Galectin-1 Interacts with the α5β1 Fibronectin Receptor to Restrict Carcinoma Cell Growth via Induction of p21 and p27*

Surface binding of galectin family members has the potential to link distinct glycan structures to growth regulation. Therefore, we addressed the antiproliferative potential of galectin-1 (Gal-1) in a panel of carcinoma cell lines. We discovered growth inhibition by Gal-1 in epithelial tumor cell lines from different origins and provide evidence that this effect requires functional interaction with the α5β1 integrin. Antiproliferative effects result from inhibition of the Ras-MEK-ERK pathway and consecutive transcriptional induction of p27. We have further identified two Sp1-binding sites in the p27 promoter as crucial for Gal-1 responsiveness. Inhibition of the Ras-MEK-ERK cascade by Gal-1 increased Sp1 transactivation and DNA binding due to reduced threonine phosphorylation of Sp1. Furthermore, Gal-1 induced p21 transcription and selectively increased p27 protein stability. Gal-1-mediated accumulation of p27 and p21 inhibited cyclin-dependent kinase 2 activity and ultimately resulted in G1 cell cycle arrest and growth inhibition. These data define a novel mechanism whereby Gal-1 regulates epithelial tumor cell homeostasis via carbohydrate-dependent interaction with the α5β1 integrin.

solution, it is suited for cross-linking and capable of forming lattice-like structures with distinct glycans, a process assumed essential for signaling after clustering of cell surface glycoprotein receptors (8).
Studies on activated T-cells have set a precedent that Gal-1 can serve as growth regulator via induction of apoptosis (9). Its widespread differential expression in malignant cells and their non-transformed counterparts raises the possibility that Gal-1 also functions as a modulator of proliferation in other cell types, such as epithelial cells. Indeed, growth inhibition of tumor cells by Gal-1 has been reported (10,11). Of note, these growth effects were unaffected by the presence of a glycan inhibitor (9) or not specified with respect to glycan inhibitors (11), whereas the proapoptotic action of galectins on activated T-cells clearly depended on ␤-galactoside-specific binding (12).
Galectins have been suggested to exert their biological effects in part through interaction with integrins (5,13), although this interaction is still poorly understood in the case of Gal-1. Integrins provide adhesion to the appropriate extracellular matrix, a central requirement for the proliferation of epithelial cells (14,15). As heterodimeric transmembrane receptors, integrins recognize and bind extracellular matrix ligands and thereby control the organization of the intracellular actin cytoskeleton (16,17). In addition to their adhesive function, integrins also initiate and modulate signal transduction cascades. Both ligand occupancy and integrin clustering are critical for the activation of integrins and trigger intracellular signaling cascades (17). Typically, integrins connect to intracellular signaling networks through recruitment, assembly, and activation of other signaling proteins in a bidirectional manner. Thus, integrin signaling converges with other signaling pathways to control central cellular processes. This cooperative signaling from integrins and other environmental cues such as soluble growth factors and cytokines is mandatory for orderly cell cycle progression (18,19) and involves multiple mechanisms, among them induction of cyclins D and A as well as sequestration and down-regulation of the cyclindependent kinase inhibitors (CKI) p21 and p27 (18). Integrins control these events in part via modulation of growth-regulatory signaling pathways such as the canonical Ras-Raf-MEK-ERK cascade (20). Based on (i) the ability of Gal-1 to regulate growth in activated T-cells and (ii) its carbohydrate-specific interaction with cell surface receptors, we explored the ability of Gal-1 to inhibit cell growth in epithelial tumor cell lines of different origin and delineated the underlying signal transduction events in unprecedented detail. polyvinylidene difluoride membranes and Renaissance chemiluminescence detection reagent from PerkinElmer Life Sciences); [␥-32 P]ATP from Amersham Biosciences; Raf-1 Raf-binding domain-agarose from Upstate Biotechnologies (Hamburg, Germany); effectene transfection reagent from Qiagen (Hilden, Germany); and fibronectin from EMP Genetech (Denzlingen, Germany). Antibodies used for immunoprecipitation and immunoblotting were from the following sources: mouse antibodies to ␣5, ␤1, cyclin E, cyclin D, p21, p27, and Sp1 from BD Transduction Laboratories (Heidelberg, Germany); rabbit antibodies to Cdk2, Cdk4, and Sp3, mouse anti-Cdc25A and anti-phosphotyrosine, and goat anti-Sp1 from Santa Cruz Biotechnology Inc. (Santa Cruz, CA); mouse anti-cyclin A from Upstate Biotechnology (Lake Placid, NY); mouse antibodies to phosphothreonine and phosphoserine from Sigma; rabbit antibodies to ERK1/2, phospho-ERK1/2, and phospho-MEK from Cell Signaling (Beverly, MA); mouse anti-pan-Ras and anti-p21 from Oncogene (Darmstadt, Germany). All secondary antibodies were from Dianova (Hamburg, Germany). Antibodies used for flow cytometry and supershift assays were as follows: mouse antibodies to ␣1, ␣2, ␣3, ␣4, ␣6, and ␣ V subunits and goat anti-␣5␤1 from Chemicon (Hofheim, Germany); mouse anti-␣5 R-phycoerythrin and anti-␤1 R-phycoerythrin from Cymbus Biotechnology (Hampshire, UK); secondary goat anti-mouse fluorescein isothiocyanate-labeled antibody from Sigma. Phorbol 12-myristate 13-acetate and PD 98059 were obtained from Merck Biosciences GmbH (Bad Soden, Germany).
Cell Culture-The human hepatoma cell lines HepG2 and Sk-Hep-1, colonic carcinoma cell lines HT-29 and Caco-2, breast carcinoma cell line T-47D, and ovarian carcinoma cell line OV90 were obtained from the American Type Culture Collection (ATCC, Manassas, VA) and cultured as recommended. The melanoma cell line SK-Mel-13 was from J. Eberle (Charité-Universitätsmedizin, Berlin, Germany). BON cells were a generous gift from C. M. Townsend (Galveston, TX) and were cultured as previously described (21). For experiments on synchronized cultures, cells were serum starved for 36 h, which resulted in G 0 /G 1 accumulation of ϳ85-95% of the cells depending on the cell type, and were then stimulated to re-enter the cell cycle by addition of fetal calf serum-supplemented medium.
Purification and Labeling of Galectins-Galectins-1, -3, and -5 (human homodimeric prototype Gal-1, murine chimera-type galectin-3, and rat monomeric prototype galectin-5) were purified after recombinant production by affinity chromatography on Sepharose 4B to which lactose had been covalently attached via divinyl sulfone activation, and biotinylation was performed under activity-preserving conditions as described (4,22,23). All galectins were routinely checked for purity by one-and two-dimensional gel electrophoresis under reducing and denaturing conditions, gel filtration, and nanoelectrospray ionization mass spectrometry as well as for activity by hemagglutination and solid phase assays as described (24 -26).
Proliferation Assays-Cells were plated in 24-well dishes at a density of 10 4 cells/well. After an attachment period of 16 h, cells were incubated with galectins or vehicle as specified, and viable cells were counted using a hemocytometer.
Stable Transfection of ␣5 Integrin and Separation of Cell Populations with High (␣5) or Low (␣5, ␣3, or ␣V) Integrin Surface Expression-␣5 Full-length cDNA in the vector pRC-CMV (pRC-␣5) was stably transfected using effectene transfection reagent according to the provided protocol. Subsequent selection of stably transfected cells was carried out with 0.8 mg/ml G418. To enrich the fraction of cells with high ␣5 surface expression in pRC-␣5-transfected HT-29 populations, cells were selected using 5 ϫ 10 8 Dynabeads M-450 (Dynal, Hamburg, Germany) coated with an antibody to ␣5. Similarly, Dynabeads M-450 coated with antibodies to ␣5, ␣3, or ␣V were utilized to enrich BON cells with low ␣5 integrin expression. In this case, cells attached to the beads were removed from the cultures to obtain integrin-depleted fractions.
Analysis of Integrin Expression by Flow Cytometry-Evaluation of integrin expression was conducted precisely as previously described (27).
Determination of Gal-1 Binding-Cells were incubated with 125 g/ml biotinylated Gal-1 for 30 min and washed twice with cold phosphate-buffered saline. Binding of labeled Gal-1 was detected by flow cytometry using a commercial streptavidin-fluorescein isothiocyanate conjugate.
Analysis of Cell Cycle Distribution by Flow Cytometry-Cell cycle distribution was determined as previously described (28). At least 10,000 cells stained with propidium iodide were analyzed on a FACS-Calibur utilizing CellQUEST software (BD Transduction Laboratories).
Immunoprecipitation and Western Blot Analysis-Whole cell lysates and nuclear extracts were prepared and Sp1 immunocomplexes were precipitated from 1-mg aliquots of nuclear extracts using 5 g of polyclonal rabbit anti-Sp1 antibody as previously described (29). Immunocomplexes with cyclin-dependent kinase (Cdk) 2 or Cdk4 were collected from 500 g of whole cell lysates using 10 g of polyclonal rabbit anti-Cdk2 or 10 g of anti-Cdk4 antibody as previously described (27,28). Equal amounts of whole cell lysates, nuclear extracts, or immunoprecipitated proteins were analyzed by immunoblotting using the respective primary antibodies, all diluted 1:1000, except for the antiphosphothreonine antibody, which was diluted 1:50.
Determination of Cdk2 and Cdk4 Activity-Cdks were immunoprecipitated as described above. The kinase reaction was started by addition of 30 l of kinase buffer (50 mM HEPES, pH 7.5, 1 mM dithiothreitol, 2.5 mM EGTA, 10 mM MgCl 2 , 10 mM ␤-glycerophosphate), 50 M ATP, 10 Ci of [␥-32 P]ATP/sample, and 1 g of glutathione S-transferase-Rb fusion protein for Cdk4 or H1-histone for Cdk2 kinase assay, allowed to proceed for 30 min, and terminated by boiling the samples in Laemmli buffer. Samples were subjected to 10% SDS-PAGE, and activities of Cdk2 and Cdk4 were determined by autoradiography of the dried gels.
Ras Activation Assay-Cells were lysed in cold magnesium lysis buffer (10 mM MgCl 2 with 1% Nonidet P-40). Equal amounts of protein (0.8 -1 mg) were used to pull down GTP-bound Ras with 10 g of Raf-1 Raf-binding domain-agarose according to the manufacturer's protocol. Active Ras was subsequently detected by immunoblotting.
Transactivation Assays-Transient transfections were performed using the effectene transfection reagent, according to the manufacturer's protocol; luciferase activity was analyzed as previously described (29). To correct for transfection efficiency, a Renilla luciferase construct pRL-TK (Promega) was cotransfected. All experiments were performed in hexaplicates and repeated at least three times.
Electrophoretic Mobility Shift Assay-Nuclear extracts were prepared using a non-ionic detergent method, and electrophoretic mobility shift assays were performed as described previously (29). In brief, equal amounts of nuclear extracts (5-10 g) were incubated either with the radiolabeled oligonucleotides comprising the Ϫ555/Ϫ512 region (agcctcggcggggcggctcccgccgccgcaaccaatggatctcc) of the human p27 promoter or the CAGA boxes-containing oligonucleotide (tcgagagccagacaaaaagcc-agacatttagccagacac). For competition experiments, nuclear extracts were preincubated with 100 ϫ molar excess of the unlabeled Ϫ555/Ϫ512 and CAGA boxescontaining oligonucleotides or the following competitor oligonucleotides: Sp1 consensus, attcgatcggggcggggcgagc; Sp1 mutant, attcgatcggttcggggcgagc; Ϫ555/Ϫ512 oligonucleotides containing site mutations within the distal Sp1 site mSp-I, agcctcggcgggatggctcccgccgccgcaaccaatggatctcc, and the proximal Sp1-binding site mSp-II, agcctcggcggggcggctcctaccgccgcaaccaatggatctcc. For supershift experiments, extracts were preincubated for 30 min with 2 g of antibodies against Sp1 and Sp3 prior to the addition of radiolabeled probes.
Statistical Analysis-Statistical analysis was performed by two-tailed Student's t-test for paired observations or one-way analysis of variance using GraphPad statistical software (GraphPad Software Inc., San Diego, CA). All data are expressed as mean Ϯ S.E. unless otherwise indicated.

Gal-1 Inhibits Growth of Epithelial Tumor Cells in a Carbohydrate
dependent Manner-To study growth-regulatory effects of Gal-1 on transformed epithelial cells, a panel of human carcinoma cell lines was treated with Gal-1 and cell numbers were recorded. Of eight cell lines analyzed, none revealed an increase in cell proliferation. In contrast, six were growth inhibited, and two (Caco-2 and HT-29) showed no response (Fig. 1A). Significant growth inhibition started at 25 g/ml Gal-1 and was consistently achieved at 50 or 100 g/ml depending on the cell line investigated. Gal-1 reduced proliferation in a time-dependent manner, as documented by growth curves of HepG2 and BON cells (Fig. 1B).
Based on the hypothesis that Gal-1 might functionally interact with ␣5␤1 integrin and thereby modify its signaling, we examined whether Gal-1-mediated growth inhibition could be prevented by preincubation with the ␣5␤1 ligand fibronectin. Although fibronectin alone did not alter cell proliferation in HepG2 cells, it abrogated Gal-1-mediated inhibition of cell growth (Fig. 2C). Similarly, a neutralizing antibody to ␣5␤1 integrin completely prevented cell growth inhibition by Gal-1 (Fig. 2D). In contrast, preincubation with either laminin or vitronectin or addition of neutralizing antibodies to ␣3␤1 and ␣V␤1 integrins did not alter Gal-1-mediated growth inhibition (data not shown).
To provide conclusive evidence for a functional role of ␣5␤1 integrin in Gal-1-mediated growth inhibition, both ␣5-deficient colon cancer cell lines were stably transfected with a cDNA encoding the ␣5 integrin subunit. Expression of ␣5 integrin in the transfected cells was confirmed by immunoblotting (Fig. 3A) and flow cytometry (data not shown). We predicted emergence of strong reactivity to Gal-1. Indeed, ␣5-transfected HT-29 (HT-29␣5) and Caco-2 (Caco-2␣5) cells bound substantially increased amounts of biotinylated Gal-1 (Gal-1-bio), relative to mock-transfected controls (Fig. 3B). Because ␤1 integrin surface expression remained unchanged (Caco-2␣5) or only slightly increased (HT-29␣5) (data not shown), increased Gal-1 binding was unlikely to result from better access to the ␤1 subunit. In support of ␣5 integrin being a Gal-1 interaction partner, binding of labeled Gal-1 to HT-29␣5 cells was substantially reduced when cells were preincubated with either fibronectin or a neutralizing antibody to ␣5␤1 integrin (Fig. 3B). Functionally, Gal-1 treatment for 96 h resulted in significant growth inhibition in ␣5-transfected clones, whereas growth of mock-transfected con-trols was unaltered (Fig. 3C). Conversely, no Gal-1-mediated growth inhibition occurred in a subset of BON cells isolated from the parental population on the basis of low ␣5 integrin surface presentation, whereas FIGURE 2. Gal-1 functionally interacts with the ␣5␤1 integrin. A, the integrin expression pattern was determined by flow cytometry using the indicated integrin-specific antibodies. B, whole cell lysates from tumor cell lines were analyzed by immunoblotting using specific monoclonal antibodies against ␣5 and ␤1. The lower band recognized by the ␤1-specific antibody likely represents pre-␤1 integrin. C and D, HepG2 cells were treated with 125 g/ml Gal-1 following 30 min of incubation with 150 g/ml soluble fibronectin (panel C, FN) or an ␣5␤1-blocking antibody (panel D, ␣5␤1 Ab), and viable cells were determined after 72 h. Data represent mean Ϯ S.E. of three independent experiments, each conducted in triplicate (***, p Ͻ0.001).

FIGURE 3. Expression of ␣5 sensitizes Caco-2 and HT-29 cells to Gal-1-mediated growth inhibition.
A, analysis of ␣5 integrin expression in stably transfected Caco-2 and HT-29 cells as well as their respective mock controls by immunoblotting using whole cell lysates and a monoclonal antibody against ␣5. B, HT-29␣5/1 and Caco-2␣5/1 cells and respective mock controls were incubated with 125 g/ml biotinylated Gal-1 (Gal-1-bio) for 10 min. HT-29␣5/1 cells were also incubated in the presence of 125 g/ml fibronectin (FN) or ␣5␤1-specific antibody (␣5␤1 Ab). After washing three times with phosphate-buffered saline to remove free labeled galectin, surface-bound biotinylated Gal-1 was determined by flow cytometry using a streptavidin-fluorescein isothiocyanate conjugate. Representative histograms are shown. C, stably ␣5-transfected HT-29 and Caco-2 cells and respective mock controls were treated with 100 and 200 g/ml Gal-1 for 96 h, and the number of viable cells was determined. Data represent mean Ϯ S.E. of three to four experiments, each conducted in triplicate (***, p Ͻ0.001). D, a BON cell subpopulation with low ␣5 surface expression was isolated based on the inability of the cells to attach to magnetic beads coated with an antibody to the ␣5 integrin subunit (␣5-depleted). Reduced ␣5 expression (left panel) and conserved ␤1 expression (right panel) of this subpopulation in comparison to unselected control cultures was confirmed by flow cytometry. Cells were then subjected to Gal-1 treatment for 96 h, and the number of viable cells was determined (lower panel). As a control, populations with reduced ␣3or␣V integrin expression were similarly selected and subjected to 96 h of Gal-1 treatment.
Gal-1 expectedly reduced cell numbers in the parental cultures (Fig.  3D). In contrast, subpopulations selected based on low ␣3 or low ␣V integrin expression did not differ from parental cells with respect to Gal-1 responsiveness (Fig. 3D). Taken together, these experiments indicated that expression of cell surface ␣5␤1 integrin was required for Gal-1-mediated growth inhibition.
Gal-1 Delays G 1 Cell Cycle Progression-To gain insight into the mechanism of Gal-1-mediated growth inhibition, we analyzed Gal-1dependent changes in cell cycle distribution by flow cytometry. HT-29␣5 cells were used throughout all further studies, because cell cycle synchronization was most reproducibly achieved in this cell system. HT-29␣5 cell populations were synchronized in G 0 /G 1 to monitor their progression through the individual cell cycle phases. Synchronization was achieved by serum starvation, which routinely retained ϳ90 Ϯ 1% of cells in the G 0 /G 1 phase (Fig. 4A). Cells were then stimulated to resume cycling by addition of fetal calf serum, and cell cycle progression was compared between Gal-1-treated cells and time-matched controls for up to 24 h.
Following release from serum starvation, control cell populations exited the G 1 phase at 16 h, proceeded through the S phase, and reached G 2 /M by 24 h. Compared with controls, a significantly increased fraction of Gal-1-treated cells remained in the G 1 phase at 16, 20, and 24 h and barely started to enter the S phase at 20 h, suggesting a G 1 cell cycle delay (Fig. 4, A and B). A very similar retention in the G 1 fraction occurred in Gal-1-treated BON and T-47D cell populations at 20 h following release from synchronization (Fig. 4C). Of particular note, Gal-1 did not increase the fraction of cells with subdiploid DNA content, suggesting that the antiproliferative effect did not result from induction of apoptosis.
Gal-1 Inhibits Cdk2 Activity via Induction of p21 and p27-To unravel the molecular mechanisms underlying Gal-1-mediated G 1 retention, we next assessed the activity of Cdk that control G 1 /S cell cycle progression, i.e. Cdk4 and Cdk2. Again, HT-29␣5 cells were synchronized and released in the presence or absence of Gal-1. In HT-29␣5 cells, basal as well as serum-stimulated Cdk4 activity was low and was not modulated by Gal-1 addition (Fig. 5A). Furthermore, we did not detect a significant Gal-1-dependent reduction of cyclin D 1 or Cdk4 expression (Fig. 5A). In sharp contrast, Cdk2 activity consistently increased with G 1 /S progression, and this induction was substantially suppressed by Gal-1 treatment (Fig. 5B).
To further investigate the mechanism of Cdk2 inhibition by Gal-1, we analyzed main regulatory components of Cdk2 complexes, i.e. cyclin E, cyclin A, and Cdk2 (Fig. 5B). The expression of cyclin E was transiently induced in both control cell populations and Gal-1-treated cultures to a comparable extent. In contrast, the subsequent rise of cyclin A expression, which occurs in a Cdk2/cyclin-E-dependent fashion, was diminished by Gal-1 at 16 and 20 h.
We next focused on regulatory molecules capable of inhibiting Cdk2 activity and determined effects of Gal-1 on the expression of the Cdk inhibitors p21 and p27 (Fig. 5B). p21 expression was low in serumstarved cells at the time of release, rapidly increased with serum stimulation (data not shown), and subsequently remained at moderate steady-state levels. Gal-1 treatment increased the cellular p21 content relative to time-matched controls at 12, 16, and 20 h (Fig. 5B). Con-  Immunoblot analyses for detection of Cdk4 and Cdk2 complex components were conducted on aliquots of whole cell lysates utilized in the respective kinase assays. C, immunoblot analyses for the CKIs p21 and p27 in Cdk2 immunoprecipitates of whole cell lysates obtained from HT-29␣5 cells. An additional immunodetection was carried out for Cdk2 as control to ascertain the quality of immunoprecipitation.
versely, p27 expression was high in serum-starved cultures and subsequently declined as control cells re-entered the cell cycle. Gal-1 treatment substantially increased the cellular p27 content relative to untreated controls (Fig. 5B). In contrast, the expression of other cell cycle-regulatory proteins such as Cdc25A phosphatase and Cdk7 (data not shown) remained unchanged, excluding unspecific effects of either Gal-1 treatment or the synchronization procedure.
The relevance of p21 and p27 for Cdk2-dependent G 1 inhibition is based on their ability to bind to and inactivate Cdk2/cyclin complexes. Therefore, we determined the composition of immunoprecipitated Cdk2 complexes in cells treated as described above (Fig. 5C). Comparable amounts of Cdk2 were detected in all samples, confirming that equal quantities of Cdk2 had been precipitated. However, Cdk2 complexes from Gal-1-treated cultures contained markedly elevated amounts of p21 and p27 compared with their time-matched controls. Thus, Cdk2 inhibition resulted from increased association with p21 and p27 CKIs.
Gal-1 Increases p21 and p27 Transcription and p27 Protein Stability-Cellular CKI content is tightly controlled via regulation of both CKI transcription and CKI degradation. To elucidate which of these mech-anisms contributes to Gal-1-mediated increase in cellular CKI content, we examined effects on CKI protein stability and on CKI promoter activity in HT-29␣5 cells. To determine the CKI protein half-life, protein synthesis was blocked by cycloheximide, and p21/p27 protein contents were subsequently monitored by immunoblotting (Fig. 6, A and B). p21 content decreased rapidly and equally in Gal-1-and vehicle-treated cell populations with a calculated t1 ⁄ 2 of ϳ1 h (Fig. 6A). In contrast, p27, which was rapidly degraded in untreated cultures (t1 ⁄ 2 ϭ 3 h), was stabilized by Gal-1 treatment (t1 ⁄ 2 ϳ8 h; Fig. 6B).
To further resolve the question whether Gal-1 might also regulate CKI transcription, luciferase reporter constructs containing the fulllength p21 or p27 promoter were utilized in transient transfection experiments. In these assays, Gal-1 treatment induced transcription of the p27 reporter construct in randomly cycling cells (Fig. 6D) as well as in synchronized populations (3.2-fold, detailed data not shown). Similarly, Gal-1 resulted in a 2-fold stimulation of the p21 reporter construct, which was comparable with induction by phorbol 12-myristate 13-acetate, a well characterized transcriptional activator of p21 (Fig.  6C). Thus, Gal-1 treatment stimulated the transcription of both p21 and p27. The latter result is of particular interest, prompting functional promoter analysis for p27.
Mapping of the Gal-1-responsive Element in the Human p27 Promoter-To identify the promoter elements that convey Gal-1 responsiveness, deletion analysis of the 5Ј-flanking region of the human p27 promoter was performed (Fig. 7A). These studies revealed that the region spanning Ϫ3568 to Ϫ549 relative to the translational start site was dispensable for the ability of Gal-1 to stimulate p27 promoter activity. In contrast, removal of an additional 38 nucleotides from Ϫ549 to Ϫ511 completely abolished Gal-1-stimulated p27 promoter activity. These results suggest that the p27 promoter region spanning Ϫ549 to Ϫ511 comprised essential regulatory elements for Gal-1 responsiveness. In HT-29 wild-type cells, Gal-1 failed to stimulate p27 promoter activity (Fig. 7A, lower panel). A survey of the Ϫ549/Ϫ511 region of the human p27 gene promoter (Fig. 7B) located two GC-rich boxes as putative binding sites for Sp-like factors (GC-I and GC-II) and a CCAAT box, which are adjacent to each other and conserved between the human and mouse promoters (33,34). To elucidate the functional relevance of these cis-acting elements for Gal-1-induced p27 transcription, a series of p27 promoter mutants carrying mutations in the GC boxes and the CCAAT box was analyzed (Fig. 7C). Mutation of the CCAAT box did not alter responsiveness to Gal-1 treatment. In contrast, functional inactivation of either one of the GC boxes completely abolished Gal-1-induced p27 promoter activity.
Gal-1 Stimulates Binding of Sp1 and Sp3 to the Gal-1-responsive Element Ϫ549/Ϫ511 of the Human p27 Promoter-To identify nuclear proteins that bind to the Ϫ549/Ϫ511 region of the p27 promoter, we performed gel shift assays using the 32 P-labeled promoter sequence from Ϫ555 to Ϫ512 and nuclear extracts obtained from HT-29␣5 cells (Fig. 8A). Two specific DNA-protein complexes were reproducibly identified. Compared with untreated controls (lane 1), treatment with Gal-1 (12 h) increased the intensity of the low mobility complex and, to a lesser extent, of the high mobility complex (lane 2). These complexes are likely to contain Sp-like factors, as they completely disappeared when competed with excess of either unlabeled Ϫ555/Ϫ512 (lane 3) or the Sp consensus (lane 4) oligonucleotide. This competition was specific, because a mutant Sp consensus oligonucleotide did not displace the complexes (lane 5). In agreement with this interpretation, the low mobility complex was supershifted by preincubation of nuclear extracts with an antibody against Sp1 (lane 6), and addition of an antibody against Sp3 abrogated the high mobility complex (lane 7). Combination Representative immunoblots are shown (upper panels). Blots were reprobed with an antibody to ␤-actin to exclude differences in protein loading (lower rows). Lower panels show the results of quantification of p21 and p27 protein contents of eight independent experiments expressed as mean percentage Ϯ S.E. of respective protein content at the onset of cycloheximide addition (designated as control) (*, p Ͻ0.05; **, p Ͻ0.01). C and D, HT-29␣5 cells were transiently transfected with either the p21-luc (C) or the p27-luc (D) reporter construct and treated for 48 h with indicated doses of Gal-1 for p27-luc and 200 g/ml Gal-1 as well as 1 nM phorbol 12-myristate 13-acetate for p21-luc. Luciferase activity was normalized for transfection efficiency. Bars represent luciferase activity as the mean fold Ϯ S.E. of vehicle-treated controls obtained from four to six experiments, each conducted six times (*, p Ͻ0.05; **, p Ͻ0.01; ***, p Ͻ0.001). NOVEMBER 4, 2005 • VOLUME 280 • NUMBER 44 of Sp1 and Sp3 antibodies almost completely supershifted both complexes (lane 8). Transient transfection assays demonstrated that both GC boxes within the Ϫ549/Ϫ511 element of the human p27 promoter were required for responsiveness to Gal-1 treatment. In line with these observations, oligonucleotides containing functionally inactivated GC boxes did not inhibit Sp1 and/or Sp3 complex formation in competition experiments (lanes 11 and 12). Finally, treatment of HT-29␣5 cells with Gal-1 for various periods of time revealed that Gal-1 time dependently induced the binding of Sp1 and Sp3 to the Ϫ555/Ϫ512 element of the p27 promoter (Fig. 8B, lanes 16 -22). This increase in DNA binding was specific and selective, as Gal-1 treatment did not change the relative amounts of Smad transcription factors bound to the 32 P-labeled CAGA boxes-containing oligonucleotide (Fig. 8B, lanes 23-25). Furthermore, Gal-1 had no effect on Sp1 and Sp3 DNA binding to the Ϫ555/Ϫ512 element of the p27 promoter in parental HT-29 cells lacking expression of ␣5 integrin (Fig. 8A, lanes 13-15).

Galectin-1-mediated Growth Inhibition
Gal-1 Stimulates Sp1-and Sp3-transactivating Capacity-To assess whether enhanced DNA binding of Sp1 and Sp3 resulted from a Gal-1induced change of cellular or nuclear Sp1 and Sp3 content, we determined protein levels of both transcription factors in whole cell lysates and nuclear extracts by immunoblotting. Compared with untreated controls, Gal-1 treatment of HT-29␣5 cells up to 24 h did not change the Sp1 and/or Sp3 content of whole cell or nuclear protein lysates (Fig. 9A).
Therefore, we explored whether Gal-1 influenced the transactivation capacity of Sp1 and Sp3, using appropriate Gal4 reporter assay systems. For these experiments, we cotransfected constructs containing the transactivation domain of the Sp1 or Sp3 protein linked to the Gal4 DNA-binding domain along with a luciferase construct containing five Gal4-binding sites as a reporter (Fig. 9B). Gal-1 treatment significantly induced the transactivation capacity of both Sp1-Gal4 and Sp3-Gal4. These data confirmed the regulation of Sp1 and Sp3 by Gal-1 and suggest that, in addition to the enhanced binding of both transcription factors shown in gel shift assays, an induction of their transactivation capacity represents a regulatory mechanism through which Gal-1 stimulates Sp1/Sp3-dependent p27 transcription.
Gal-1 Inhibits Threonine Phosphorylation of Sp1-The transactivating properties of Sp1 are modulated by several posttranslational modifications, including phosphorylation on tyrosine, serine, and threonine residues. To assess whether Gal-1-mediated stimulation of Sp1 transactivation was because of changes in the phosphorylation status, Sp1 was immunoprecipitated from Gal-1-treated cells, and the abundance of phosphotyrosine, -serine, and -threonine was determined by immunoblotting using phospho-specific antibodies (Fig. 9C). To ensure that equal amounts of Sp1 had been precipitated, the blots were subjected to immunoblotting with an Sp1 antibody. The levels of tyrosine and serine phosphorylation of Sp1 remained unchanged up to 18 h of Gal-1 exposure. In contrast, the relative abundance of phosphothreonine was substantially reduced by Gal-1, suggesting that a Gal-1-mediated decrease in threonine phosphorylation of Sp1 accounted for the observed increase of Sp1 transactivation.
To establish the functional relevance of Gal-1-induced reduction of Sp1 threonine phosphorylation with respect to p27 promoter regulation, the full-length p27 reporter construct was cotransfected with either wild-type Sp1 (CMV-Sp1) or Sp1 harboring a mutation at the threonine phosphorylation site (CMV-Sp1.mThr 453 /mThr 739 ), and the ensuing effects of Gal-1 treatment on p27 promoter activity were determined. Upon cotransfection of wild-type Sp1, Gal-1 treatment reproduced the stimulation of the p27 promoter that was observed with endogenous Sp1. As predicted from our hypothesis, cotransfection of the threonine phosphorylation-deficient mutant per se stimulated p27 promoter activity when compared with wild-type Sp1. Importantly, p27 promoter activity had lost Gal-1 responsiveness under these conditions, supporting the functional involvement of Sp1 threonine phosphorylation in p27 regulation by Gal-1.
Gal-1 Inhibits the Ras-MEK-ERK Signaling Pathway-To link the Sp1-dependent induction of p27 to upstream signaling events initiated by Gal-1, we focused on signals that are (i) elicited or modulated by the ␣5␤1 fibronectin receptor and (ii) capable of modifying Sp1 transcriptional activity via threonine phosphorylation. As ERK1/2 have been proposed to directly phosphorylate Sp1 on threonine residues, this profile applies to the ERK signaling module, which is also critically implicated in G 1 /S cell cycle progression.
Initially, ERK phosphorylation was determined using a phospho-specific ERK antibody as an indicator of kinase activity (Fig. 10A, upper  panel). When synchronized cultures were analyzed, activity was low in serum-starved cells but substantially increased upon serum stimulation in control cells. In contrast, Gal-1 treatment suppressed ERK activity to the level of serum-starved controls at all time points analyzed. Gal-1induced differences in ERK activity were due to activity rather than  12). To identify distinct components of DNAprotein complexes in supershift experiments, nuclear extracts were incubated with antibodies against Sp1 and Sp3 prior to incubation with the 32 P-labeled probe (lanes 6 -8). To confirm that Gal-1 regulation of Sp1 DNA binding required the presence of ␣5 integrin, nuclear extracts from parental, ␣5-integrin-deficient HT-29 cells (HT-29 WT) were treated with Gal-1 or vehicle, and Sp1/ Sp3-DNA complexes were detected in gel shift assays (lanes 13-15). B, nuclear extracts from HT-29␣5 cells treated with Gal-1 or vehicle for the indicated time periods were incubated with 32 Plabeled Ϫ555/Ϫ512 oligonucleotide, and Sp1/ Sp3-DNA complexes were detected in gel shift assays (lanes 16 -22). As controls, aliquots from the same nuclear extracts were incubated with the 32 P-labeled CAGA boxes-containing oligonucleotide and competed with 100-fold molar excess of cold oligonucleotide to ensure specificity (lanes [23][24][25]. Experiments at each condition were repeated at least three times from independently prepared nuclear extracts; representative gels are shown. NOVEMBER 4, 2005 • VOLUME 280 • NUMBER 44 regulation of expression, because immunoblots using a regular ERK1/2 antibody documented equal cellular ERK content. Furthermore, ERK1/2 activity remained entirely unchanged when ␣5-deficient HT-29 cells (mock controls) were exposed to Gal-1 treatment (data not shown), suggesting that the modulation of ERK activity was due to interaction of Gal-1 with glycans of the ␣5␤1 fibronectin receptor. In line with the inhibition of ERK1/2 activity, Gal-1 also reduced the activity of the immediate upstream ERK1/2 kinase MEK, as determined based on a reduction in the extent of phosphorylation (Fig. 10A).

Galectin-1-mediated Growth Inhibition
To further clarify whether the Gal-1-dependent inhibition of MEK and ERK activity resulted from impaired proximal signaling input, we examined the activity of the small GTPase Ras, which represents a prototype upstream activator of ERK signaling and is also regulated by ␣5␤1. We utilized the Ras-binding domain of Raf to selectively precipitate GTP-bound, active Ras and subsequently detected active Ras by immunoblotting using an antibody that recognizes H-Ras, N-Ras, and K-Ras homologues (Fig. 10B). Again, synchronized cultures were studied, the time course being modified to accommodate an early time point at 6 h of treatment (Fig 10B). This time course revealed an increase of Ras activity in response to serum stimulation, which peaked at 20 h in control cells and was virtually abolished in Gal-1-treated cells. Thus, Gal-1 potently inhibited the activity of three core components of the Ras-Raf-MEK-ERK signaling cascade.
To further corroborate the link between ERK inhibition and p27 induction, the specific MEK inhibitor PD 98059 was utilized to block ERK1/2 activity. This artificial signal should mimic Gal-1 activity if our concept is correct. A concentration of 20 M PD 98059 was required to inhibit ERK phosphorylation to an extent comparable with that with Gal-1 (Fig. 10C). In this concentration range, PD 98059 also (i) reduced threonine phosphorylation on Sp1, (ii) stimulated the binding of Sp1 and Sp3 to the 32 P-labeled Ϫ555 to Ϫ512 sequence of the p27 promoter, (iii) increased cellular p27 content determined by immunoblotting, and (iv) resulted in G 1 cell cycle inhibition (Fig. 10D). Thus, inhibition of ERK1/2 activity was sufficient to activate Sp1 transcription factors and to induce p27, thereby precisely resulting in the biological effects elicited by Gal-1 treatment.
Gal-1-dependent Induction of p27 Requires Inhibition of the Ras-MEK-ERK Signal Transduction Pathway-To further corroborate the functional relevance of Ras, MEK, and ERK inhibition for Gal-1-mediated p27 induction, the ability of constitutively active Ras (K-RasV12) and MEK (MEK S222E) variants to prevent the Gal-1-mediated induction of the p27 luciferase reporter construct was tested (Fig. 10E). Importantly, cotransfection of either K-RasV12 or MEK S222E constructs counteracted Gal-1-dependent p27 promoter activation but had no measurable effect on basal promoter activity. Conversely, transfection of a dominant negative MEK variant (MEK S222A), as well as the presence of PD 98059, mimicked Gal-1-mediated p27 promoter induction (Fig. 10E).
In a second approach, we explored whether inhibition of the Ras-MEK-ERK pathway also accounted for Gal-1-mediated induction of Sp1-transactivating capacity. Again, a construct containing the transactivation domain of the Sp1 protein linked to the Gal4 DNA-binding domain (Sp1-Gal4) and a luciferase construct containing five Gal4binding sites as a reporter (Gal4-luc) were utilized (Fig. 10F). Cotransfection of either K-RasV12 or MEK S222E constructs completely blocked the Gal-1-mediated increase of Sp1 transactivation capacity.
Taken together, these experiments establish Gal-1-dependent inhibition of Ras, MEK, and ERK as functionally required signaling events that link Gal-1 binding to the ␣5␤1 fibronectin receptor on the cell FIGURE 9. Gal-1 inhibits threonine phosphorylation of Sp1 and increases its transactivating capacity. A, Sp1 and Sp3 protein levels in whole cell lysates and nuclear extracts from Gal-1-treated HT-29␣5 cells were analyzed by immunoblotting at the indicated time points. The Sp3 antibody binds two Sp3 isoforms, p84 and p91. B, HT-29␣5 cells were transiently transfected with the Gal4-luc plasmid, which contains five Gal4-binding sites in front of a minimal promoter fused to the luciferase gene and constructs containing either the Sp1 (Sp1-Gal4) or Sp3 (Sp3-Gal4) transactivation domain fused to a Gal4-DNA-binding domain. Cells were subsequently incubated with 200 g/ml Gal-1 or vehicle for 48 h, and luciferase activity was expressed as relative light units (RLU). Bars represent the mean Ϯ S.E. of four experiments, each performed six times (*, p Ͻ0.05). C, HT-29␣5 cells were treated with 200 g/ml Gal-1 for the indicated periods of time, nuclear extracts were prepared, and Sp1 was immunoprecipitated. Immunoblots with the indicated antibodies were performed to determine Gal-1-mediated changes in Sp1 phosphorylation of threonine, serine, and tyrosine residues. To ensure comparable precipitation of Sp1, membranes were subsequently stripped and reprobed with a Sp1-specific antibody. Shown are representative immunoblots of three independent experiments. D, HT-29␣5 cells were cotransfected with plasmids encoding the Ϫ3568-p27-luc reporter construct and either wild-type Sp1 (CMV-Sp1) or mutant Sp1.mThr 453 /mThr 739 (CMV-Sp1mut) and then treated with Gal-1 (200 g/ml) or vehicle for 48 h before luciferase activity was determined. Data represent the luciferase activites as the mean fold increase Ϯ S.E. of vehicle-treated wild-type Sp1transfected cells obtained from three experiments, each conducted six times (*, p Ͻ 0.05, ns, not significant; **, p Ͻ0.01).
surface to Sp1-dependent transcriptional induction of p27 and subsequent cell cycle inhibition.

DISCUSSION
Gal-1 is a carbohydrate-dependent inducer of apoptosis in activated T-cells. The current study has extended the biological function of Gal-1 to growth control in human epithelial cancer cells, where Gal-1 acts as a modifier of integrin-dependent cell cycle regulation. In addition, we have mapped the chain of signaling events responsible for this type of Gal-1-mediated growth inhibition. Specifically, we have provided evidence that Gal-1 (i) binds to epithelial tumor cells via its lectin domain in an ␣5␤1-dependent manner, (ii) inhibits the Ras-MEK-ERK signaling cascade, which integrates the mitogenic response to various stimuli, (iii) relieves ERK-dependent suppression of Sp1 transactivation capacity, (iv) induces Sp1-dependent p27 (and p21) gene transcription and protein stability, and (v) inhibits Cdk2 activity and subsequent G 1 /S cell cycle progression, which ultimately leads to growth inhibition (summarized in Fig. 11).
In detail, we have reported that Gal-1 inhibits growth in carcinoma cell lines of different origin, suggesting that the growth-suppressive function of Gal-1 remains conserved across transformed epithelial cells derived from different tissues. Of note, two cell lines were entirely resist-ant to Gal-1 action and thereby allowed the identification of the fibronectin receptor ␣5␤1 as a critical mediator of Gal-1 responsiveness. Our results in ␣5-transfected cell lines and their ␣5-deficient counterparts indicated that ␣5 confers the ability of Gal-1 cell surface binding. Importantly, modified Gal-1 responsiveness occurred without overt changes of ␤1 integrin surface expression but required the presence of the ␣5 subunit. In view of the previously reported binding of Gal-1 to ␤1 integrin subunits (13), Gal-1 likely contacts both integrin subunits with the ␣ subunit conveying specificity to the interaction in the carcinoma cell lines we investigated. This is an intriguing demonstration for selective interaction of Gal-1 with distinct target structures on epithelial tumor cells, despite the ubiquitous abundance of ␤-galactosides on the cell surface. Structural research will have to resolve how Gal-1 recognizes distinct target sites in complex glycans, as was accomplished recently for ganglioside G M1 , the main binding partner for growth regulation on neuroblastoma cells (7).
Both Gal-1 binding and subsequent growth inhibition were abrogated by coincubation with fibronectin or a neutralizing ␣5␤1 antibody, implying a functional interaction of Gal-1 with the ␣5␤1 fibronectin receptor. In previous reports, Gal-1 was demonstrated to interact with ␣7␤1 or ␣1␤1 integrin in mesenchymal cells and ␤1 integrin in T-cells, suggesting that cell type-specific expression patterns of integrins and  NOVEMBER 4, 2005 • VOLUME 280 • NUMBER 44 their glycans contribute to differential effects of Gal-1 in mesenchymal, lymphoid, and epithelial cancer cells (35)(36)(37). In the case of human smooth muscle cells Gal-1 binding did not cross-link ␤1 integrin but nonetheless led to a transient increase in tyrosine phosphorylation of two cytoskeleton-associated proteins and to modulation of cell attachment, possibly due to an effect of Gal-1 on ␤1 conformation (13). Cytoskeletal reorganization also occurs following interaction of the tandem repeat-type lectin galectin-8 with ␣3␤1 and ␣6␤1 integrins on Chinese hamster ovary cells or human endothelial cells (38,39). So far, however, the functional relevance of integrin binding for Gal-1-dependent growth-regulatory effects has not been conclusively demonstrated. Of interest, not an integrin but the pentasaccharide of ganglioside G M1 serves as Gal-1 contact site responsible for growth inhibition of neuroblastoma cells (6,40).

Galectin-1-mediated Growth Inhibition
Intriguingly, the antiproliferative effects of Gal-1 resulted from p27/ p21-mediated inhibition of Cdk2 activity with subsequent retention of cells in the G 1 phase of the cell cycle. This pattern reflects the exact opposite of the well studied events that characterize fibronectin-induced ␣5␤1 signaling (20,41,42). Depending on the type of binding partner the integrin can serve as a more versatile relay station for intracellular signaling than appreciated so far. Importantly, our data ascribe to the glycan chains of the integrin the capacity to modulate effector pathways. Accordingly, minor changes in glycan structure such as introduction of a bisecting GlcNAc into a N-glycan may modulate Gal-1 binding and give the integrin surface receptor a new biological meaning (43,44). Thus, the current study has expanded our understanding of integrin perception and/or interpretation of the cellular environment.
When delineating Gal-1 effector pathways, we first observed a pronounced, time-dependent reduction of Ras activity in Gal-1-treated cells, suggesting a Gal-1-mediated disruption of the well established ␣5␤1 integrin-dependent and shc-mediated stimulation of Ras (20). Second, as a consequence of Ras inhibition, Gal-1 reduced the activity of ERK1/2 as well as the upstream activator MEK. This represents a novel observation, as previous studies in T-cells (45) and mesenchymal hepatic stellate cells (46) both linked Gal-1 to ERK activation, which in the latter case led to a mitogenic response. However, neither of these studies identified the molecular link between ERK activation and Gal-1. Again, the inhibition of ERK activity observed in the current study could be accounted for by inhibition of ␣5␤1 signaling, as ERK activation by integrins represents a well established event in integrin-mediated growth control (20), and ␣5-deficient HT-29 cells failed to down-regulate ERK activity in response to Gal-1. In fact, integrins may utilize multiple pathways for ERK activation, which are attributed to the ␤1 or the ␣5 subunit (20). For example, ␤1 subunits stimulate ERKs via focal adhesion kinase-dependent mechanisms, whereas ␣5 integrin utilizes caveolin, the Src family kinase Fyn, and Ras (20,47). Given that ERK inhibition by Gal-1 occurs subsequent to reduced Ras activity and requires the presence of ␣5, a disruption of caveolin-dependent ERK activation by Gal-1 provides a particularly appealing scenario.
Within the context of ␣5␤1-dependent G 1 cell cycle progression, ERK activation is required to relieve p27-mediated Cdk inhibition (48). In this regard, ␣5␤1 appears to control the proteolytic decay of p27 primarily via posttranslational mechanisms (49). Conversely, the stabilization of p27 by Gal-1 observed in the current study may represent a consequence of ERK inhibition. However, additional mechanisms have been implicated in integrin-dependent control of p27 protein stability. Skp2, an F-box family protein known to facilitate the ubiquitination and subsequent degradation of p27, was found transcriptionally induced in HepG2 cells with ␤1 integrin overexpression (50). At present, we cannot exclude mechanisms different from ERK inhibition that may participate in p27 stabilization at the level of the protein.
In addition to stabilization, however, we have provided a second mechanism of p27 up-regulation by Gal-1, i.e. Gal-1-stimulated p27 gene transcription. Transcriptional control of p27 expression represents a novel finding within the context of Gal-1-mediated biological effects. Fittingly, a recent survey of genes differentially expressed in human umbilical vein endothelial cells upon fibronectin engagement of ␣5␤1 integrin revealed a Ͼ3-fold reduction of p27 mRNA expression (51). These data are in intriguing agreement with the functional antagonism between Gal-1 and fibronectin observed in the current study, which results in transcriptional activation of p27.
According to this concept, the p27 promoter should be responsive to Gal-1 stimulation. This is indeed the case. Using a combination of progressive 5Ј-deletion analysis and systematic mutagenesis of the p27 promoter, we identified the Ϫ549/Ϫ511 region as the Gal-1-responsive element. This region harbors two GC-rich sequences that serve as consensus sites for Sp-like transcription factors. Both Sp consensus sites are required for Gal-1 responsiveness of the p27 promoter, suggesting a coordinate function. Gal-1 treatment stimulated Sp1 and Sp3 binding to this cis-regulatory element and increased Sp1/3 transactivation capacity. Both processes were regulated in an ERK-dependent manner, such that ERK inhibition permits increased Sp1-dependent p27 transcription. Enhanced Sp1 activity did not result from increased nuclear Sp1 concentration or from changes in the Sp1/Sp3 ratio, which have both been shown to affect Sp1-dependent gene transcription (52,53). FIGURE 11. Scheme of Gal-1-mediated growth inhibition via transcriptional induction of the p27 gene promoter. Functional interaction of Gal-1 with the fibronectin receptor ␣5␤1 inhibits the Ras-MEK-ERK signaling pathway, resulting in reduced threonine phosphorylation of Sp1, increased Sp1 transactivation and DNA binding, and consecutive induction of p27 gene transcription. Accumulation of p27 inhibits Cdk2 activity and ultimately results in G 1 cell cycle arrest and growth inhibition. Gal-1-mediated effects on most steps within this cascade can be either blocked (I-) or mimicked (4) as indicated. For reasons of simplicity, Gal-1-induced p21 transcription is not included in the scheme.
In this context, posttranslational modifications such as O-GlcNAcylation and phosphorylation on serine and threonine (or tyrosine) residues provide dynamic and bidirectional control of Sp1 transcriptional capacity (53)(54)(55). In the current study, Gal-1 treatment did not stimulate serine or tyrosine phosphorylation to increase Sp1-dependent p27 transcription but rather reduced Sp1 phosphorylation at threonine residues. As threonine phosphorylation has been shown (i) to occur in response to ERK1/2 activation (56) and (ii) to restrain Sp1 transcriptional and DNA binding capacity (57,58), the Gal-1-mediated reduction of Sp1 phosphothreonine content provides a plausible molecular link between ERK inhibition and transcriptional induction of p27. Furthermore, activation of Sp1 could also offer a molecular link to the transcriptional induction of p21 that occurred in response to Gal-1 treatment. In fact, the p21 promoter represents an established prototype target of Sp1-dependent gene regulation (59).
In summary, we established in the current study a novel antiproliferative mode of action for Gal-1 in epithelial tumor cells. It is based on specific interference with mitogenic ␣5␤1 fibronectin receptor signaling. As both Gal-1 and ␣5␤1 are expressed in non-transformed epithelial tissues, our results may also delineate a pathway that controls epithelial homeostasis using integrin glycosylation to specify target structures. Loss of either Gal-1 or ␣5␤1 integrin expression, which have been described during the process of malignant transformation in a variety of tissues (60,61), would relieve Gal-1 restriction of cell cycle progression and thereby favor epithelial tumor cell growth. Conversely, use of Gal-1 or Gal-1-mimetic compounds could provide an attractive novel biotherapeutic approach in cancers that have conserved expression of the fibronectin receptor.