Galectin-3 phosphorylation is required for its anti-apoptotic function and cell cycle arrest.

Galectin-3, a beta-galactoside-binding protein, is implicated in cell growth, adhesion, differentiation, and tumor progression by interactions with its ligands. Recent studies have revealed that galectin-3 suppresses apoptosis and anoikis that contribute to cell survival during metastatic cascades. Previously, it has been shown that human galectin-3 undergoes post-translational signaling modification of Ser(6) phosphorylation that acts as an "on/off" switch for its sugar-binding capability. We questioned whether galectin-3 phosphorylation is required for its anti-apoptotic function. Serine to alanine (S6A) and serine to glutamic acid (S6E) mutations were produced at the casein kinase I phosphorylation site in galectin-3. The cDNAs were transfected into a breast carcinoma cell line BT-549 that innately expresses no galectin-3. Metabolic labeling revealed that only wild type galectin-3 undergoes phosphorylation in vivo. Expression of Ser(6) mutants of galectin-3 failed to protect cells from cisplatin-induced cell death and poly(ADP-ribose) polymerase from degradation when compared with wild type galectin-3. The non-phosphorylated galectin-3 mutants failed to protect cells from anoikis with G(1) arrest when cells were cultured in suspension. In response to a loss of cell-substrate interactions, only cells expressing wild type galectin-3 down-regulated cyclin A expression and up-regulated cyclin D(1) and cyclin-dependent kinase inhibitors, i.e. p21(WAF1/CIP1) and p27(KIP1) expression levels. These results demonstrate that galectin-3 phosphorylation regulates its anti-apoptotic signaling activity.

acids that controls its cellular targeting; a repetitive collagenlike sequence rich in glycine, tyrosine, and proline, which serves as a substrate for matrix metalloproteinases; and a COOH-terminal domain of galectin-1 with a globular structure encompassing the carbohydrate-binding site (1-3). Gal-3 was shown to be involved in cell growth, cell adhesion, differentiation, and tumor progression through binding to complementary glycoconjugates (1, 4 -8).
Recent studies (9 -12) have revealed that Gal-3 inhibits apoptosis induced by anti-Fas antibody, staurosporine, chemotherapeutic reagent, tumor necrosis factor, radiation, and nitric oxide. Gal-3 also prevents cells from undergoing anoikis, a specific form of apoptosis induced by loss of cell-substrate interactions, by inducing G 1 arrest (13). Although Gal-3 does not belong to the Bcl-2 family, it contains the Asp-Trp-Gly-Arg amino acid anti-death sequence, which is a highly conserved sequence within the BH1 domain of the Bcl-2 family (9,10,14,15). Recent reports (16 -21) have shown that phosphorylation of Bcl-2 at Ser 70 appears to be critical for its anti-apoptotic function. However, the role of this phosphorylation is debatable due to conflicting reports demonstrating that such a phosphorylation may activate or inactivate Bcl-2 anti-apoptotic function (17)(18)(19)(20)(21). It was argued that these contradictory conclusions stemmed from results obtained from using different cell types and death signaling molecules (16 -23). Due to the functional anti-death mimicking of Bcl-2 and Gal-3 and the fact that both undergo post-translational modification of serine phosphorylation, we questioned the role of Gal-3 phosphorylation in its anti-apoptotic activity.
In 3T3 fibroblasts, quiescent cells express phosphorylated Gal-3 both in the cytoplasm and in the nucleus, whereas proliferating cells show an increased level of phosphorylated Gal-3 ratio in the cytoplasm (24). It was also shown that canine Gal-3 is phosphorylated at the NH 2 -terminal Ser 6 (major) and Ser 12 (minor) (25), and the major acidic residues on both sides of Ser 6 make it a substrate for casein kinase I and /or for casein kinase II (24,25). The human Gal-3 undergoes phosphorylation by casein kinase I only at Ser 6 , and the phosphorylation significantly reduces its binding ability to its ligands, e.g. laminin and asialomucin, whereas dephosphorylation fully restores the sugar binding activity (26), implying that phosphorylation of Gal-3 at Ser 6 serves as an "on/off" switch for its sugar-binding capabilities. The function of Gal-3 phosphorylation to its other biological activities, however, remained unknown.
We investigate the cellular function of Gal-3 phosphorylation and its impact on apoptosis. Gal-3 mutants generated by sitedirected mutagenesis were transfected into the breast carcinoma BT-549 cells that are Gal-3 null. Ser 6 mutation resulted in loss of Gal-3 anti-apoptotic activity and its involvement in cell cycle arrest in response to loss of cell-substrate interactions as compared with the wild type species.

EXPERIMENTAL PROCEDURES
Cells and Monolayer Culture Conditions-The breast carcinoma cell line BT-549 was obtained from Dr. E. W. Thompson (Vincent T. Lombardi Cancer Research Center, Georgetown University Medical Center, Washington, D. C.). Cells on tissue culture dishes were grown in DMEM (Invitrogen) supplemented with 10% heat-inactivated FBS, 2 mM glutamine, nonessential amino acids, and antibiotics (Invitrogen) and maintained at 37°C in a humidified atmosphere of 95% air and 5% CO 2 .
Site-directed Mutagenesis-To generate Ser 6 to Ala (S6A) and to Glu (S6E) point mutations on Gal-3 cDNA, a Quick Change Site-directed Mutagenesis Kit (Stratagene, La Jolla, CA) was employed with appropriate primers as described previously (2). Briefly, pGEM (7ϩ) vector containing human wild type Gal-3 cDNA was used as a template for PCR to generate S6A and S6E point mutations. After amplification, the template DNA was cleaved with DpnI restriction enzyme, and the DpnI-treated DNA was transferred into Escherichia coli XL1-Blue supercompetent cells. Recombinant plasmid pGEM (7ϩ)/Gal-3 mutant S6A and S6E were purified and sequenced. The sequences were confirmed by the Macromolecular Core Facility of Wayne State University.
Stable Transfection of Gal-3 cDNAs-Wild type and Ser 6 mutant Gal-3 cDNAs were excised from pGEM (7ϩ) containing them with EcoRI and inserted into a mammalian expression vector pBK-CMV (Stratagene) at the EcoRI site in the sense direction. The proper orientation of the cDNA insert was confirmed by restriction enzyme analysis. Then each purified plasmid DNA was transfected into BT-549 using LipofectAMINE (Invitrogen). After 48 h, 800 g/ml G418 (Invitrogen) was added to the culture for 14 days to obtain stable transfected clones. The resultant clones were named BT-549/V for the control vectortransfected cells, BT-549/Gal-3WT1 and BT-549/Gal-3WT2 for the wild type Gal-3-transfected cells, BT-549/Gal-3S6A1 and BT-549/Gal-3S6A2 for the S6A mutant Gal-3-transfected cells, and BT-549/Gal-3S6E1 and BT-549/Gal-3S6E2 for the S6E mutant Gal-3-transfected cells, respectively. These transfected cells were maintained in complete DMEM containing 400 g/ml G418 sulfate.
Gal-3 Expression in Parental BT549 and Transfected Cells-Subconfluent cells were harvested and washed with CMF-PBS. Two million cells were then lysed with 500 l of SDS-PAGE sample buffer (1% SDS, 62.5 mM Tris-HCl, pH 6.8, 10% glycerol). Each protein concentration was measured using BCA protein assay reagent (Pierce). Fifteen-g aliquots of the cell lysates were boiled for 5 min after adding 5% ␤-mercaptoethanol and 1% bromphenol blue, separated by 12.5% SDS-PAGE, and transferred to a polyvinylidene difluoride membrane. Expression levels were detected using monoclonal rat anti-Gal-3 antibody (TIB166; American Type Culture Collection, Manassas, VA) and the enhanced chemiluminescence (ECL) system. To confirm whether the loading amount was same in all lanes, actin expression levels of each cell line were compared using anti-actin antibody (Ab-1; Oncogene Research Products, Boston) after stripping the membrane with Restore TM Western blot Stripping Buffer (Pierce).
Metabolic Labeling and Immunoprecipitation-For labeling in vivo with 32 P, 8 ϫ 10 5 cells were grown overnight in 35-mm dishes. To deplete endogenous phosphate before labeling, the cells were washed twice with phosphate-free DMEM (Invitrogen) and incubated in phosphate-free DMEM supplemented with 5% dialyzed FBS for 2 h at 37°C. The cells were then labeled with 0.2 mCi/ml 32 P (ICN Biomedicals, Costa Mesa, CA) in phosphate-free DMEM supplemented with 10% dialyzed FBS for 5 h at 37°C. After two washes with ice-cold phosphatefree DMEM, the cells were lysed with 400 l of ice-cold lysis buffer (50 mM Tris-HCl, pH 7.5, 120 mM NaCl, 0.5% Nonidet P-40, 2 mM EDTA, 5 g/ml leupeptin, 10 g/ml aprotinin, 500 M phenylmethylsulfonyl fluoride, 200 M sodium orthovanadate, 100 mM sodium fluoride) and incubated for 30 min at 4°C. The cell lysates from two dishes (1.6 ϫ 10 6 cells) were combined and then centrifuged at 12,000 ϫ g for 5 min, and the supernatant was precleared by overnight agitating at 4°C with 30 l of a 1:2 slurry of protein A-Sepharose 6MB (Amersham Biosciences) in ice-cold lysis buffer. After adjusting radioactivity and volume (800 l) of each sample, immunoprecipitation was initiated by adding 2 g of polyclonal rabbit anti-Gal-3 antibody (2) to the precleared supernatant. The reaction mixture was incubated at 4°C for 2 h and then 30 l of 1:2 slurry of protein A-Sepharose 6MB in ice-cold lysis buffer was added to the mixture. After agitating at 4°C for 1 h, the immunoprecipitates were washed five times with ice-cold lysis buffer. The washed precipitates were boiled for 5 min in SDS-PAGE sample buffer containing 5% ␤-mercaptoethanol and then separated by 12.5% SDS-PAGE. Gal-3 phosphorylation state was analyzed by PhosphorImager (Molecular Dynamics, Sunnyvale, CA), and immunoprecipitated Gal-3 was detected by immunoblotting using TIB166. Equal loading amount was confirmed by checking actin expression in the same membrane.
Cell Viability and CDDP Treatment-Cell viability was assessed by a trypan blue dye exclusion test. Briefly, 8 ϫ 10 5 cells were cultured in a 60-mm dish with 25 M CDDP (Sigma) for the indicated times. Both attached and detached cells were then thoroughly collected, and cell viability was determined by trypan blue exclusion.
Apoptosis Assays-Apoptosis was assessed by the PARP degradation and the accumulation of cells at sub-G 1 fraction. Briefly, cells incubated with or without 25 M CDDP for 72 h were completely harvested. For the former analysis, the cells were lysed with SDS lysis buffer as described above, and 20-g aliquots of the cell lysates were separated by 6% SDS-PAGE, transferred to a polyvinylidene difluoride membrane, and probed with mouse anti-PARP antibody (C-2-10; Biomol Research Laboratories, Plymouth Meeting, PA). For the latter analysis, the cells were fixed with 80% ethanol at Ϫ20°C for at least 4 h. After a wash with CMF-PBS, the cells were resuspended at a concentration of 1 ϫ 10 6 cells/ml with staining solution that consisted of 50 g/ml propidium iodide (Sigma), 0.1% Triton X-100, 0.1 mM EDTA, pH 8.0, and 50 g/ml RNase A in CMF-PBS. The samples were then incubated at room temperature for 20 min in the dark and analyzed by flow cytometer (FACSCalibur; Becton Dickinson, San Jose, CA). CellQuest (Becton Dickinson) and ModFit LT (Verity Software House, Topsham, ME) were used as software for the analysis.
Cell Suspension-Poly-HEMA (Sigma) was solubilized in methanol (50 mg/ml) and diluted in ethanol to a final concentration of 10 mg/ml. To prepare poly-HEMA-coated dishes, 4 ml of poly-HEMA solution were placed onto 100-mm dishes and dried in a tissue culture hood. The poly-HEMA coating was repeated twice, followed by three washes with CMF-PBS. One million five hundred thousand cells were plated onto poly-HEMA-coated dishes for 24 h.
Anoikis and Cell Cycle Distribution-Anoikis was assessed by the PARP degradation and the sub-G 1 fraction of propidium iodide-labeled cells, and cell cycle distribution was assessed by each cell cycle phase fraction of propidium iodide-labeled cells. Briefly, cells incubated as monolayer or in suspension culture for 24 h were collected and used for PARP immunoblot and flow cytometric analyses as described above. Cells grown for 24 h as monolayer or in suspension were harvested and lysed with SDS lysis buffer. Equal amounts of the cell lysates were separated by SDS-PAGE and transferred to a polyvinylidene difluoride membrane. Expression levels of cyclins were measured using the following antibodies: anti-cyclin A (SC-239; Santa Cruz Biotechnology, Santa Cruz, CA), anti-cyclin D 1 (DCS-6; Sigma), anti-cyclin E (HE-12; PharMingen, San Diego, CA), anti-p21 WAF1/CIP1 (SX118; PharMingen), anti-p27 KIP1 (M-197; Santa Cruz Biotechnology).
Next, we examined whether wild type and Ser 6 mutant Gal-3 were phosphorylated in vivo. BT-549/V, /WT1, and /S6A1 cells were metabolically labeled with 32 P as described under "Experimental Procedures," and cell extracts were immunoprecipitated with anti-Gal-3 polyclonal antibody. Western blot analy-sis using monoclonal anti-Gal-3 antibody revealed that the immunoprecipitates of both BT-549/WT1 and /S6A1 cell extracts contained Gal-3 (Fig. 1II, B), whereas only BT-549/WT1 cell extracts expressed the phosphorylated form of Gal-3 (Fig.  1II, C). These results confirmed that the human Gal-3 is phosphorylated in vivo at Ser 6 as described previously (2) for BT-549 cells transfected with wild type Gal-3 cDNA. In addition, we have recently described endogenous Gal-3 phosphorylation in human colon cancer cells (26). Western blot analysis with anti-phosphoserine antibody confirmed serine phosphorylation in the wild type-transfected cells (data not shown).
Ser 6 Mutation Alters the Anti-apoptotic Activity of Gal-3-Previously, it was shown (10) that Gal-3 inhibits CDDP-induced apoptosis. Thus, we examined the effect of Gal-3 phosphorylation on its anti-apoptotic activity. All clones were treated with 25 M CDDP, and subsequently cell viability was determined by a trypan blue dye exclusion. CDDP induced cell death in BT-549/P, /V, /S6A, and /S6E cells when compared with BT549/WT cells (Fig. 2I). BT-549/P, /V, /S6A, and /S6E cells showed similar sensitivity to CDDP treatment; a 72-h exposure to CDDP resulted in a decrease in cell viability of 55% of these cells. On the other hand, more than 75% of BT-549/WT cells were refractory even after a 72-h exposure to CDDP.
To determine whether CDDP-treated cells underwent apoptosis, we examined the apoptosis-specific cleavage of PARP, an early event in apoptosis resulting from the activation of interleukin-1␤ converting enzyme/Ced-3 family members (2). Fig. 2II showed that CDDP induced the cleavage of intact PARP (M r 116,000) into the inactive fragment (M r 85,000) only in the BT-549/P, /V, /S6A, and /S6E cells, whereas in the wild type Gal-3-transfected BT-549/WT cells, PARP remained intact.
We further evaluated the sub-G 1 fraction of propidium iodide-labeled cells by flow cytometry to confirm that CDDP induced apoptosis to the cells as described previously (10). The sub-G 1 population of BT-549/V and /S6A cells occupied more than 30% of the DNA histogram after a 72-h exposure to 25 M CDDP, whereas that of BT-549/WT occupied only about 10% (Fig. 2III, D-F). It should be mentioned that we obtained similar DNA histograms for BT-549/P and /S6E cells compared with BT-549/V and /S6A cells, respectively (data not shown). These results demonstrate that wild type Gal-3 inhibits CDDPinduced apoptosis, whereas abrogation of its phosphorylation ability results in loss of inhibitory capability.
Ser 6 Mutant Gal-3 Fails to Inhibit Anoikis and to Induce Cell Cycle Arrest-A recent study revealed that Gal-3 inhibits anoikis-inducing G 1 arrest (13). We therefore examined the effect of Gal-3 phosphorylation on its regulation of anoikis and cell cycle arrest. Anoikis was induced by culturing cells in suspension for 24 h on a non-adhesive substrate of poly-HEMA. Subsequently, we presented the data of BT-549/WT1, /S6A1, and /S6E1 as their representative to avoid redundancy. We monitored the PARP degradation as a marker for anoikis. No PARP degradation was detected in all clones when cells were cultured as a monolayer (Fig. 3I); however, when cells were cultured in suspension, proteolytic cleavage of PARP was readily detected in BT-549/V, /S6A, and /S6E cells, whereas it was significantly inhibited in BT-549/WT cells (Fig. 3I).
Flow cytometric DNA analysis to determine the cell cycle phase distribution of cells grown as monolayer or in suspension for 24 h was performed next, considering that cell cycle regulation is associated with a loss of cell-matrix interactions (13,28). In monolayer cultures, the cell cycle phase distribution of both BT-549/V, /WT, and /S6A cells showed a similar pattern (Fig. 3II, A-C). In suspension cultures, however, there were two prominent differences in the DNA histogram pattern among these three clones (Fig. 3II, D--F). First, the percentages of BT-549/V and /S6A cells in sub-G 1 phase were 10.6 and 10.7%, respectively, whereas only 0.1% of BT-549/WT cells were in sub-G 1 phase. Second, a dramatic decrease in S phase (from 47.8 to 8.8%) and a dramatic increase in G 1 phase (from 38.1-75.1%) were detected in BT-549/WT cells in response to a loss of cell-substrate interactions, whereas no such remarkable change was observed in BT-549/V and /S6A cells. These results of the PARP degradation and flow cytometric DNA analysis suggest that Gal-3 protects BT-549 cells from anoikis by cell cycle arrest at G 1 phase in response to a loss of cell-matrix contact and that a mutation at Ser 6 abrogates such a function of Gal-3.
Wild Type Gal-3 Alters Expression Levels of Cell Cycle Regulators-To have a further understanding of the differences in cell cycle responses between BT-549/WT and /S6 mutant cells following loss of cell-substrate interactions, we examined the Gal-3 effect on the expression levels of cell cycle regulators. It was shown that expression of cyclin D 1 , an early G 1 cyclin that promotes cell cycle into late G 1 phase, is dependent upon cell adhesion (28,29). As shown in Fig. 4, cyclin D 1 expression was up-regulated in BT-549/WT cells when cultured as monolayer. Moreover, cyclin D 1 expression in BT-549/WT cells was further up-regulated in suspension culture, showing that wild type Gal-3 induces cyclin D 1 overexpression independent of cellmatrix interactions. Cyclin E is a nuclear protein essential for the G 1 to S phase transition (28,29). Expression of cyclin E was down-regulated in BT-549/WT cells, and the expression level was not significantly altered by the culture conditions (Fig. 4). Expression of cyclin A, a late G 1 to S cyclin, is coincident with and necessary for the onset of S phase (28). In a monolayer culture, cyclin A expression was up-regulated in BT-549/WT cells and was significantly down-regulated in BT-549/WT in response to a detachment of cells from the substrate (Fig. 4). p21 WAF1/CIP1 and p27 KIP1 are CDK inhibitors that mainly act as negative regulators of G 1 to S phase transition (30, 31). As shown in Fig. 4, the expression of p21 WAF1/CIP1 and p27 KIP1 was up-regulated in BT-549/WT cells when cultured as monolayer, and their expression was further up-regulated in suspension culture. Taken altogether, these results suggest that in response to loss of cell-substrate interactions, wild type Gal-3 up-regulates cyclin D 1 expression resulting in passing the apoptosis-sensitive point in early G 1 and up-regulates expression of CDK inhibitors (p21 WAF1/CIP1 and p27 KIP1 ) and down-regulates cyclin A expression resulting in late G 1 arrest.
In contrast, such alteration in expression levels of cyclins and CDK inhibitors was observed in neither BT-549/V nor /S6 mutant cells in response to loss of cell-substrate interactions (Fig. 4). The clear difference in the expression levels of cell cycle regulators between BT-549/WT and /S6 mutant cells in response to loss of cell-substrate interactions demonstrates that G 1 arrest induced by Gal-3 to avoid anoikis depends on its ability to undergo phosphorylation. tion regulates its carbohydrate recognition (24). Studies using site-directed mutagenesis have shed new light on diverse protein functions (2, 10, 15-18, 20, 21); thus we employed this method to gain an insight into the relationship between Gal-3 phosphorylation and its biological function. From the data, we have concluded that Gal-3 phosphorylation is required for its anti-apoptotic activity and anti-anoikis activity with G 1 cell cycle arrest.
It is now well established that Gal-3 overexpression correlates with increased metastatic potential in some cancers (1,6,8). Gal-3 may render this property to tumor cells by its antiapoptotic and anti-anoikis activities, which are thought to be critical for anchorage-independent cell survival in the circulation during dissemination. We show that the anti-death activity of Gal-3 is regulated by its phosphorylation. Moreover, the enhanced metastatic potential related to Gal-3 overexpression seems to be promoted by the property that cell surface Gal-3 mediates homotypic cell aggregation and tumor cell adhesion to endothelial cells and to extracellular matrix through binding with its complementary glycoconjugates, which is also regulated by its post-translational modification, i.e. Ser 6 phosphorylation of human Gal-3 significantly reduces the saturation binding to its complementary glycoconjugates and dephosphorylation fully restores the binding (5,7,26). Taken together, Gal-3 phosphorylation may play a pivotal role in the biological function of Gal-3.
Although Gal-3 does not belong to the Bcl-2 family, it con- tains sequence and functional similarities to Bcl-2; Gal-3 has the four amino acid anti-death motif (Asp-Trp-Gly-Arg) that is highly conserved within the BH1 domain of the Bcl-2 family (9,10,14,15), and both molecules undergo post-translational modification of serine phosphorylation. Recently, it has been shown (16 -21) that phosphorylation of Bcl-2 at Ser 70 appears to be critical for its anti-apoptotic function; however, the role of this phosphorylation is uncertain due to conflicting reports (17)(18)(19)(20)(21) showing that such a phosphorylation may activate or inactivate Bcl-2 anti-apoptotic function. These contradictory conclusions may result from the fact that these studies were performed using different types of tumor cells and apoptotic signals or from the fact that multiple kinases have been implicated in the phosphorylation of Bcl-2, including Raf-1 kinase (22), protein kinase C (17), protein kinase A (23), Jun NH 2 -terminal kinase/ stress-activated protein kinase (20), and v-cyclin-CDK6 (21). On the other hand, the human Gal-3 is phosphorylated at Ser 6 by casein kinase I (25,26). Casein kinase I was shown to phosphorylate tumor necrosis factor receptor, which negatively regulates receptor-mediated tumor necrosis factor signaling for apoptosis (33). Furthermore, casein kinase I was suggested to be a conserved component of the Wnt pathway (34) and to be a positive regulator of this pathway and a link between upstream signals and the complexes that regulate ␤-catenin (35). Casein kinase I might also act in the ␤-catenin-independent pathway (36). Recently, it was found that Wnt signaling can inhibit drug-induced apoptosis suggesting that it may exhibit its oncogenic potential through a mechanism of anti-apoptosis (37).
Thus, it appears that casein kinase I plays a role in regulation of apoptotic function of diverse molecules and signaling pathways.
We demonstrated here that Gal-3 also modulates its antiapoptotic activity and cell cycle arrest in response to a loss of cell-substrate interactions via its phosphorylation. Thus, the finding of a new similarity in post-translational modification between Gal-3 and Bcl-2 should lead to further understanding of regulation of Gal-3 signaling. Taken together, these finding indicate that the regulation of Gal-3 phosphorylation transduces signal(s) for carbohydrate binding and anti-apoptotic function.