Cell surface localization of importin α1/KPNA2 affects cancer cell proliferation by regulating FGF1 signalling

Importin α1 is involved in nuclear import as a receptor for proteins with a classical nuclear localization signal (cNLS). Here, we report that importin α1 is localized to the cell surface in several cancer cell lines and detected in their cultured medium. We also found that exogenously added importin α1 is associated with the cell membrane via interaction with heparan sulfate. Furthermore, we revealed that the cell surface importin α1 recognizes cNLS-containing substrates. More particularly, importin α1 bound directly to FGF1 and FGF2, secreted cNLS-containing growth factors, and addition of exogenous importin α1 enhanced the activation of ERK1/2, downstream targets of FGF1 signalling, in FGF1-stimulated cancer cells. Additionally, anti-importin α1 antibody treatment suppressed the importin α1−FGF1 complex formation and ERK1/2 activation, resulting in decreased cell growth. This study provides novel evidence that functional importin α1 is located at the cell surface, where it accelerates the proliferation of cancer cells.

Furthermore, many studies have recently reported that importin α 1 is highly expressed in diverse types of cancers, including breast cancer, hepatocellular carcinoma, lung cancer, melanoma, and ovarian cancer [13][14][15][16] . Such aberrant importin α 1 expression is often correlated with an adverse outcome in patients 13 . Although subcellular localization of importin α 1 is diffuse throughout cells 17 , it has been shown that importin α 1 is also detected in the sera of lung cancer patients 18 . However, it is still poorly understood how importin α 1 is involved in cancerous processes.
In this study, using a combination of flow cytometric, biochemical, and confocal microscopic approaches, we show for the first time that importin α 1 is localized to the cell surface in several human cancer cell lines. Furthermore, we found that importin α 1 at the cell surface is associated with a growth factor, FGF1, thereby enhancing its signalling pathway and accelerating the proliferation of cancer cells. This is the first evidence showing that proteins that ordinarily function within cells can localize to the cell surface where they participate in novel physiological activities.

Results
Importin α1 is localized to the cell surface in some cancer cell lines. Recently, we performed cellbased proteomic experiments using human vascular endothelial cells to screen for cell surface protein targets that may be involved in systemic sclerosis 19 . Among this proteomic data, we noticed that importin α 1 (Importin subunit alpha-1) was included as a potential cell surface protein 19 (Supplementary Table S1). Furthermore, we performed another proteomic analysis aimed at novel cell surface marker discovery, by using colon cancer cells and tissues. Membrane fraction proteins that had been separated by homogenization and centrifugation also included importin α 1 (Supplementary Table S1). Given that high levels of importin α 1 expression have been reported in various types of cancers 13 , we assessed whether importin α 1 is actually localized at the cell surface by performing flow cytometric analysis using two different antibodies against importin α 1 in several types of cancer cell lines. These included the lung cancer cell lines A549 and PC9, gastric cancer cell lines KATOIII and AGS, the colon cancer cell line HCT116, hepatocellular carcinoma cell lines HepG2, Hep3B, and HLE, breast cancer cell lines MRK-nu-1, MCF-7, SKBr3, and MDA-MB-231. We also analysed three normal cell lines: a human fibroblast cell line TIG-1, human dermal microvascular endothelial cells (dHMVECs), and a normal human mammary epithelial cell line MCF-10A. These cell lines express endogenous importin α 1 to some extent (Fig. 1a). As shown in Fig. 1b, we detected cell surface importin α 1 in PC9, HCT116, KATOIII, HepG2, Hep3B, and HLE cells, while no signals were detected in A549 and AGS cells, or in the three normal cell lines. More particularly, relatively higher levels of cell surface importin α 1 were detected in all hepatocellular carcinoma cell lines. On the other hand, none of the four tested breast cancer cell lines expressed importin α 1 at the cell surface, suggesting that the cell surface localization of importin α 1 may occur in a cell type-dependent manner. Similar results were obtained using the other anti-importin α 1 antibody ( Supplementary Fig. S1). Thus, for further studies, we mainly focused on colon cancer and hepatocellular carcinoma cell lines.
To confirm the findings above, we next performed a cell surface biotinylation assay, in which cell surface proteins of HCT116, HepG2, and AGS cells were first biotinylated, and the cell lysates were then incubated with avidin beads to precipitate the cell surface biotinylated proteins, followed by immunoblot analysis. As shown in Fig. 1c, we detected a single band of importin α 1 in the biotinylated cell surface fraction of HCT116 and HepG2 cells, whereas faint importin α 1 was detected in that of AGS cells.
Furthermore, we attempted to visualize cell surface expression of importin α 1 in HCT116 cells using confocal laser scanning microscopy. Intracellular importin α 1 showed diffuse localization throughout the cell after detergent-based cell permeabilization after fixation (Fig. 1d, + Triton), as reported previously 17 . In contrast, cell surface staining of importin α 1 was clearly observed in non-permeabilized cells, with only faint signals in the cytoplasm (Fig. 1d, − Triton). In addition, we found that importin α 1 was co-localized with Cholera toxin B (CtxB), which is a known cell surface marker, confirming that importin α 1 was expressed on the cell surface.
To further confirm that importin α 1 is localized to the cell surface, HCT116 cells were transiently transfected with the N-terminal Flag-tagged importin α 1 expression vector. Flow cytometric analysis using either anti-Flag antibody (N-terminus, Fig. 1e) or anti-importin α 1 antibody (which recognizes the C-terminal side, Fig. 1f) both showed positive signals in importin α 1-overexpressing cells, indicating that the whole importin α 1 molecule is exposed to the extracellular milieu. Of note, the amount of cell surface importin α 1 was markedly augmented in HCT116 cells expressing Flag-tagged importin α 1, as compared with that in the cells transfected with the control vector (Fig. 1f). Taken together, we concluded that importin α 1 is present on the surface of these cancer cells.
Cell surface importin α1 is functional. The predominant feature of cNLS is a sequence comprising one or two short clusters of basic amino acids, such as lysine or arginine. The best characterized NLS is the monopartite type found in the SV40 large T-antigen (PKKKRKV), which is recognized by importin α 1. Therefore, next, to address whether the cell surface importin α 1 is functional, we examined whether it binds to a typical NLS-containing substrate, the SV40 large T antigen NLS-fused GST and GFP (GST-NLS-GFP), which is known to bind to importin α 1 20 (Fig. 2a). HCT116 cells were incubated with GST-NLS-GFP or GST-GFP (as a negative control). Immunofluorescent analysis revealed GST-NLS-GFP on the cell surface (Fig. 2b). Comparable data was also obtained by flow cytometric analysis using HCT116 cells (Fig. 2c).
Furthermore, to examine whether the association of GST-NLS-GFP with HCT116 cells is dependent on importin α 1, we manipulated the cell surface expression levels of importin α 1 by knockdown or overexpression procedures. Knockdown of importin α 1 by siRNA significantly decreased the amount of cell surface importin α 1 ( Supplementary Fig. S2), leading to a substantial decrease of GST-NLS-GFP association with the surface of HCT116 cells (Fig. 2d). In contrast, the overexpression of importin α 1 resulted in a significant increase in the intensity of GST-NLS-GFP fluorescence on the cell surface (Fig. 2e). These data indicated that the cell surface Scientific RepoRts | 6:21410 | DOI: 10.1038/srep21410 importin α 1 is able to recognize proteins with cNLSs. Collectively, these data indicated that cell surface importin α 1 is functional, and not denatured.
Importin α1 is detected in cultured medium. It has previously been reported that importin α 1 is detected in the sera of patients with lung cancer 18 . Therefore, we assessed whether importin α 1 could be detected in cultured medium of cancer cell lines. After the removal of dead cells and debris by both centrifugation and filtration, the importin α 1 protein levels in the supernatants of the cell culture medium were analysed. Immunoblot analysis showed the presence of importin α 1 in the supernatants of HCT116 and HepG2 cultures (Fig. 3a), whereas no importin α 1 was detected in the supernatants of AGS (Fig. 3a) and A549 cultures (data not shown), showing that cell surface localization and extracellular release of importin α 1 was correlated. Trypan blue exclusion assays indicated that the protein release was not due to a lack of cell integrity (cell viability > 90%; data not shown). Additionally, no extracellular release of lamin A/C was detected (Fig. 3a). Thus, these results demonstrated that cancer cells release importin α 1 into the extracellular space under normal culture conditions, and that this is correlated with its cell surface localization.
Next, to exclude the possibility that importin α 1 in cultured medium was enclosed by a lipid layer, e.g., exosomes, we measured free importin α 1 in cultured medium by a sandwich ELISA using two different anti-importin α 1 antibodies. Under non-denatured conditions, free importin α 1 was detected in the supernatant of HepG2 cultures, while the amount of importin α 1 in the supernatant of AGS culture was comparable to background levels (Fig. 3b), indicating that cell surface importin α 1-positive cancer cells release a free form of importin α 1 into the extracellular milieu.
Extracellular importin α1 associates with heparan sulfate on the cell surface. These findings led us to suppose that the importin α 1 released into the extracellular space is attached to the cell surface. To verify this, we added recombinant Flag-importin α 1 proteins to the cultured medium. As shown in Fig. 4a,b, we found  apparent binding of the recombinant importin α 1 to the cell surface of not only HCT116 cells and HepG2 cells, but also to that of AGS cells that were negative for cell surface localization of endogenous importin α 1. These data indicated that importin α 1 interacts with the cell membrane after its release into the extracellular space.
Next, we attempted to address how the released importin α 1 binds to the cell surface. It is well known that many growth factors, such as fibroblast growth factors (FGFs), vascular endothelial growth factor (VEGF)-A and hepatocyte growth factor (HGF), have cationic amino-acid clusters 21,22 , and interact with heparan sulfate (HS), which includes heparin, which in turn is composed of one or more unbranched anionic polysaccharide(s) known as glycosaminoglycans (GAGs) [22][23][24] . Since importin α 1 contains a hydrophilic cluster, including lysine Flag staining was visualized using confocal laser microscopy. Scale bars: 10 μ m. (b) Living cancer cell lines were treated with PBS or recombinant importin α 1 (100 ng/ml). These cells were stained with anti-importin α 1 antibody (Ab), and subjected to flow cytometric analysis. Dashed line, isotype control Ab. (c) GST or GSTtagged importin α 1 at concentrations of 0.5, 5, or 50 pmol were immobilized on heparin-beads. The bound proteins were analysed using immunoblotting with anti-GST Ab. GST-tagged-FGF1 and -FGF2 (50 pmol) were used as positive controls. (d) GST-tagged importin α 1 mutants, GST-GFP or IBB domain-fused GST-GFP at concentrations of 50 pmol were immobilized on heparin beads. The bound proteins were analysed using immunoblotting with anti-GST Ab. Band intensities of each protein bound to heparin-sepharose were normalized with that of each input sample using the Image J program. (e) Supernatants of cultured cancer cell lines were collected and immobilized on heparin-conjugated beads. Immunoblot analysis was performed using Abs to importin α 1. (f) Living HCT116 cells were treated with recombinant Flag-tagged importin α 1 (100 ng/ml) in the presence or absence of heparin (10 U/ml). These cells were stained with anti-importin α 1 Ab (Left panel) or anti-Flag antibody (Right panel), and subjected to flow cytometric analysis. Gray area, isotype control Ab. (g) Living HCT116 cells were pretreated with or without heparinase III (0.2 U/ml) for 6 h, and then treated with recombinant Flag-tagged importin α 1 (100 ng/ml). These cells were stained with anti-importin α 1 antibody, and subjected to flow cytometric analysis. Gray area, isotype control Ab. and arginine residues, in its importin β binding (IBB) domain, conferring cationic properties to the protein, we hypothesized that importin α 1 might be recruited to the cell surface through binding to HS. To assess this, we performed an in vitro solution-binding assay, in which recombinant importin α 1 was mixed with heparin-conjugated beads. As expected, we found that heparin directly bound to importin α 1 (Fig. 4c). To identify the domains of importin α 1 important for binding to heparin, we prepared several mutants of importin α 1, including an IBB domain-deleted mutant (Δ IBB) and a C-terminal region-deleted mutant (Δ C). The in vitro binding assay indicated that the association of the Δ IBB mutant with heparin was markedly weaker than that of the wild type and the Δ C mutant of importin α 1 (Fig. 4d). In addition, we found that the IBB domain-fused GST-GFP proteins strongly bound to the heparin-beads (Fig. 4d), indicating that importin α 1 was capable of associating with heparin via its IBB domain. In addition, when we applied heparin-beads to the cultured medium of cancer cells to determine whether the association between importin α 1 and HS occurred in the extracellular space, rather than within cells, the heparin-conjugated beads significantly concentrated importin α 1 in the cultured medium of HCT116, HepG2, PC9, and Hep3B cells, but not that of AGS and A549 cells (Fig. 4e and Supplementary Fig. S3).
To clarify the interaction between HS and importin α 1 on the cell surface, HCT116 cells were incubated with recombinant importin α 1 in the presence of free heparin. Association of importin α 1 with the cells was clearly reduced by the addition of heparin (Fig. 4f). Moreover, when HCT116 cells were incubated with heparinase III, an enzyme that cleaves HS in cell surface proteoglycans, prior to importin α 1 treatment 25 , we observed an apparent reduction of importin α 1 binding to the cells (Fig. 4g). From these results, we concluded that importin α 1 that has been released into the extracellular environment binds to HS on the cell surface.
Cell surface importin α1 affects FGF1 signalling. We next investigated which endogenous proteins associate with the cell surface importin α 1. It has been reported that some secreted proteins, such as FGFs and epidermal growth factors (EGFs), are localized to and function in the nucleus, and that most of these proteins have their own cNLSs that are recognized by the importin α family and translocated into the nucleus [26][27][28][29] (Supplementary Table S2). This raised the possibility that these proteins interact with the cell surface importin α 1. Therefore, we selected candidates among these NLS-containing secreted proteins by ELISA screening. We found that the cNLS-containing secreted growth factors, such as FGF1, FGF2, insulin-like growth factor-binding protein (IGF-BP)3, and IGF-BP5, but not IFN-γ , bound to importin α 1 ( Supplementary Fig. S4). Furthermore, using an in vitro binding assay, we confirmed that recombinant GST-tagged FGF1, FGF2, and IGF-BP5 proteins interacted with importin α 1 (Fig. 5a). In this study, we focused on FGF1 for further experiments.
We investigated whether the cell surface importin α 1 affects FGF1 signalling in cancer cells. In HCT116 cells, the exogenous addition of FGF1 slightly increased cell proliferation (Fig. 5b). Under the same assay conditions, further addition of importin α 1 slightly, but significantly, accelerated the proliferation of FGF1-stimulated HCT116 cells (Fig. 5b).
Next, to determine how the cell surface importin α 1 is involved in the cellular response to FGF1, we monitored the activation of extracellular signal-regulated kinase 1/2 (ERK1/2), major targets of the FGF signalling pathway [30][31][32] . We found that the addition of importin α 1 increased the phosphorylation of ERK1/2 in FGF1-stimulated HepG2 and HCT116 cells at an early stage (10 min) after stimulation (Fig. 5c), although this effect did not last up to late stages (1− 6 h; Fig. 5d), meaning that importin α 1 may function at the start of FGF1 signalling.
To confirm the involvement of the cell surface importin α 1 in FGF1 signalling, we performed an RNAi study using HepG2 cells. Knockdown of importin α 1 resulted in marked down-regulation of the activation of ERK1/2 in HepG2 cells in the presence of FGF1, and this suppressive effect was restored by the addition of exogenous importin α 1 under the FGF1 stimulation (Fig. 5e). These results indicated that extracellular importin α 1 affects FGF1 signalling on the cell surface.
To further confirm that importin α 1 affects FGF1 signalling, we applied anti-importin α 1 monoclonal antibody (mAb) to block the function of the extracellular and cell surface importin α 1. We found that anti-importin α 1 mAb impaired the interaction of importin α 1 with FGF1 (Fig. 5f), and decreased the importin α 1-mediated activation of ERK1/2 in response to FGF1 (Fig. 5g). These results indicated that the interaction of importin α 1 with FGF1 plays an important role in the FGF1-mediated proliferative signalling (Fig. 5g).
Finally, we treated cancer cells with the anti-importin α 1 mAb to establish whether the anti-importin α 1 mAb can indeed diminish the proliferation of cancer cells. In HCT116 and HepG2 cells, cell proliferation was significantly reduced by anti-importin α 1 mAb treatment ( Fig. 5h and Supplementary Fig. S5), while the growth of AGS cells was not affected by the antibody (Fig. 5i). Taken together, these results demonstrated that the cell surface importin α 1 plays crucial roles in the proliferation of colon cancer and hepatocellular carcinoma cells.

Discussion
This study demonstrates the novel cellular localization and function of importin α 1. That is, importin α 1 is localized to and functions at the cell surface of some cancer cells. Indeed, we showed that cell surface importin α 1 associates with growth factors, such as FGFs (FGF1 and FGF2), and enhances FGF1 signalling. Furthermore, we found that the treatment with an anti-importin α 1 antibody impaired FGF1 signalling, resulting in the suppression of cancer cell proliferation.
Importin α 1 has been regarded as a novel cancer marker in several cancers, because of its overexpression in some tumour tissues that correlates with tumour progression 14,15,18,33 . Hepatocellular carcinoma is one of the cancers that show an aberrant expression of importin α 1 14 . Decreased proliferation of hepatocellular carcinoma cell lines after knockdown of importin α 1 has been shown in a previous report 14 . Our present study showed that importin α 1 is detected at relatively higher levels at the cell surface of at least three hepatocellular carcinoma cell lines (HepG2, Hep3B, and HLE) than that of other types of cancer cell lines. Moreover, we found that cell growth was suppressed when the HepG2 cells were treated with a monoclonal antibody specific for importin α 1. Therefore, these results suggested that importin α 1 localized to the cell surface is involved in the cell growth of at least some hepatocellular carcinomas.
Several studies have so far reported that some proteins that were originally identified as having a cytoplasmic or nuclear localization, such as calreticulin (endoplasmic reticulum, ER), GRP78/BiP (ER), hnRNP-K (nucleus), and nucleolin (nucleus) 19,[34][35][36][37][38][39] , can be expressed on the cell surface. Furthermore, it has been shown that most of these are translocated to the cell surface in response to cellular stresses, such as ER stress (GRP78/BiP) and cold stimulation (hnRNP-K). On the other hand, we showed that importin α 1 was expressed on the cell surface of some cancer cells under physiological conditions, similar to nucleolin 39 . Furthermore, our data revealed that although the total protein levels of importin α 1 in AGS cells was comparable to those in HCT116 cells and HepG2 cells, importin α 1 was not detected on the cell surface of AGS cells. These results suggested a novel intracellular trafficking, by which soluble intracellular proteins are transported to the cell surface and released from within cells in a cell condition-and/or cell type-dependent manner.
It has recently been reported that importin α 1 is present in the sera of lung cancer patients 18 . Consistently, we also found that importin α 1 was detected in culture media of several cancer cell lines. Furthermore, we found that exogenously added importin α 1 was associated with the cell surface and bound to HS on the cell surface. These results suggested that the released importin α 1 binds to the cell surface of adjacent cells via HS in tumour tissues.
It is known that FGFs can regulate cellular functions through an evolutionarily conserved signalling module that functions in both invertebrates and vertebrates 40 . The FGF family proteins, including the prototype members FGF1 and FGF2, are potent regulators of cell proliferation, differentiation, migration, and survival. On the cell surface, FGFs bind to two types of receptors; the high-affinity tyrosine kinase FGF receptors (FGFRs) and the low-affinity HS proteoglycan (HSPG). Activated FGFR initiates downstream signalling cascades, such as the pathways involving Ras/MAPK, phosphoinisitide 3-kinase (PI3K)/AKT, and phosholipase Cγ /PKC. The ERK signalling cascade is reported to control the proliferation of multiple cell types in response to growth factor treatment 30 . On the other hand, HSPG serves to recruit FGFs to FGFRs, and to stabilize the FGF− FGFR axis to maintain their downstream signalling 21 . In this study, we showed that importin α 1 interacts with both FGFs and HS. Furthermore, we found that the addition of importin α 1 recombinant proteins to culture cells enhanced the phosphorylation of ERK1/2 at the initial step of FGF signalling after FGF1 stimulation. Taken together, we propose the following model. (1) Importin α 1 is released from within cells and the released importin α 1 binds to the cell surface via HS. (2) The cell surface importin α 1 helps to recruit FGF, to accelerate FGF signalling at the initial step. Thus, it is biologically important to know how importin α 1 is transported to the cell surface and released from within cells.
It is also known that FGFs have cNLSs and accumulate in the nucleus in an importin α /β -dependent manner. However, it is still unknown how extracellular FGFs traverse the cell membrane to enter the cell. Thus, it is interesting to know whether cell surface importin α 1 is also involved in the internalization of extracellular FGFs. Further studies will be required to address these questions.
Cell surface proteins specific to cancer cells and soluble proteins in serum that promote tumorigenesis have been considered as potential targets for the treatment of cancers. Therapies targeted at the extracellular importin α 1 (both the released form and the cell surface-localized form), as well as those that are targeted at released importin α 1-mediated induction of FGF1 signalling, are expected to perturb the proliferation of importin α 1-releasing cancer cells, thereby providing a novel option for the treatment of cell surface importin α 1-related cancers. . pGEX-6P2-3 × Flag-importin α 1 (human) was obtained as described previously 8 . pGEX-2T constructs (GST-GFP, GST-NLS-GFP, and GST-IBB-GFP) were obtained as described control), and subjected to immunoblot analysis. (g) Starved HCT116 cells were incubated in FGF1 (20 ng/ml) in the presence or absence of recombinant importin α 1 (20 ng/ml) with anti-importin α 1 mAb (25 or 250 ng/ml) or control mouse IgGs for 10 min, and subjected to immunoblot analysis with indicated antibodies. (h,i) Cell proliferation of HCT116 cells (h) and AGS cells (i) were measured for the indicated time courses after treatment without or with normal mouse IgGs as isotype control or anti-importin α 1 mAb (250 ng/ml), in triplicate for each condition, using Cell Counting Reagent. Data are means ± SD from three independent experiments. *P < 0.005.
Pull-down assay. The proteins were added to transport buffer (20 mM HEPES/NaOH, pH 7.4, 110 mM potassium acetate, 2 mM magnesium acetate, 5 mM sodium acetate, 0.5 mM EGTA/NaOH, 2 mM DTT) in the presence or absence of anti-importin α 1 mAb, and mixed with glutathione-sepharose 4B beads (GSH-beads, GE Healthcare) or heparin-sepharose beads (Sigma). The mixtures were incubated at 4 °C for 1 h and then washed with transport buffer. Bound proteins were eluted with sample buffer for sodium dodecylsulfate-polyacrylamide gel electrophoresis (SDS-PAGE).
Sandwich enzyme-linked immunosorbent assay (ELISA). ELISA was performed using MaxiSorp plate (Thermo Fisher Scientific) coated with 0.5 μ g/well of recombinant proteins. Candidate proteins or conditioned medium were incubated with plate-bound each protein for 1 h. The antigen-antibody complexes were detected with 1:5,000-diluted HRP-conjugated second antibodies with 3, 3′ , 5, 5′ -tetramethylbenzidine (TMB; Dako Cytomation, Glostrup, Denmark) as the substrate. OD was read at 450 nm using a Bio-Rad Microplate Reader Model 680 (Bio-Rad Laboratories, Hercules, CA, USA). Immunofluorescence analysis. Cultured cells grown on glass coverslides were fixed in 4% paraformaldehyde for 15 min at room temperature; then, cells were permeabilized with PBS containing 0.1% Triton-X100 and 1 mg/mL bovine serum albumin (BSA) for 5 min. The cells were washed three times with PBS before incubation at 4 °C overnight with anti-importin α 1 mouse mAb (1:200 dilution; BD Bioscience). After washing three times with PBS, the cells were incubated at room temperature for 30 min with the appropriate Alexa Fluor 488-, or 647-conjugated secondary antibodies and stained with Hoechst 33342 (Thermo Fisher Scientific) for detection of nuclei. Cells were observed by confocal fluorescence microscopy (TCS SP8, Leica, Mannheim, Germany).
Statistical analysis. Data are presented means ± SD and were assessed for statistical significance using the unpaired Student's t test.