Cell Cycle Arrest and Inhibition of Anoikis by Galectin-3 in Human Breast Epithelial Cells¹

Cell-matrix interactions regulate many cellular responses including gene expression, differentiation, and cell survival (1, 2, 3). The ECM³ is a complex network of macromolecules that is composed of collagens, laminin, fibronectin, proteoglycans, and many soluble molecules including growth factors (4). The ECM of both the prostate and the lactating mammary gland undergo drastic involution upon removal of the appropriate trophic hormones (5). Because hormone ablation induces secretion of ECM-degrading enzymes that disrupt the interaction of epithelial cells with the ECM, it was hypothesized that the loss of cell-matrix interactions induces anoikis, a specific form of apoptosis (6).

Integrins are heterodimeric cell surface receptors that mediate cell adhesion to ECM (7, 8). Neutralizing antibodies against integrins induce cell detachment, followed by anoikis in epithelial cells, suggesting a role for integrin signaling in the regulation of anoikis (9). Integrin and growth factor receptor signals coregulate the turnover of phospholipids and play nonredundant roles in cell cycle progression (10, 11). Growth factor signaling triggers cell cycle progression from G₀ to G₁, whereas cell adhesion mediates the transition through the restriction point contained in late G₁. Cyclin D₁ expression and the activities of cyclin E-associated kinases, the functions of which are critical for G₁-S transition, are shown to depend on cell adhesion (10, 11). This may explain why nonadherent cells fail to transit the cell cycle. Although normal epithelial cells require cell-extracellular matrix interactions for cell survival and growth, invasive and motile mesenchymal cells survive without these interactions and become arrested at the G₁ stage of the cell cycle (6). This shows that anoikis sensitivity is associated with cell cycle regulation.

Because retaining anchorage-independent viability of disseminating cells in the circulation is critical for tumor cell metastasis, we investigated the role of galectin-3 in the cellular response to the loss of cell contact. The human galectin-3 is a M_r 31,000 glycoprotein that belongs to a family of galactoside-binding proteins that are highly expressed in many human tumor cells including breast carcinoma cells (12, 13, 14, 15, 16). Structurally, galectin-3 is composed of two distinct domains: an NH₂-terminal of 12 amino acids preceding a collagen-like sequence, and a globular COOH-terminal domain containing the carbohydrate-binding site (12, 17). Although galectin-3 is not a bcl-2 family member, it contains the same four amino acid motif (NWGR) conserved in the BH1 domain of the bcl-2 gene family. This motif is critical for bcl-2 antiapoptotic activity (18). Consistently, we and others have found that galectin-3 is a novel antiapoptotic gene product (19, 20) and showed that substitution of the Gly¹⁸² residue with Ala in the NWGR motif of galectin-3 abrogates its antiapoptotic activity (19). Thus, similar to the bcl-2 protein (18), the NWGR motif appears to be critical for the antiapoptotic function of galectin-3.

Although the exact function of galectin-3 is still debatable, galectin-3 expression has been associated with transforming and metastatic potentials (12, 13, 14, 15, 16). Ectopic expression of galectin-3 in the human breast cancer cell line BT549 resulted in the acquisition of a potent tumorigenic phenotype in nude mice (15, 21). We questioned whether oncogenic activity of galectin-3 involves anti-anoikis activity. Because cell cycle regulation is associated with anoikis (6, 9, 11), we have investigated the roles of galectin-3 on cell cycle regulation during anoikis. We report that galectin-3 prevents anoikis in human breast epithelial cells. We also present data that suggest that galectin-3-mediated inhibition of anoikis may result from its ability to induce cell cycle arrest at late G₁ through expression modulation of cyclins and their inhibitors.

The human breast cancer cell line BT549 was obtained from Dr. E. W. Thompson (Vincent T. Lombardi Cancer Research Center, Georgetown University Medical Center, Washington, DC). Galectin-3-transfected BT549 cell clones were established previously by introducing an expression vector containing human galectin-3 cDNA in either the sense or the antisense orientation into BT549 parental cells (19, 21). Galectin-3 antiapoptotic activity was characterized previously using three independent, galectin-3-expressing BT549 clones (11811, 11913, and 11914; Ref. 19). In the present study, we have also used these three clones. We present data mostly obtained using clone 11914 to avoid redundancy. Hereafter, galectin-3-transfected BT549 clone 11914, antisense galectin-3-transfected BT549 clone 41421, and neo-resistant control vector-transfected BT549 (19) are referred to as BT549-Gal_wt, BT549-Gal_AS, and BT549neo, respectively. The mutant galectin-3-expressing BT549 in which the Gly¹⁸² of the NWGR motif was substituted to Ala (Ref. 19) is referred as to BT549-Gal_m. Cells on tissue culture dishes (Sarstedt, Newton, NC) were grown in RPMI 1640 supplemented with 10% heat-inactivated fetal bovine serum, 100 units/ml penicillin, 100 μg/ml streptomycin, 2 mm l-glutamine, and 0.5 μg/ml fungizone in a 95% air and 5% C0₂ incubator at 37°C.

PolyHEMA (purchased from Aldrich Chemical Co., Madison, WI) was solubilized in methanol (50 mg/ml) and diluted in ethanol to a final concentration of 10 mg/ml. To prepare polyHEMA-coated dishes, 4 ml of polyHEMA solution were placed onto 100-mm Petri dishes and dried in a tissue culture hood. The polyHEMA coating was repeated twice, followed by three washes with PBS. Anoikis was induced by culturing 1.5 × 10⁶ cells on polyHEMA- coated 100-mm dishes in a humidified 5% CO₂ incubator with RPMI 1640 supplemented as described above.

The percentage of cells in each cell cycle phase was determined by flow cytometry (PAS-II; Partec AG, Münster, Germany). Cells grown for 24 h on monolayer or in suspension (polyHEMA-coated dishes) were trypsinized, suspended in culture medium, centrifuged at 150 × g for 5 min, and fixed with 70% ethanol. The fixed cells were centrifuged, and the cell pellet was resuspended in Hoechst staining solution at a concentration of 1 × 10⁶ cells/ml and incubated for 3 min at room temperature. The staining solution consisted of 3 μg/ml Hoechst 33258 (Sigma Chemical Co., St. Louis, MO) in Tris buffer (2 mm MgCl₂, 0.1% Triton X-100, 154 mm NaCl, and 100 mm Tris, pH 7.5). The stained cells (30,000) were analyzed by flow cytometry. Optical filtration included a UG-1 excitation filter, a 420-nm dichroic filter, and 435-nm long pass emission filter. Single-channel data were acquired and subsequently analyzed with a computer program (Phoenix Flow Systems, San Diego, CA).

Protein extracts were prepared by lysing 1 × 10⁶ cells with 50 μl of 2× SDS sample buffer, and protein concentration was measured using BCA protein assay reagents (Pierce, Rockford, IL). Extracts were boiled for 10 min, chilled on ice, and subjected to SDS-PAGE analysis, followed by electrophoretic transfer to a nitrocellulose membrane. Expression levels were measured using the following antibodies: anti-PARP (C-2–10; Biomol Research Laboratory, Plymouth Meeting, PA), anti-galectin-3 (American Type Culture Collection, Rockville, MD), anti-human bcl-2 (DAKO A/S, Denmark), anti-cyclin D₁ (Ab2; Oncogene Research, Cambridge, MA), anti-cyclin E (HE67; PharMingen, San Diego, CA), anti-cyclin A (BF683; Santa Cruz Biotechnolgy, Santa Cruz, CA), anti-p15 (C-20; Santa Cruz Biotechnology), anti-p21^WAF1/CIP1 (Cal Biochem, San Diego, CA), anti-p27^KIP1 (M-197; Santa Cruz Biotechnology), and anti-Rb (SC4112; Santa Cruz Biotechnology).

Extracts were prepared by lysing the cells in ice-cold buffer containing 50 mm Tris (pH 7.5), 150 mm NaCl, 25 mm NaF, 1 mm sodium vanadate, 1 mm EDTA, 2 mm benzamidine, 1 mm phenylmethylsulfonyl fluoride, 25 mm NaF, 0.5% NP40, and 10 μg/ml leupeptin and aprotinin. After removing the precipitate by centrifugation at 14,000 rpm for 15 min at 4°C, the cyclin D₁ and cyclin E complexes were collected using the anti-cyclin D₁ or anti-cyclin E antibodies and protein G-Sepharose. The complex was washed three times with 1.0 ml of washing buffer (1% deoxycholate, 0.5% Tween 20, and 50 mm Tris-HCl, pH 7.5) and once with 1 ml of 50 mm Tris (pH 7.5) and 10 mm MgCl₂. Kinase activity of the cyclin D₁-immune complex was measured using 2.5 μm GST-pRb protein (purchased from Santa Cruz Biotechnology), and cyclin E-complex using histone H1 protein (Boehringer Manheim, Indianapolis, IN) as substrate in 50 mm Tris-HCl (pH 7.5), 10 mm MgCl₂, 1 mm DTT, 0.5 mm NaF, 0.1 mm sodium orthovandate, 5 μg/ml leupeptin, and [γ-³²P]ATP (0.01–0.3 nCi/pmol). After kinase reaction at 30°C for 30 min, samples were mixed with 2× SDS buffer and subjected to SDS-PAGE analysis. [γ-³²P]ATP-labeled Rb and histone H1 proteins were visualized by autoradiography.

Anoikis was induced by culturing cells in suspension using polyHEMA-coated tissue culture dishes, which prevent cell adhesion, as described in “Materials and Methods.” To determine whether galectin-3 inhibits anoikis, we examined the apoptosis-specific cleavage of PARP, an early event in apoptosis resulting from the activation of interleukin 1β-converting enzyme (caspase)/Ced-3 family members (22). As shown in Fig. 1 A, proteolytic cleavage of PARP was readily detected in the parental control and BT549-Gal_AS cells, whereas it was significantly inhibited in BT549-Gal_wt cells, even after 3 days in suspension culture. Galectin-3 inhibition of anoikis was also confirmed by nuclear morphological analysis. The nuclei were stained using Hoechst 33258 after 48 h in suspension culture, as described previously (23). The parental control and the BT549-Gal_AS cells showed fragmented nuclei that were consistent with nuclear morphological changes in other apoptotic cells (23). In contrast, no significant changes in nuclear morphology could be observed in BT549-Gal_wt cells (data not shown).

Because cell cycle regulation is associated with sensitivity to anoikis (6, 9, 11), flow cytometric DNA analysis was performed to determine the cell cycle phase distribution of cells grown on monolayer or in suspension for 24 h. In exponentially growing monolayer cultures, the percentage of BT549neo in G₁ phase was 46.8%, whereas only 26.2% of BT549-Gal_wt was in the G₁ phase (Fig. 2, A and C). In suspension culture of BT549neo, the G₁ population decreased to 25.4%, and the sub-G₁ population increased from 7 to 25.4% (Fig. 2, A and B). Flow cytometric DNA analysis, morphological analysis, and PARP cleavage showed that BT549 cells underwent apoptotic cell death in response to the loss of cell-substrate interactions. In contrast, no sub-G₁ population of BT549-Gal_wt was detected, regardless of the state of cell-substrate interactions, whereas a dramatic increase in the G₁ population (from 26.2 to 78.6%) was detected in BT549-Gal_wt in the absence of cell-substrate interactions (Fig. 2, C and D). These results show that galectin-3 overexpression renders BT549 cells able to respond to the loss of cell- ECM contact by growth arrest at G₁ without detectable apoptotic cell death.

After the loss of cell-substrate contacts, the G₁ population increase of BT549-Gal_wt was accompanied by a 6.7-fold decrease in the proportion of cells in S phase, suggesting that galectin-3 overexpression inhibits G₁ to S phase progression. Cyclin E-associated kinases are activated late in G₁ phase, and their activities are thought to be critical for the progression to S phase (24, 25, 26). Cyclin A-associated kinase activities are then activated concomitantly with the onset of S phase and involved in maintenance of S phase (27, 28, 29). Therefore, we sought to determine whether expression of these cyclins is modulated by galectin-3 when cultured on monolayer or in suspension. As shown in Fig. 3, expression of cyclin E was down-regulated by galectin-3, especially after the loss of cell-substrate interactions. In monolayer culture, cyclin A expression was up-regulated by galectin-3, in agreement with a proportional increase in BT549-Gal_wt in S phase compared with BT549neo. However, cyclin A expression was significantly down-regulated in BT549-Gal_wt upon cell detachment, whereas the loss of cell-substrate interactions had no effect on cyclin A expression in BT549neo. These data (Fig. 3, A and B) suggest that galectin-3-mediated G₁ arrest in the absence of cell anchorage involves drastic down-regulation of cyclin E and cyclin A expression.

Kinase activities of cyclin E and cyclin A complexes are regulated by CDKIs, including p21^WAF1/CIP1 and p27^KIP1 (30, 31, 32). Therefore, we also examined the effects of galectin-3 overexpression on CDKI expression with regard to cell-matrix interactions. As shown in Fig. 3,C, expression of both p21^WAF1/CIP1 and p27^KIP1 was up-regulated by galectin-3 and further increased after cell-matrix detachment (Fig. 3,C). Galectin-3 modulation of cyclin E and cyclin A as well as their inhibitors in suspension culture is consistent with cell cycle arrest at mid-G₁ as shown in flow cytometric analysis (Fig. 2,D). However, the immunoblot analysis showed that galectin-3 down-regulated cyclin E and slightly up-regulated p21^WAF1/CIP1 and p27^KIP1, even in proliferating monolayer culture. To assess the functional significance of changes in expression of cyclin E and its inhibitors, we measured cyclin E-associated kinase activity in vitro using histone H1 as substrate. Surprisingly, cyclin E-associated kinase activity was not drastically altered by galectin-3 on monolayer culture (Fig. 4 A), whereas its activity is significantly down-regulated by galectin-3 in suspension culture. The cyclin E-associated kinase activity agrees with G₁-S transition in monolayer culture of BT549-Gal_wt and G₁ arrest in suspension culture. In monolayer culture, it is not clear how cyclin E-associated kinases are activated in BT549-Gal_wt cells with low levels of cyclin E protein as efficiently as in BT549neo (see “Discussion”).

To determine whether the in vivo kinase activities of cyclin complexes were modulated by galectin-3, we examined the status of Rb phosphorylation in monolayer or suspension cultures of three independent galectin-3-overexpressing BT549 clones. The Rb protein is hypophosphorylated in G₀ and early G₁ phases and binds to a series of cellular proteins including E2F transcription factor. Rb protein becomes hyperphosphorylated by D-type cyclin-dependent kinases and further phosphorylated by cyclin E- and A-dependent kinases (33, 34). The hyperphosphorylated pRb becomes inactive and is no longer capable of binding to its interacting molecules such as E2F. This hyperphosphorylated state is maintained through S, G₂, and most of M phases. As shown in Fig. 4,B, hyperphosphorylated Rb was dominant in three clones, indicating that most of the cells are in S to M phases when cultured on monolayer (also see Fig. 2,C). In contrast, only hypophosphorylated Rb was detected in suspension cultures of BT549-Gal_wt clones, showing that these cells fail to enter S phase, and cdk activities are down-regulated, as predicted by flow cytometric analysis and the immunoblot assay (Figs. 2 and 3). It should be noted that the Rb protein in the control cells was barely detectable, so the phosphorylation status of Rb protein in BT549neo was not determined.

Recent studies showed that cell cycle progression is coregulated by growth factors and cell anchorage. Expression of the D-type cyclins, early G₁ cyclins, requires cell adhesion (10, 11). To examine whether galectin-3 modulates the expression of the D-type cyclins, immunoblot analysis of cyclin Ds was performed. As shown in Fig. 3 A, cyclin D1 expression was induced by galectin-3 when cultured on monolayer. Interestingly, galectin-3 up-regulation of cyclin D₁ expression was further enhanced in suspension cultures. In BT549 cells, neither cyclin D₂ nor D₃ was detected, irrespective of galectin-3 expression levels (data not shown). This suggests that galectin-3 up-regulates cyclin D₁ expression independently of cell-substrate interactions.

The D-type cyclins are often inhibited by a family of inhibitors including p15^INK4B, p16^INK4A, and p19^INK4D (35). When the levels of these inhibitors were examined in BT549 cells, neither p16^INK4A nor p19^INK4D was detectable, and expression of p15^INK4B was not significantly altered by galectin-3, regardless of cell adhesion (Fig. 3,C). We next questioned whether galectin-3 induction of cyclin D1 expression lead to increased kinase activity. We examined whether cyclin D₁ in BT549-Gal_wt cells was complexed to its functional partners (CDKs). The cyclin D₁ complex was immunoprecipitated, and its associated kinase activity was measured using purified GST-pRb fusion protein as substrate. The Rb fusion proteins of M_r 46,000 were more efficiently phosphorylated by cyclin D₁-associated kinases in BT549-Gal_wt cell lysates compared with those in BT549-neo in the absence of cell anchorage (Fig. 5). Although in vitro kinase assay showed that galectin-3 induction of cyclin D1 lead to activation of cyclin D1-associated kinases, the phosphorylated form of Rb was not detected in BT549-Gal_wt suspension culture (Fig. 4 B). This is in agreement with previous reports that maintenance of hyperphosphorylation of Rb in vivo requires cyclin E and cyclin A-dependent kinase activities (33, 34).

The significance of the NWGR motif for galectin-3 regulation of anoikis was evaluated using a galectin-3 mutant in which the Gly¹⁸² of the NWGR motif was substituted to Ala (Ref. 19; Fig. 6). Immunoblot analysis ensured that the mutant and wild-type galectin-3 proteins were stably expressed after the loss of cell-substrate contact (Fig. 3,A). To explore whether mutant galectin-3 prevents anoikis, BT549-neo, BT549-Gal_wt and BT549-Gal _m cells were cultured on polyHEMA-coated dishes for 24 h. Unlike in BT549-Gal_wt cells, apoptosis-specific cleavage of PARP protein occurred in both BT549-Gal_m cells and control cells (Fig. 1, B and C).

We next examined the significance of the NWGR motif of galectin-3 for cell cycle regulation after loss of cell-substrate interactions. As shown in Fig. 3, mutant galectin-3 failed to modulate expression of cyclin D₁, cyclin E, cyclin A, p21^WAF1/CIP1, and p27^KIP1. In addition, in vitro kinase assay of the cyclin D₁ and cyclin E complexes showed that a substitution of Gly¹⁸² to Ala in the NWGR motif of galectin-3 abolished its ability to modulate kinase activity of the cyclin D₁ and cyclin E complexes after ECM detachment (Fig. 5). These results reemphasize that the integrity of the NWGR motif in the BH-1-like domain of galectin-3 is a prerequisite for its antiapoptotic activity as well as its ability to modulate expression of cell cycle regulators.

Although survival of “stationary” epithelial cells depends on cell-cell and cell-matrix interactions, invasive and motile mesenchymal cells survive without these interactions (36). Cancer cells often acquire a “fibroblast-like” phenotype and become invasive. The present study showed that galectin-3 overexpression renders breast epithelial cells (BT549) able to respond to the loss of cell-ECM contact by growth arrest at G₁ without detectable apoptotic cell death (anoikis), which resembles the response of fibroblast to loss of cell-ECM contact. Acquisition of anti-anoikis activity may be critical for anchorage-independent cancer cell survival in circulation during metastasis and might explain why galectin-3 expression is enhanced in some tumors with malignant phenotypes (12, 13, 14, 15, 16). A key step in the apoptosis cascade is the activation of members of the interleukin-1β converting enzyme/Ced-3 family (caspases; Ref. 37). This commitment step appears to be regulated by the Ced-9/bcl-2 family (37, 38, 39). Our study clearly suggests the existence of a non-bcl-2 family member that regulates caspase-mediated apoptosis. Like the bcl-2 family, the galectin-3 family includes both proapoptotic and antiapoptotic members. Galectin-1 and galectin-9 are reported to play a role in the thymocyte-epithelial interactions and induce apoptosis of activated T lymphocytes (40, 41, 42, 43). These studies and ours suggest that some members of the galectin family are novel regulators of apoptosis. It is noted that among galectin family members, only galectin-3 has the NWGR motif present in the BH1 domain of the bcl-2 gene family. Galectin-3 may interact with apoptosis-regulating molecules through the NWGR motif and replace bcl-2 activity.

The present study suggests that galectin-3 regulates the cell cycle through modulation of the cell cycle regulating gene expression. Galectin-3 up-regulates cyclin A expression, in agreement with an increase in BT549-Gal_wt in S phase compared with BT549neo. This is consistent with a recent report that the cellular expression of galectin-3 is proliferation dependent (44). Galectin-3 does not alter cyclin E-associated kinase activity, although it down-regulates cyclin E protein expression in monolayer culture. The question is then how cyclin E-associated kinases are efficiently activated in these cells. Many isoforms of cyclin E are detected in breast cancer cells, and activities of these isoforms appear to be regulated by different mechanisms (45). It should be noted that although galectin-3 down-regulates the major cyclin E protein, the lower molecular weight cyclin E protein was induced by galectin-3 in monolayer culture (Fig. 3 B). Galectin-3-induced low molecular weight cyclin E isoform may be efficiently activated in monolayer culture. Or, galectin-3 up-regulation of cyclin D1 may be critical for efficient activation of cyclin E. Cyclin E-associated kinase was previously suggested to be active after it is phosphorylated by cyclin D1-associated kinase (46). Lastly, galectin-3-induced p21^WAF1/CIP1 and p27^KIP1 may regulate cyclin E-associated kinase activities in a bidirectional manner. It was shown previously that the low levels of p21^WAF1/CIP1 and p27^KIP1 promote the assembly of active kinase complexes, whereas at higher concentrations, they inhibit activity (47). Thus, it is possible that galectin-3-induced p21^WAF1/CIP1 and p27^KIP1 in monolayer culture may function as adaptor proteins of cyclin E complexes, which are critical for G₁-S transition, whereas further induction in suspension culture results in inhibition of kinase activity.

Cyclin D₁ overexpression has been associated with a poor prognosis (48, 49, 50, 51, 52), dysregulation of the cell cycle, increased cell proliferation, increased EGF receptor expression, and p53 abnormalities in a transgenic model (53). Cyclin D₁ expression was shown to depend on adhesion in epithelial cells. Interestingly, galectin-3 up-regulates cyclin D₁ expression independent of anchorage. It is not clear whether galectin-3 induction of cyclin D₁ is associated with anti-anoikis activity of galectin-3 and/or with its oncogenic activity. Mounting evidence suggests that apoptosis regulation is tightly linked to regulation of the cell cycle. Although apoptosis can be induced at any point of the cell cycle, sensitivity of apoptosis greatly differs, depending on cell cycle points. Our study suggests that galectin-3 inhibition of anoikis may be related to its ability to regulate critical decision points between cell cycle progression and induction of apoptosis. Galectin-3 induction of cyclin D₁-associated kinase activity may help cells pass the apoptosis-sensitive point in early G₁. Similarly, galectin-3 down-regulation of cyclin E and cyclin A expression may be involved in cell cycle arrest at an anoikis-resistant point as summarized in Fig. 7.

At present, it remains unclear how galectin-3 regulates expression of the cyclins and their inhibitors. Galectin-3 was shown to be a nuclear matrix protein that binds to RNA and single-stranded DNA (54) and identified as a splicing factor in cell-free splicing assay (55). Nuclear galectin-3 may directly modulate gene expression through regulation of transcription and/or mRNA splicing. Or, cytoplasmic galectin-3 may modulate the activity of signaling molecules (e.g., kinases, phosphatases, and proteinases), which, in turn, transduce signals to regulate a set of gene expression. Galectin-3-mediated changes in the levels of the cyclins and their inhibitors may result from changes in gene expression of the adhesion-related genes. Cell adhesion transduces biochemical signaling, including focal adhesion kinase activation that is critical for anoikis regulation as well as cell cycle transition (7, 56, 57, 58). Whether cell cycle regulating genes are primary or secondary effects of galectin-3 overexpression, the present study showed that galectin-3 inhibition of anoikis is associated with cell cycle regulation. However, it remains to be fully investigated whether cell cycle regulation is a prerequisite for galectin-3 regulation of anoikis.

We thank Dr. Erik Thompson of the Lombardi Cancer Center for the gift of the BT-549 parental cells, Dr. Yong J. Lee for flow cytometric analysis, and Mary Ann Krug for preparation of the manuscript.