Importin-mediated Nuclear Translocation of Galectin-3*

Galectin-3 (Gal-3), a member of a β-galactoside-binding protein family, is involved in RNA processing and cell cycle regulation through activation of transcription factors when translocated to the nucleus. We have previously shown that Gal-3 can import into the nucleus through at least two pathways; via passive diffusion and/or active transport (Nakahara, S., Oka, N., Wang, Y., Hogan, V., Inohara, H, and Raz, A. (2006) Cancer Res. 66, 9995-10006). Here, we investigated the process mediated by the active nuclear transport of Gal-3 and have identified a nuclear localization signal (NLS)-like motif in its protein sequence, 223HRVKKL228, that resembles p53 and c-Myc NLSs (378SRHKKL383, 322AKRVKL327), respectively. Moreover, trimers of enhanced green fluorescence protein (3×GFP) fused with this NLS-like sequence, which is too large to passively diffuse through the nuclear pores, accumulated in the cell nuclei. To gain insights into this newly identified nuclear import mechanism, the interaction between Gal-3 and importins (importins α and β) that carry the NLS harboring nuclear proteins into the nucleus, was investigated. Pull-down assays and bimolecular fluorescence complementation (BiFC) analysis revealed that wild-type Gal-3, but not mutant Gal-3 (R224A), binds to importin-α. Down-regulation of importin-β by RNA interference (RNAi) efficiently abrogates its nuclear accumulation. Furthermore, we provide evidence that impaired nuclear translocation of mutant Gal-3 protein (R224A) results in accelerated degradation compared with the wild-type protein. Thus, these results suggest that Gal-3 is translocated to the nucleus, in part, via the importin-α/β route and that Arg224 amino acid residue of human Gal-3 is essential for its active nuclear translocation and its molecular stability.

This protein can be found not only in the cytoplasm, but also in the nucleus and even in the extracellular spaces (4,5). It shuttles between the cytoplasm and the nucleus, where it undergoes post-translational modification, i.e. phosphorylation of a serine residue by the protein kinase; casein kinase 1 (CK1) that signals its export to the cytoplasm and serves as a molecular switch for the sugar binding ability (6 -9). The nuclear localization of Gal-3 is probably associated with normal cell proliferation because it is a required factor in the splicing of pre-mRNA (10). In cancer, Gal-3 plays a significant role in tumor progression (11)(12)(13)(14)(15)(16). It exerts a growth-promoting effect and cell cycle regulation through induction of cyclin D1 and c-Myc in the nuclei of human breast epithelial cells (11)(12)(13). In particular, up-regulation of cyclin D1 expression was induced through enhancement or stabilization of the nuclear protein-DNA complex formation at the cAMP-responsive element of the cyclin D1 promoter (14). In the nuclei of papillary thyroid cancer cells, Gal-3 interacts with the thyroid-specific TTF-1 and up-regulates its transcriptional activity, contributing to proliferation (15). In prostate cancer cells, nuclear Gal-3 suppressed malignancy whereas cytoplasmic Gal-3 promoted tumorigenicity through a mechanism yet to be determined (16). In vivo, the cytoplasmic versus nuclear expression of Gal-3 is associated with tumor invasion and metastasis (17)(18)(19)(20)(21)(22). Loss of nuclear Gal-3 expression was reported in colon and prostate carcinomas (17,18). Similarly, the levels of nuclear Gal-3 were markedly decreased during the progression from normal to cancerous states in tongue carcinomas (19). In lung carcinoma, the nuclear Gal-3 is a predictive factor of recurrence and/or worse clinical outcome (20,21). In patients with esophageal squamous cell carcinoma, elevated expression of Gal-3 in the nuclei was suggested to be an important pathological parameter related to histological differentiation and vascular invasion (22).
Analysis of in vitro nuclear import assay using digitonin-permeabilized cells revealed that there are at least two pathways for the nuclear import of Gal-3, passive diffusion and active transport dependent on its N-terminal region (23). However, the possible role of the carbohydrate recognition domain (CRD) in Gal-3 nuclear translocation process was difficult to determine using this system because it bound to the cell surface glycoproteins and/or glycolipids through its ␤-galactoside binding site, resulting in inhibition of transport (23). Thus, the complete mechanism of Gal-3 nuclear import is not fully understood and a different approach avoiding the use of digitonin-permeabilized cells is required to study and establish the critical role of CRD for Gal-3 nuclear import.
Here, we have identified an NLS-like sequence, 223 HRVKKL 228 , in the C-terminal region of the human Gal-3 that is similar to p53 and c-Myc NLSs (24 -26), and studied its function(s). NLS-containing proteins, like p53, are transported into the nucleus by the importin (karyopherin) ␣/␤ complex (26 -28). Importin-␣ is the receptor subunit that recognizes the NLS, a cluster of basic amino acid residues that complex with importin-␤ for passage through the nuclear pore (29,30). The presence of several isoforms of importin-␣ proteins in mammals recognizing variants of the classic NLS was reported (31), suggesting that the nuclear import process might be different for each protein depending in part on specifically regulated biological functions. This study also shows that deletion of the C-terminal region of Gal-3 protein without the NLS-like motif  results in complete impairment of nuclear accumulation, whereas a different C-terminal deletion protein that includes this motif  can be accumulated in the nucleus. In addition, substitution of Arg to Ala at position 224 (R224A) of the human Gal-3 efficiently abolished nuclear localization. To conclude, the studies shown here revealed that Gal-3 directly binds to importin-␣ proteins and complexes with the importin-␣/␤ complex in vivo. Moreover, the failure to translocate from the cytoplasm to the nucleus resulted in a rapid degradation of Gal-3.

EXPERIMENTAL PROCEDURES
Cell Culture and Transfection-COS-7, HeLa and HT1080 cells were purchased from American Type Culture Collection (ATCC, Manassas, VA). These cells were cultured in Dulbecco's modified Eagle's medium (Invitrogen, Carlsbad, CA) supplemented with 5% fetal bovine serum (Atlanta Biologicals, Lawrenceville, GA) and antibiotics (50 international units/ml penicillin and 50 g/ml streptomycin) (Cellgro, Herndon, VA) under 5% CO 2 at 37°C. Transfection into COS-7 or HT1080 cells was performed using Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol. For siRNA transfection to down-regulate the expression level of human importin-␤1, karyopherin ␤1 siRNA (h) (sc-35736, Santa Cruz Biotechnology, Santa Cruz, CA) was purchased and used. Selection was performed via the addition of 1000 g/ml G418 (Cellgro) to establish stable transfectants. Cell culture with addition of the reagents was performed as follows. In the case of leptomycin B treatment, at 24 h after transfections, COS-7 cells were treated with 30 ng/ml of leptomycin B (Sigma) for 4 h at 37°C. In the case of cycloheximide (CHX) treatment, at 24 h after transfection COS-7 cells were treated with 100 ng/ml of CHX (Sigma) for the time indicated.
GFP Expression and Immunofluorescence Microscopy-Cells on Lab-Tek II chamber slide (8-well) (Nalge Nunc International, Naperville, IL) were fixed with precooled (Ϫ30°C) methanol for 30 min at 4°C or 4% paraformaldehyde for 20 min at 25°C. In the case of GFP expression, cells were washed with PBS and counterstained with 5 g/ml Hoechst 33258 (Molecular Probes, Eugene, OR), then mounted with Crystal/Mount reagent (Biomeda, Foster City, CA). In the other cases, cells were permeabilized with 0.5% Triton X-100/PBS for 5 min after fixation. They were blocked with 10% goat serum in PBS for 30 min at room temperature and then incubated with anti-Myc antibody (Invitrogen) (1:400 dilution in PBS) for Gal-3/C detec-tion, anti-Xpress antibody (Invitrogen) (1:400 dilution in PBS) for Gal-3/N detection at room temperature for 1 h, respectively. After washes with PBS, cells were incubated for 1 h with fluorescein isothiocyanate (FITC)-conjugated anti-mouse IgG (Zymed Laboratories Inc., San Francisco, CA) for Myc and Xpress diluted 1:100. After washes with PBS, slides were then mounted with a drop of mounting reagent. Fluorescent signal was recorded with a Zeiss Laser Scanning Microscope 310 or an Olympus BX40 microscope.
In Vitro Binding Assay-Recombinant GST-fused importin proteins (importin-␣1, ␣3, ␣5, ␤1) were expressed and purified as described previously (35). Recombinant Gal-3 proteins (wild type, R224A mutant) were expressed and purified as described previously (23,32). Lysates from bacteria expressing GST (20 g) or GST-importins (each 15 g) were incubated with GSH-Sepharose 4B (Amersham Biosciences) and resuspended as 1:1 slurry with transport buffer (23). The beads were mixed with 1 g of Gal-3 and dissolved in final 150 l of transport buffer. After incubation for 1.5 h at room temperature, the beads were washed five times with transport buffer and subject to SDS-PAGE followed by Western blot analysis.
Lactose Precipitation-Lactose-conjugated agarose was used to check lactose binding capacity of wild-type Gal-3 and its mutants (10). Briefly, whole cell extracts (200 g) from each transfected cell line were mixed with 15 l of 1:1 lactose-conjugated agarose slurry (Sigma) followed by agitation at 4°C for 1.5 h. At the same time, 10% of whole cell extracts was saved as an input. After three washes with 0.5% Triton X-100 in PBS, samples were subject to SDS-PAGE followed by Western blot analysis.
In Vitro Degradation Assay-Recombinant GFP-fused Gal-3 and mutant Gal-3 (R224A) proteins were expressed and purified as described above. Each GFP-fused Gal-3 (25 ng) was mixed with 10 g of whole cell lysates (protease inhibitor-free) from COS-7 cells and dissolved in final 15 l of PBS. After incubation at 37°C for indicated time, samples were subject to SDS-PAGE followed by Western blot analysis.

RNA Isolation and Reverse Transcription (RT)-PCR-Total
RNAs from transfected COS-7 cells were isolated with TRIzol reagent (Invitrogen) according to the manufacturer's instructions. RT-PCR analysis was performed using AccessQuick RT-PCR system (Promega, Madison, WI) with 1 g of RNA from each cell as a template. For detection of GFP mRNA expression, 40 cycles (95°C for 30 s, 55°C for 30 s, 72°C for 1 min) were performed with the following primers: 5Ј primer, 5Ј-CGGGA-TCCATCGCCACCATGGT-3Ј and 3Ј primer, 5Ј-CGGGATC-CCTTGTACAGCTCGTC-3Ј. For detection of ␤-actin mRNA as an internal control, 25 cycles (95°C for 30 s, 60°C for 30 s, 72°C for 1 min) were performed with the following primers: 5Ј primer, 5Ј-TGACGGGGTCACCCACACTGTGCCCAT-3Ј and 3Ј primer, 5Ј-CTAGAAGCATTTGCGGTGGACGATG-GAGGG-3Ј. After RT-PCR, samples were electrophoresed on 1.0% agarose gel containing SYBR Gold nucleic acid gel stain reagent (Molecular Probes), and the intensity of each band was detected with a Kodak Digital Science Image System.

Subcellular Localization of NLS-like Sequence Deleted or
Point-mutated Gal-3 Proteins-To address the possibility that the 223 HRVKKL 228 sequence of Gal-3 is functionally homologous with p53 and c-Myc NLSs (Refs. 24 and 25, Fig. 1A) and is involved in Gal-3 nuclear import, we initially constructed several GFP-fused Gal-3 expression vectors, transfected them into cells, and analyzed subcellular localization of those products with a fluorescence microscope. We constructed fusion proteins whereby wild-type Gal-3 fused with GFP at its C terminus (WT/GFP) and its point mutants in the target sequence by sitedirected mutagenesis where we have substituted Arg 224 , Lys 226 , and Lys 227 with Ala (R224A, K226A, K227A), respectively, and in addition constructed two C-terminal-deleted mutants proteins with or without the above sequence (1-229/GFP and 1-222/GFP), respectively (Fig. 1B). To confirm the correct expression of the deletion and point mutants of EGFP-fused Gal-3 as well as 2ϫ EGFP (2ϫ GFP) proteins, Western blotting of the transfected COS-7 cell lysates was performed using antibodies to GFP and Gal-3 (TIB166) (Fig. 1C). All plasmids expressed proteins of the expected molecular weights in Western blotting of either GFP (Fig. 1C, green), and/or Gal-3 (Fig. 1C, red) indicated the correct expression of Gal-3 polypeptides (Fig.  1C, yellow). Thereafter, each plasmid was transfected into COS-7 cells and cultured for 24 h at 37°C. The cells were then Galectin-3 and Nuclear Import DECEMBER 22, 2006 • VOLUME 281 • NUMBER 51 fixed and visualized under a fluorescence microscope. The representative microscopic fields are depicted in Fig. 1D. EGFP expression of 1-222/GFP, which does not include the above NLS-like Gal-3 sequence, was outside the nucleus (Fig. 1D, panels a and aЈ). However, 1-229/GFP Gal-3-protein, which contains that sequence, was found to be localized in the nuclear compartment (Fig. 1D, panels b and bЈ), implying that the consecutive amino acid residues (223-229) of Gal-3 partially control its cellular distribution. Moreover, high expression of R224A mutant protein was located predominantly outside the nucleus (Fig. 1D, panels d and dЈ), whereas the GFP expression patterns of K226A and K227A protein mutants were diffusely spread throughout the cytoplasm (Fig. 1D, panels e, eЈ, f, and f Ј). Whereas, the expression of wild-type Gal-3 protein was found to be randomly distributed in both the nucleus and cytoplasm (Fig. 1D, panels c and cЈ). The fluorescence in the cells transfected with 2ϫ GFP showed the predominant cytoplasmic localization of EGFP (Fig. 1D, panels g and gЈ) as described previously (23). However, we observed that the expression pattern of each protein had variety and diversity, therefore, to evaluate and determine the cellular distribution of each protein, we categorized their patterns into three major groups (mostly localized in the nucleus, N Ͼ C; equally localized in the cell, n ϭ C; mostly spotted around the nucleus (spot C) and counted over 300 cells that expressed GFP in each plasmid-transfected cells. As shown in Fig. 1E, the nuclear localization ratio of 1-229/GFP was ϳ10%, whereas the ratio of 1-222/GFP was ϳ0%. In addition, the spotted pattern ratio of 1-222/GFP highly increased compared with R224A/ GFP (Fig. 1D, 1-222, R224A, shadowed bar, p Ͻ 0.0001). There were no significant differences of subcellular distribution among WT/GFP, K226A/GFP, and K227A/Gal-3 proteins (Fig. 1E, WT, K226A, K227A, p Ͼ 0.05).
Because GFP fusion proteins might change the original biological function of Gal-3 because of its large molecular mass, and because it was reported that the region fused with tags could affect the localization pattern of several proteins (37), we constructed both N-and C-terminal His-tagged Gal-3 expression vectors. In addition to the wild type, one C-terminal-deleted Gal-3-(1-222) and three point mutant proteins (described above), we constructed a mammalian expression vector for the deleted N-terminal region Gal-3 (Gal-3/N-(63-250)) ( Fig. 2A). To determine how these exogenous His-tagged Gal-3 products distribute in the cell, we first performed immunofluorescence analysis after transfection of each plasmid into COS-7 cells cultured for 24 h, and visualized them under the microscope. The representative immunostaining data using anti-Myc antibody of C-terminal His-tagged Gal-3 (Gal-3/C) are depicted in Fig. 2B. As shown in Fig. 2B, wild-type Gal-3 (Gal-3/C(WT)) was evenly distributed throughout the cell, slightly different from the pattern of GFP-fused wild-type Gal-3 distribution (Fig. 1D, panels c and cЈ), probably because of the difference between direct observation of GFP expression and indirect immunofluorescent methodology. The staining patterns of R224A mutant and C-terminal-deleted Gal-3-(1-222) proteins were similar to the results of GFP-fused protein expression, respectively (Fig. 2B, 224, 1-222). The staining patterns of both K226A and K227A mutants were almost similar to that of wild type (data not shown). We confirmed there was no staining with control normal mouse IgG (Fig. 2B, CTRL). The representative immunostaining patterns using anti-Xpress antibody of N-terminal His-tagged Gal-3 proteins (Gal-3/N) were shown in Fig. 2C. The patterns of wild-type Gal-3 and three point mutants (R224A, K226A, K227A) were almost similar to those of C-terminal His-tagged Gal-3 (Fig. 2C, WT, 224, the data of 226 and 227 are not shown), indicating that any tags of the terminal regions do not alter the subcellular localization of Gal-3. As a new insight, N-terminal-deleted Gal-3 (Gal-3/N-(63-250)) could also exist in the nucleus (Fig. 2C, 63-250). We also confirmed there was no staining with normal mouse IgG (Fig. 2C, CTRL). Next, the observed distributions of each Gal-3 protein were confirmed by Western blot analysis using anti-His antibodies after subcellular fractionation of each plasmid transfected COS-7 cell. The results of the Western blot analyses in subcellular localization strongly support the lack of nuclear localization of Gal-3/C(R224A), Gal-3/C-(1-222), and Gal-3/N(R224A) proteins, respectively, whereas the other mutant proteins including wild-type Gal-3 were localized in the nucleus (Fig. 2, D and E, upper  panel). To show the successful subcellular fractionation, we reprobed the same membranes with anti-␤tubulin antibodies for cytoplasmic fraction and anti-histone H1 antibodies for nuclear fraction (Fig. 2, D and E, middle and lower panel). In addition, we performed these experiments using additional cell lines (BT-549, human breast carcinoma and HEK 293, human fetal kidney cell), and the results were found to be similar (data not shown). Taken together, Arg 224 residue in the NLS-like motif of Gal-3 is essential for nuclear localization of Gal-3.
Because the possible role of a sugar binding activity for Gal-3 nuclear import was questioned (23), we examined whether the nuclear translocation of mutated Gal-3 protein (R224A) lost its sugar binding capacity. Thus, we examined the sugar binding of each mutant Gal-3 protein by lactose-agarose precipitation (LP) assay using the endogenous expressed Gal-3 protein present in cell lysates of COS-7 cells transfected with His-tagged Gal-3 plasmids. The result indicated that Gal-3(R224A) also maintained its sugar binding activity similar to the wild-type Gal-3 proteins (Fig. 3A, upper panel). To validate the efficiency of this methodology, the same membranes were reprobed with anti-Gal-3 antibodies, positive control (Fig. 3A, middle panel) or anti-␤-actin antibodies as a negative control of lactose binding (Fig. 3A, lower panel).
Next, we examined whether Gal-3(R224A) protein was predominantly distributed outside the nucleus because of the lack of a nuclear retention. Because nuclear localization of Gal-3 can be attributed to a nuclear export signal (NES) located at the C-terminal region, 228 LNEISKLGI 236 in human Gal-3 (39), and the Arg 224 residue is located near this NES, one might argue that the cytoplasmic localization of Gal-3(R224A) is caused by the lack of nuclear retentive capacity by enhancement of its nuclear export. Leptomycin B (LMB) is known to inhibit the nuclear export of Gal-3 through NES (23, 39); thus, we treated GFP-fused Gal-3-transfected cells with LMB and followed its

. Subcellular localization of His-tagged Gal-3. A, schema of Gal-3 fused with Myc-His in its C-terminal region (I, Gal-3/C) and with His-Xpress in its N-terminal region (II, Gal-3/N). B, immunofluorescent staining (ϫ630) with anti-Myc antibody was performed to detect Gal-3/C expression after transfection to COS-7 cells. No staining was observed in vector only transfected cells (CTRL). C, immunofluorescent staining (ϫ630) with anti-Xpress antibody was performed to detect Gal-3/N after transfection. Parental cells were used as a control stain (CTRL). D, subcellular localization of various Gal-3/C was investigated after transfection into COS-7 cells.
Cytoplasmic protein (Cy, each 30 g) and nuclear protein (N, each 15 g) were subjected to SDS-PAGE (12.5%) followed by Western blotting with anti-His for detection of Gal-3/C. The same membrane was reprobed with anti-␤-tubulin, anti-histone H1 for detection of cytoplasmic and nuclear fraction, respectively. As a control, vector only was expressed and analyzed (CTRL). An arrow indicates nonspecific band detected in nuclear fraction of the cell. E, subcellular localization of various Gal-3/N was investigated after transfection into COS-7 cells. Cytoplasmic protein (Cy, each 30 g) and nuclear protein (N, each 20 g) were subjected to SDS-PAGE (12.5%) followed by Western blotting with anti-His for detection of Gal-3/N. As a control, vector only was expressed and analyzed (CTRL). An arrow indicates nonspecific band that is mainly detected in the nuclear fraction of the cell. DECEMBER 22, 2006 • VOLUME 281 • NUMBER 51 effect on the localization of each Gal-3 protein. The representative data are depicted in Fig. 3B. LMB significantly inhibited the nuclear export of the WT/GFP protein (Fig.  3, A and B, WT, black bar, p ϭ 0.01). Under the same experimental condition, the localization of R224A/ GFP was not changed because of the LMB treatment (Fig. 3, B and C, 224, p Ͼ 0.05), suggesting that R224A/ GFP protein cannot migrate into the nucleus, similarly, the localization of 1-222/GFP protein in which NES was deleted, was also not changed (Fig. 3, B and C, 1-222, p Ͼ 0.05). Taken together, these results suggested that Gal-3(R224A) protein lacks nuclear import activity independent of its ␤-galactoside binding.

Galectin-3 and Nuclear Import
NLS-like Sequence of Gal-3-Additional studies are required to evaluate the essential role of the NLSlike sequence ( 223 HRVKKL 228 ) as well as the Arg 224 residue for Gal-3 nuclear import. Nuclear localization was observed for 1-229/GFP, K226A/GFP, and K227A/GFP, but not for 1-222/GFP and R224A/GFP (Fig. 1D), therefore, we focused on further characterizing the putative nuclear localization domain of amino acid residues 222 NHRVKKLNE 230 (aa 222-230) and the significance of its Arg residue. To confirm that aa 222-230 can mediate the nuclear import of Gal-3 and the Arg 224 residue can play a critical role in that process, plasmids encoding aa 222-230 and its Arg residue mutated peptides fused with trimer of EGFP (3ϫ EGFP), respectively, to avoid the passive diffusion into the nucleus were constructed (3ϫ GFP/ RVKKL, 3ϫ GFP/AVKKL, Fig. 4A), and the subcellular localization of those fusion proteins as well as 3ϫ GFP was examined 48 h after transfection in COS-7 (Fig. 4B) and BT549 (Fig. 4C) cells by fluorescence microscopy. As expected from the deletion mutant analysis, 3ϫ GFP/RVKKL was predominantly directed into the nucleus (Fig. 4, B and C, panel aЉ). On the other hand, the replacement of Arg to Ala in that sequence (3ϫ GFP/ AVKKL) or trimer of EGFP (3ϫ EGFP) abrogated the nuclear localization in both COS-7 and BT549 cells (Fig. 4, B and C,  panels bЉ and cЉ). Collectively, these results indicate that the 222 NHRVKKLNE 230 sequence of human Gal-3 constitutes a functional NLS, one in which Arg 224 in particular plays a critical role.
Gal-3 Importins Interaction-Because nuclear import of Gal-3 has a sequence motif similar to that of p53 and c-Myc NLSs (Fig. 1A) and this 222 NHRVKKLNE 230 sequence should be a functional NLS (Fig. 4), we examined whether Gal-3 could directly bind to importin-␣ proteins, e.g. binding partners of NLScontaining proteins. And because importin-␣ proteins have been categorized into three major groups according to the structure (31, 40), we first performed GST pull-down assay using three importin-␣ proteins (␣1, ␣3, ␣5). As shown in Fig.  5A, wild type Gal-3 protein directly binds to all importin-␣ proteins, whereas failing to bind to importin-␤ or GST alone (Fig. 5A, upper  panel). On the other hand, mutant Gal-3 (R224A) protein could not efficiently bind to any of the importin-␣ proteins (Fig. 5A, middle  panel), suggesting that the Arg 224 residue is essential for Gal-3 binding to importin-␣ proteins and continuous basic amino acid residues of that region function as an NLS. The proper expression of each recombinant protein was confirmed with Western blotting of GST (Fig. 5A,  lower panel).
To further investigate the interaction of Gal-3-importins in vivo, we tested the complex formation of these proteins using BiFC analysis. The BiFC approach is based on the formation of a fluorescent complex when two fragments of a fluorescent protein are brought together by an interaction between proteins fused to the fragments (33,34). Complementary fragments of YFP (YN or YC) were fused to the N-terminal region of importin-␤, importin-␣1, Gal-3(WT), and Gal-3(R224A), respectively (Fig. 5B). A selected pair of these plasmids was co-transfected into COS-7 cells, and we confirmed the expression of the two proteins by Western blotting using polyclonal anti-GFP antibody (Fig. 5C). Two distinct protein bands in each lane indicate the successful double protein expression in each transfection.  microscopy (ϫ400). Panels a-c, fluorescent signals of the cells expressing EGFP-labeled proteins; panels aЈ-cЈ, nuclear DNA stained with Hoechst 33258 of the cells in the same fields as in panels a-c, respectively; panels aЉ-cЉ, combined images of panels a-c and aЈ-cЈ, respectively. DECEMBER 22, 2006 • VOLUME 281 • NUMBER 51
Importin-␤ Contributes to the Nuclear Translocation of Gal-3-Next, we examined whether Gal-3 is carried by the importin-␣/␤ complex into the nucleus, and used the RNA interference (RNAi) approach to inhibit the expression of importin-␤ to evaluate its effect on the cellular localization of Gal-3/GFP. The cellular localization of SV40NLS fused with 2ϫ EGFP (NLSϩ2 ϫ GFP, Ref. 23), which is a characterized carrier protein of the importin-␣/␤ complex, is used as a control. Each plasmid and a siRNA duplex targeted to human importin-␤ were co-transfected to HT1080 cells. At 48 h after transfection, the cells were harvested, and the expression level of endogenous importin-␤ was analyzed by Western blotting. As shown in Fig. 6A, this treatment specifically reduced the expression level of importin-␤ protein as compared with ␤-tubulin. The transfected cells with importin-␤ siRNA and Gal-3/GFP showed distinct perinuclear localization with several punctuate cytoplasmic staining patterns (Fig. 6B, panels b and bЈ, Fig.  6C), whereas the non-transfected cells with siRNA showed the entire cellular distribution of Gal-3/GFP (Fig. 6B, panels a and aЉ, Fig. 6C). Similarly, NLSϩ2 ϫ GFP in the cell transfected with siRNA was predominantly localized in the cytoplasmic space (Fig. 6B, panels d and  dЉ, Fig. 6C). These results demonstrate the functional and the specific interaction of importin-␤ is required for Gal-3 nuclear translocation.
Outcome of the Lack of Nuclear Localization Ability of Deleted and Mutated Gal-3 Proteins-Because the mutant Gal-3(R224A) protein did not show typical cytoplasmic distribution, we hypothesized that it may be rapidly degraded. Thus, we have used CHX, a reagent that can inhibit de novo protein synthesis (41), to establish whether the rate at which Gal-3 protein level declines with this CHX treatment is a gauge of protein instability (42). As shown in Fig. 7A, we examined the stability of wild type of Gal-3 (Gal-3/C(WT)) and mutant Gal-3 (Gal-3/ C(R224A)) at different time points after CHX addition to the transfected COS-7 cells. It appeared that Gal-3/C(R224A) protein, which cannot enter the nucleus, was rapidly degraded (Fig.  7A, 224), whereas wild-type Gal-3 was very stable over the 6 h of the assay (Fig. 7A, WT). We also examined the stability of Gal-3/C-(1-222), and the results were similar to that of Gal-3/ C(R224A) (data not shown). In addition, we also monitored the amount of Gal-3 proteins secreted into the conditioned medium after CHX treatment and found that Gal-3/C(WT), but not Gal-3/C(R224A), was secreted out of the cell (data not shown). Next, recombinant proteins were introduced to evaluate the stability of mutant Gal-3 (R224A) protein using an in vitro degradation assay. As shown in Fig.  7B, both recombinant Gal-3 proteins (Gal-3/GFP(WT) and Gal-3/ GFP(R224A)) were degraded in ϳ6 h when incubated with COS-7 cell lysates, with no detectable difference in the degradation rate between them. These results suggest that lack of nuclear translocation of Gal-3 (1-222, R224A) protein results in its in vivo rapid degradation. Thus, we tested whether the protein expression level of Gal-3(R224A) would decrease after prolonged culture. Cell selection was performed after 3 weeks to establish the stable transfectants of Gal-3/ GFP(WT) and Gal-3/GFP(R224A). Three different transfectants were established, e.g. one Gal-3/GFP(WT) transfectant (Fig. 7C, WT) and two Gal-3/GFP(R224A) transfectants (Fig. 7C, 224a, and b), and performed Western blotting and RT-PCR for detection of GFP. Gal-3/GFP(WT), but not Gal-3/GFP(R224A), could be detected at the protein level (Fig.  7C, upper panel). In contrast, Gal-3/ GFP(R224A) could be clearly detected at the mRNA level as well as Gal-3/GFP(WT) (Fig. 7C, lower  panel). We confirmed the control of no protein and no mRNA expressions of GFP in parental cells (Fig.  7C, CTRL) and detected ␤-actin expression as an internal control in all cells (Fig. 7C, ␤-actin). In addition, we observed these transfectants under the microscope and again the expression of Gal-3/ GFP(R224A) protein was not detected (data not shown). Taken together, these results suggest that Arg 224 mutation as well as C-terminal deletion affect the Gal-3 protein stability and expression.

DISCUSSION
Here, we report that a consecutive basic amino acid residues in the C-terminal domain of the human Gal-3, i.e. 223 HRVKKL 228 that resembles a sequence of one of the p53 NLSs (p53 has three NLSs in its sequence, Refs. 24 and 26) or FIGURE 6. Effect of importin-␤ siRNA on Gal-3 nuclear accumulation. Total cell lysates (15 g) were prepared from HT1080 cell 48 h after specific importin-␤ siRNA transfection and subjected to Western blotting of importin-␤ and ␤-tubulin using the same membrane (A). Images of cells (B, ϫ400) and relative proportion of cells (C) expressing Gal-3/GFP or NLSϩx2GFP in control cultures or after importin-␤ siRNA transfection. Panels a-d, fluorescent signals of the cells expressing EGFP-labeled proteins; panels aЈ-dЈ, nuclear DNA stained with Hoechst 33258 of the cells in the same fields as in panels a-d, respectively; panels aЉ-dЉ, combined images of panels a-d and aЈ-dЈ, respectively. ((Ϫ), control culture; (ϩ), siRNA transfection)). DECEMBER 22, 2006 • VOLUME 281 • NUMBER 51 that of c-Myc NLS (25) (Fig. 1A), could function as its NLS. A single amino acid substitution of Arg 224 residue in this motif results in the protein retention in the cytoplasm, whereas both Lys 226 and Lys 227 residues are not essential for Gal-3 nuclear import despite consecutive basic amino acids (Fig. 1, D and E). This agrees with reports showing the significance of an Arg residue, but not Lys, in the nuclear import of protein (43)(44)(45). In particular, a single mutation of Arg 306 to Ala or even Lys in the major NLS of p53 abolished its nuclear import (42). Next, we examined the interaction between Gal-3 and importin-␣ proteins, key molecules essential for NLS-containing the protein nuclear import. Gal-3 (WT) could bind directly to impor-tin-␣ proteins, whereas mutant Gal-3 (R224A) protein could not (Fig. 5A). BiFC analysis using Gal-3(WT) and Gal-3(R224A) proteins supported these findings (Fig. 5D). In addition, downregulation of importin-␤ by siRNA treatment abrogated nuclear localization of Gal-3, suggesting that Gal-3 could be imported to the nucleus via the importin-␣/␤ complex.

Galectin-3 and Nuclear Import
However, a question arises regarding the biological significance of Gal-3 nuclear import by means of an importin-␣/␤ complex because we previously reported that Gal-3 migrated into the nucleus by two pathways; passive diffusion and active transport dependent on its N-terminal sequence (23). In this regard, it is important to point out that the localization patterns of 1-222/GFP, 1-229/GFP, and R224A/GFP were different in the transfected cells (Fig. 1E). The difference between 1-222/ GFP and 1-229/GFP subcellular distribution might show the dependence of Gal-3 nuclear import on the importin-␣/␤ complex (Fig. 1). Regarding the subcellular distribution pattern, wild-type Gal-3 is predominantly localized in a cytoplasmic compartment despite the presence of functional importin-␣/␤ complex. It is probable that Gal-3 nuclear migration depends in part on the affinity to importin-␣ proteins, which is likely to be weak (Fig. 5A). To support this, the localization of trimer of EGFP fused with Gal-3 NLS-like sequence did not show the typical nuclear accumulation (Fig. 4) as shown when SV40NLS harboring protein was expressed (23). In BiFC analysis, the number of cells expressing maturated YFP, i.e. YN/␤ ϩ YC/␣1 Ͼ Ͼ YN/WT ϩ YC/␣1 Ͼ YN/␤ ϩ YC/WT, may also suggest a low affinity of Gal-3 to importin-␣ proteins (data not shown). Moreover, Davidson et al. (46) implied that the static observations of the nuclear versus cytoplasmic localization of murine Gal-3 must reflect a balance among at least four key parameters, i.e. nuclear import, nuclear export, cytoplasmic anchorage, and binding and retention in the nucleus, so subcellular localization of Gal-3 should be regulated in a complicated manner in vivo.
Regarding nuclear export, there are also some similarities between Gal-3 and p53. Serine phosphorylation is the only known post-translational modification for Gal-3, and its phosphorylation seems to be related to changes in its subcellular distribution (4,39,47). In particular, phosphorylation of Gal-3 is required for its nuclear export in response to chemotherapeutic drugs (9). In the case of p53, two serine residues located within the N-terminal NES (two NESs have been found in p53) are phosphorylated following DNA damage (26,48), resulting in blockage of p53 nuclear export leading to its nuclear retention. Of note, ubiquitination can also regulate the nuclear-cytoplasmic shuttling of p53 (26,49). Ubiquitination of p53 protein in the nucleus signals its export from the nucleus followed by cytoplasmic degradation in the proteasome system. Here, we found that loss of nuclear migration ability results in a rapid cytoplasmic degradation of Gal-3 mutant proteins (1-222, R224A) (Fig. 7). One might think that this degradation is not caused by simple inhibition of nuclear import but induced by rapid protein misfolding and/or aggregation followed by trapping in certain cytoplasmic compartments. In this regard, we performed double staining with Mitotracker or Lysotracker (Molecular Probes) using Gal-3/GFP-transfected cells, but no mutant Gal-3 could be detected in either the mitochondria or lysosomes (data not shown). Also, we checked the rate of deg- FIGURE 7. Rapid degradation of Gal-3 that cannot enter the nucleus in a living cell. A, COS-7 cells transiently transfected with each Gal-3/C expression vector were treated with cycloheximide (CHX(ϩ)) or only Me 2 SO ((Ϫ)) and harvested at the indicated time. Each fraction was subjected to Western blotting of anti-His and anti-␤-actin antibodies. (WT; Gal-3/C(WT)-transfected cells, 224; Gal-3/C (224)-transfected cells)). B, recombinant GFP-fused Gal-3 (WT, 224) was incubated with whole cell extracts (10 g) from COS-7 cells. After incubation for indicated time, samples were subjected to Western blotting of GFP and ␤-actin. (WT, EGFP-fused wild-type Gal-3; 224, EGFP-fused mutant Gal-3 (R224A)). C, proteins and RNAs were extracted from COS-7 cells stably transfected with wild-type and mutant Gal-3/GFP. Equal amount of proteins (20 g) were subjected to Western blotting of anti-GFP and anti-␤actin antibodies (I). RT-PCR was performed for detection of GFP and ␤-actin mRNA expression, respectively, and each amplified cDNA was electrophoresed (II). (CTRL: parental COS-7 cells, WT; Gal-3/GFP(WT) transfected COS-7 cells, 224a, b; Gal-3/GFP (224) transfected COS-7 cells, and these two cell lines were independently established.) radation by addition of lactacystin and chloroquine, a proteasome and lysosome inhibitor, respectively, but there were no significant differences with these treatments (data not shown). Taken together, the detailed mechanism of mutant Gal-3 rapid degradation is still unclear, but we believe that active nuclear import and serine phosphorylation of Gal-3 are necessary for stability expression of Gal-3 protein.
In summary, we hypothesized that, like p53, the nuclear transport of Gal-3 plays a role in the regulation of its turnover in a living cell. The Gal-3 protein could be imported into the nucleus by either of passive diffusion, N-terminal region-dependent active transport, or importin-␣/␤ complex-dependent transport. This Gal-3 nuclear localization is dependent on the Arg 224 residue involved in the interaction with importin-␣/␤ proteins leading to nuclear import of Gal-3, a process likely to be essential for its protein stabilization.