα is in dimer-monomer equilibration in the concentration range of 0.6~1.2 mM, with a dissociation constant KD of~20 μM [36]. We found that the Xenopus importin-α has an arginine residue (R105) that is aligned to the K108 of importin-α1 (S2 Fig.). Thus, the R105 possibly interacts with the P1 0 -binding pocket of another protomer, promoting the closedhomodimerization.
The first structural knowledge about the autoinhibition mechanism was obtained from the crystal structure of mouse-importin-α2 (m-importin-α2) [37], [38]. In that crystal structure, the IBB domain binds to the major, but not the minor, NLS binding site; thus the NLS binding affinity is partly autoinhibited. It has been unclear whether m-importin-α2 forms a homodimer or not [37], although the m-importin-α2 has a K102 and the P1 0 -binding pocket consists of the same residues as in h-importin-α1 (S2 Fig.). Further studies should be addressed about the multimerization property of the m-importin-α2.

A possible autoinhibition state of h-importin-α1
Together with the results obtained from the AUC-SV, ITC, and crystallography, we can add a potential autoinhibition mode of the h-importin-α1 to the conventional scenario, as shown in Fig. 5. In the canonical scheme, the self-bound IBB domain autoinhibits NLS binding to the NLS binding sites in h-importin-α1. Our current study reveals that the self-bound IBB domain also autoinhibits homodimerization. The IBB domain is dissociated from the NLS binding sites, e.g., by the competitive binding of NLS peptides to the NLS binding sites and/or the binding of the IBB domain to importin-β (Fig. 5A). Upon the IBB domain dissociation, the himportin-α1 can homodimerize as an autoinhibition state (Fig. 5B), which possibly lowers the KD (~6 μM) for NLS comparing to that of yeast (~10 nM) [34]. Through the association/dissociation of the IBB domain and molecular h-importin-α1, the NLS binding to the h-importin-α1 is controlled by both of the self-binding of IBB domain and homodimerization.

Conclusions
In h-importin-α1, NLS binding affinity is autoinhibited by the self-binding of IBB domain. In addition, the closed homodimerization possibly works as an autoinhibition state when the IBB domain dissociates from the NLS binding sites. There still remains unresolved subjects whether the current closed-homodimer is physiologically relevant in vivo. Further studies are required to address the functional aspects of the closed-homodimer, by mutagenesis and cellular biological experiments.