The Involvement of CD146 and Its Novel Ligand Galectin-1 in Apoptotic Regulation of Endothelial Cells*

Background: CD146 is a glycosylated adhesion molecule involved in the control of vessel integrity. Results: Galectin-1 directly binds to CD146 and this interaction is involved in the control of endothelial cell apoptosis. Conclusion: Galectin-1 is identified as a novel ligand for CD146. Significance: CD146 acts as a decoy-receptor to down-regulate Galectin-1-mediated apoptosis. CD146 is a highly glycosylated junctional adhesion molecule, expressed on human vascular endothelial cells and involved in the control of vessel integrity. Galectin-1 is a lectin produced by vascular cells that can binds N- and O-linked oligosaccharides of cell membrane glycoproteins. Because both CD146 and Galectin-1 are involved in modulation of cell apoptosis, we hypothesized that Galectin-1 could interact with CD146, leading to functional consequences in endothelial cell apoptosis. We first characterized CD146 glycosylations and showed that it is mainly composed of N-glycans able to establish interactions with Galectin-1. We demonstrated a sugar-dependent binding of recombinant CD146 to Galectin-1 using both ELISA and Biacore assays. This interaction is direct, with a KD of 3.10−7 m, and specific as CD146 binds to Galectin-1 and not to Galectin-2. Moreover, co-immunoprecipitation experiments showed that Galectin-1 interacts with endogenous CD146 that is highly expressed by HUVEC. We observed a Galectin-1-induced HUVEC apoptosis in a dose-dependent manner as demonstrated by Annexin-V/7AAD staining. Interestingly, both down-regulation of CD146 cell surface expression using siRNA and antibody-mediated blockade of CD146 increase this apoptosis. Altogether, our results identify Galectin-1 as a novel ligand for CD146 and this interaction protects, in vitro, endothelial cells against apoptosis induced by Galectin-1.

integral membrane glycoprotein of 113 kDa that belongs to the immunoglobulin superfamily with a characteristic V-V-C2-C2-C2 domain structure (1). Using the anti-human CD146 monoclonal antibody (S-Endo1) developed in our laboratory, we have previously demonstrated that CD146 is expressed on human vascular endothelial cells, whatever vessel caliber or anatomical region (2). CD146 is also overexpressed on some tumoral cells, activated T-cell subpopulation and pericytes (3)(4)(5)(6)(7)(8)(9). It has been previously shown that in cultured endothelial cells, CD146 is concentrated at the intercellular junctions and controls inter-endothelial cell cohesion, paracellular permeability, monocyte transmigration, and angiogenesis (10 -15). Moreover, antibody-mediated cross-linking of CD146 in HUVEC (Human Umbilical Vein Endothelial Cells) induces tyrosine phosphorylation of the non-receptor tyrosine kinase p59fyn, of FAK (Focal Adhesion Kinase) and of paxillin, indicating a role of CD146 in the reorganization of the cytoskeleton (16). Altogether, these data identified CD146 as an adhesion molecule of the endothelial junction. However, very few are known about its ligands. It has been reported that CD146 functions as a homophilic adhesion molecule, but this interaction is still debated as a direct biochemical evidence for homophilic interaction with the purified receptor is lacking (5,17,18). Laminin 411, also named ␣4␤1␥1 integrin, that is expressed along the vascular endothelium, was recently discovered as a specific ligand for CD146 (19). In this study, the authors demonstrated that blocking the interaction of mouse CD146 with laminin 411 inhibits T cell adhesion (19).
CD146 is a highly glycosylated protein since glycosylations account for about 35% (1) of its apparent molecular weight, suggesting that this protein could interact with sugar-binding proteins such as lectins. Interestingly, Galectin-1 is a 14 kDa lectin widely expressed in mammalian organism that is produced by vascular, interstitial, epithelial, and immune cells (20). This lectin is known to specifically bind galactoside derivatives of glycoproteins on cell membranes (21). Indeed, various membrane glycoconjugates have been identified as binding partners of Galectin-1 such as ␤1 integrins, CD2, CD3, CD4, CD43, CD45, and GM1 ganglioside. In addition, Galectin-1 binds to a number of extracellular-matrix components in a dose-dependent and ␤-galactoside-dependent manner. For example, laminin and fibronectin, which are highly N-glycosylated, interact with Galectin-1 (22). Because Galectin-1 is a divalent lectin, it can also form complexes with receptor partners participating in cell-matrix recognition. Thus, Galectin-1 is a multifunctional protein involved in a variety of biological activities such as cell-matrix interactions, immune response, inflammatory response, tumorigenesis, and apoptosis of both tumor and immune cells (23,24).
Galectin-1 and CD146 showed intriguing common biological functions. First, Galectin-1 expression knock-down in the zebrafish model results in impaired vascular network formation (25) whereas suppression of CD146 expression in the same model results in poorly developed intersomitic vessels (26), suggesting that both Galectin-1 and CD146 are involved in vascular development in this model. Second, tumor growth in Galectin-1-null mice is markedly impaired because of insufficient tumor angiogenesis (25). In a similar way, tumor growth was inhibited in mice following the administration of monoclonal antibody targeting CD146, and this effect was accompanied by a reduction of tumor blood vessel density (27). In addition, Galectin-1 is involved in melanoma progression (28,29) as well as CD146, which level of expression was correlated with an advance tumor stage in melanoma (30). Third, Galectin-1 was described to activate VEGFR-2 signaling (31,32), and it has been recently shown that VEGFR-2 phosphorylation is dependent of CD146 (33). Fourth, as well as CD146, Galectin-1 showed altered expression and localization in activated and tumor endothelial cells (34). Finally, it has been previously shown that endothelial cells secrete Galectin-1 upon inflammatory conditions (34). Moreover, Galectin-1 can induce apoptosis of T lymphocytes and can regulate bcl-2 expression (35,36). Along the same line, CD146 expression is both up-regulated at the membrane and secreted after TNF␣ treatment (12), and its overexpression leads to activation of survival and anti-apoptotic pathways (37). These data combined with the strong glycosylation of CD146 prompted us to look for a direct interaction between endothelial CD146 and Galectin-1 protein as well as to determine its potential role in endothelial cell functions. Thus, we hypothesized that Galectin-1 could interact with CD146 and that the binding of Galectin-1 to CD146 might modulate endothelial cell apoptosis.

EXPERIMENTAL PROCEDURES
Cells-HUVEC were obtained from different donors and prepared according to a previously described protocol (38). This work has been approved by the local ethical committee N°2 006-A006-16-45 Assistance Publique Hôpitaux de Marseille. Written parental informed consents have been obtained. They were cultured on 0,2% gelatin, in EBM-2 (Endothelial Basal Medium) (Lonza, CC3156) supplemented with EGM-2 (Endothelial Growth Medium) (Lonza, CC4176). When divided, HUVEC were detached using trypsin incubation during 2 min, followed by addition of fetal calf serum, which permits trypsin neutralization.

Transfection of CD146 Complementary DNA in Fibroblastic
Cell Lines-CD146 transfected fibroblasts were obtained as previously described (11). Briefly, CD146 cDNA was inserted in the EcoRI sites of the mammalian expression vector PCI-neo. The fibroblastic L929 cell line was cultured in Dulbecco modified Eagle's medium-Ham F12 supplemented with 10% FCS and was transfected using Lipofectamine (Invitrogen). The transfected cells were selected by resistance to geneticin (800 mg/ml; Invitrogen) and stable clones were obtained by dilution and 3-fold subcloning. Cells were further selected on the basis of CD146 expression determined by Western blotting and flow cytometry using S-Endo1 anti-CD146 antibody.
Enzyme-linked Immunoadsorbent Assay (ELISA)-Human CD146-Fc was obtained as previously described for the mouse CD146-Fc (39). Briefly, CD146-Fc fusion protein contains the extracellular part of human CD146 fused to the Fc moiety of a human IgG1. The coding sequence of human extracellular CD146 was obtained by PCR using specific primers (forward 5Ј-GCTAGCGGAGCAGCCTGCGCCTGAG-3Ј, reverse 5Ј-GGATCCCGCCGGCTCTCCGGCTCCG-3Ј) and cloned in pCDM7 vector. Recombinant CD146-Fc protein was obtained after transient transfection of the COS-7 eukaryotic cell line and was purified as previously described (39). Quality and purity of CD146-Fc were assessed using Coomassie Blue and Western blotting with anti-CD146 antibodies (7A4 clone Biocytex) and HRP-coupled donkey anti-human-Ig antibodies (BD).
Human His-tagged Galectin-1 and Galectin-2 proteins were obtained and purified as previously described (40). Recombinant human ICOS Ligand Fc (Inducible T-cell COStimulator) was provided by Jacques Nunes (UMR 891, Marseille, France) and used as a negative control. 10 g/ml of CD146-Fc proteins were coated on ELISA plate (VWR, 1315801) overnight in PBS at 4°C. After BSA (bovine serum albumin) saturation, distinct concentrations of His-tagged Galectin-1 and Galectin-2 proteins were incubated for 1 h at room temperature. Interactions were revealed by an incubation with a monoclonal antibody directed against His tag (BD, cat 552565) followed by an incubation with donkey anti-mouse IgG-HRP (1:1000, Jackson Laboratories). Revelation was performed by addition of TMB (Tetramethylbenzidine) (Sigma, T0440-100-1). In competition experiments using sugars, galectin incubation was performed in the presence of 0.2 M of either maltose or lactose.
Surface Plasmon Resonance Assay-Surface plasmon resonance measurements were performed on a Biacore T100 apparatus (Biacore GE Healthcare) at 25°C in HBS-EPϩ buffer (10 mM Hepes, pH 7.4, 150 mM NaCl, 3.4 mM EDTA, 0.0005% surfactant P20). In all Biacore experiments, HBS-EPϩ buffer (Biacore GE Healthcare) served as running buffer and sensorgrams were analyzed with Biacore T100 evaluation software. CD146and Ligand Fc proteins were immobilized covalently to carboxyl groups in the dextran layer on a Sensor Chip CM5. The sensor chip surface was activated with EDC/NHS (N-ethyl-NЈ-3-dimethylaminopropyl) carbodiimidehydrochloride and N-hydroxysuccinimide (Biacore GE Healthcare). Proteins were diluted at 10 g/ml in coupling buffer (10 mM acetate, pH 5.2) and injected until the appropriate immobilization level was reached (i.e. 600 to 800 Resonance Units RU). Deactivation of the remaining activated groups was performed using 100 mM ethanolamine pH 8 (Biacore GE Healthcare). Then, a solution of Galectin-1 (1.7 M) was injected for 2 min through CD146-Fc and ICOS Ligand-Fc channels.
For steady state experiments, serial dilutions from 4 nM to 2 M of soluble Galectin-1 were injected for 6 min at a constant flow rate of 40 l/min on dextran layers containing immobilized CD146 recombinant proteins and allowed to dissociate for 1 min before regeneration by a 8 s injection of 500 mM NaCl and 10 mM NaOH buffer. The flow cell containing immobilized ICOS Ligand-Fc proteins was used as a negative control for blank subtraction.
The resulting sensorgrams were analyzed by global fitting using the appropriate model. In this model, the equilibrium dissociation constant K D is obtained by calculating the slope from a pseudo-Scatchard plotting of Req versus Req/C. The curves show the specific signal obtained after subtraction of the background (obtained using immobilized ICOS Ligand-Fc). For solution inhibition experiments, Galectin-1 proteins, at a constant concentration of 50 g/ml, were pre-incubated with increasing concentrations of lactose or maltose (from 0 to 50 mM, Sigma Aldrich) and injected for 2 min at a flow rate of 10 l/min onto the CD146 chips. After each cycle, sensorchips were regenerated by 8 s injection of 500 mM NaCl and 10 mM NaOH buffer at flow rate of 40 l/min.
Assessment of Endothelial Cell Apoptosis-Endothelial cell apoptosis was determined by Annexin-V/7AAD-positive staining. HUVEC were plated in 6-well plates and incubated at 37°C in a 5% CO 2 atmosphere. All experiments were performed between passages 2 and 6. When cells reached 70% of confluence, they were stimulated with Galectin-1 during 24 h. Annexin-V/7AAD co-staining was rapidly performed after HUVEC detachment, with the Annexin-V-FITC/7AAD kit from Beckman Coulter (cat PNIM 3614). Positive cells were analyzed on a FC500 flow cytometer (Beckman Coulter). Apoptosis was also evaluated with a cell death detection ELISA kit (Roche Applied Science), according to the manufacturer's instructions.
Neutralizing anti-CD146 (S-Endo1 clone, Biocytex, cat 5050) monoclonal antibody and its corresponding isotypic control (Biocytex) were used at the final concentration of 5 g/ml in this assay. The caspase-family inhibitor Z-VAD-FMK (Biovision, 1010 -100) was used at the concentration of 2 M.
Statistical Analysis-Data are expressed as mean Ϯ S.E.M. Student tests and Mann Whitney tests for independent samples were used. Statistical analysis was performed with Prism software for Windows (SPSS Software). Differences were considered significant at p Ͻ 0.05.

Galectin-1 Interacts with CD146 in Endothelial Cells-To
investigate whether Galectin-1 interacts with endothelial CD146, we first performed immunoprecipitation of Galectin-1 from endothelial cells with a specific rabbit anti-Galectin-1 serum. Blotting of the resulting precipitate showed the interaction of Galectin-1 with CD146 in HUVEC (Fig. 1A). Pre-immune rabbit serum, used as negative control (IP control, Fig.  1A), was not able to precipitate CD146 and Galectin-1 from endothelial cells. We further performed several immunofluorescence analyses by confocal microscopy to detect a colocalization of CD146 and Galectin-1. Once secreted, Galectin-1 binds rapidly to the surface of many cell surface proteins and to the extracellular matrix. This feature makes difficult the identification of a specific colocalization of CD146 and Galectin-1. Nevertheless a partial colocalization of CD146 and Galectin-1 was observed when CD146-transfected fibroblastic cell line was exposed to exogenous Galectin-1 (Fig. 1B). To test whether endogenous Galectin-1 could also colocalize with endothelial CD146, endothelial cells were stimulated with TNF-␣, described to induce Galectin-1 secretion. We also detected some colocalization areas using endothelial progenitor cells (Fig. 1C).
The Interaction of Galectin-1 with CD146 Is Direct and Specific-To further determine whether the Galectin-1/CD146 interaction is direct, we performed biochemical assays using CD146-Fc (Fc-tagged CD146 extracellular domain) and soluble Galectin-1 (His-tagged Galectin-1). CD146-Fc was purified from supernatant of transiently transfected COS-7 cells. Coomassie Blue staining after SDS-PAGE separation showed that purity of the protein was more than 90% ( Fig. 2A, left panel). Both CD146 and human-Fc moieties are detected by Western blot (Fig. 2A, right panel).
Using ELISA assay, the interaction between CD146 and Galectin-1 was detected in a dose-dependent way, whereas no binding was observed with ICOS Ligand-Fc-coated plates, used as a negative control (Fig. 2B). ELISA assays demonstrated that Galectin-1 binds to CD146, whereas Galectin-2, which is structurally closely related to Galectin-1, does not (Fig. 2B). These data suggest a specific interaction between CD146 and Galectin-1, as Galectin proteins were expressed and purified according to the same protocol.
The surface plasmon resonance (SPR) study confirmed a direct interaction between CD146 and Galectin-1, whereas no binding was observed with ICOS Ligand-Fc (Fig. 2C, upper  panel). Serial concentrations of Galectin-1 were used in SPR assay (Fig. 2C, middle panel) to determine the equilibrium dissociation constant. These data were plotted under a pseudo-Scatchard graph in which the slope of the line corresponds to the K D . The dissociation constant at the equilibrium (K D ) was equal to 3.10 Ϫ7 M (Fig. 2C, lower panel).
The Interaction of Galectin-1 with CD146 Is Dependent of Glycosylation Motifs-CD146 is a highly glycosylated protein as glycosylations account for about 35% of its apparent molecular weight. Moreover, Galectin-1 is able to establish bivalent or multivalent interactions with N-and O-glycans. Thus, we analyzed the glycosylation status of CD146-Fc protein. We investigated the potential shift of molecular weight of CD146-Fc exposed to either neuraminidase or glycosidases. A shift of CD146 molecular weight was observed using N-glycosidase whereas no change was observed using either neuraminidase or O-glycosidase (Fig. 3A). These data show that CD146-Fc protein is mainly N-glycosylated protein.
Galectin-1 can bind to its cognate ligands by regular interaction with ␤-galactoside sugars or by direct protein-protein interaction (40,41). SPR and ELISA studies were performed in the presence of lactose, which is a competitor of Galectin-1 binding to ␤-galactosides. A dose-dependent inhibition of Galectin-1 binding to CD146-Fc was observed, whereas noncompeting maltose had no effect (Fig. 3B, ELISA assay and Fig.  3C, SPR assay). These results indicate that glycosylation motifs are key players of the interaction.
CD146 Plays a Protective Role in Galectin-1-induced Endothelial Cell Apoptosis-To further investigate whether their interaction could affect endothelial cell apoptosis, we performed 7AAD/Annexin-V staining, used as markers for cell integrity and phosphatidylserine surface exposure. Apoptosis of HUVEC was increased after exposure to increasing concentrations of Galectin-1 (0, 2, 10, and 50 g/ml), whereas Galectin-2 had a dose dependent effect even but in a lower extend (Fig. 4A). In addition, the Galectin-1 apoptotic effect on endothelial cells was confirmed using histones release quantification. HUVEC exposed to 5 and 10 g/ml of Galectin-1 released cytoplasmic histones 2.78 Ϯ 0.69 and 4.42 Ϯ 0.19 times more than unstimulated HUVEC (control) (Fig. 4B).
To address the role of CD146 in Galectin-1-induced apoptosis, we did a specific siRNA silencing of CD146 using nucleo-fection of HUVEC. CD146-silenced HUVEC no longer expressed CD146 as demonstrated by the absence of CD146 staining, using both Western blot and flow cytometry, on transfected HUVEC (Fig. 4C). When CD146 was absent, HUVEC were more susceptible to apoptosis upon Galectin-1 treatment (Fig. 4D, left panel). This effect was dependent of the Galectin-1 concentration. Quantification of endothelial apoptotic cells after exposure to 20 g/ml of Galectin-1 during 24 h showed a significant increase of apoptosis in CD146-silenced HUVEC (3.34 Ϯ 0.17-fold/cont, n ϭ 9) when compared with control siRNA-transfected HUVEC (1.95 Ϯ 0.24-fold/cont, n ϭ 5; p ϭ 0.001) (Fig. 4D, right panel).
We further examined the contribution of CD146 in Galectin-1-induced endothelial apoptosis using blocking anti-CD146 antibody (S-Endo1 clone, 5 g/ml) and its corresponding IgG isotypic control, at the same concentration. Endothelial cells were treated with Galectin-1 (20 g/ml) alone or after previous exposure to anti-CD146 antibody. Incubation of HUVEC with the anti-CD146 antibody, before exposure to Galectin-1, significantly increased by 1.60 Ϯ 0.34-fold (n ϭ 5 each; p ϭ 0.032) the percentage of Annexin-V/7AAD positive endothelial cells (Fig.  4E), whereas the addition of anti-CD146 alone did not affect endothelial apoptosis.
We further investigated the mechanisms by which CD146/ Galectin-1 interaction modulates endothelial cell apoptosis. To determine whether this effect is caspase-dependent, the same experiment was performed using the caspase-family inhibitor

DISCUSSION
The present study identifies Galectin-1 as a novel ligand for CD146 and demonstrates that this interaction is involved in the control of endothelial cell apoptosis. We demonstrated that CD146 specifically interacts with Galectin-1 as no binding was detected with Galectin-2. Using real-time analysis by SPR, we determined a K D of 3.10 Ϫ7 M for this interaction, consistent with reported affinity of Galectin-1 interaction to pre-BCR (41). It is known that Galectin-1 can bind either in a sugar-dependent or independent way to their ligands (22). We showed that CD146-Fc protein is sialylated and mainly N-glycosylated and that the lactose inhibits the interaction. Altogether, these results demonstrated that CD146 interacts with Galectin-1 in a sugar-dependent manner.
Many of the proposed regulatory effects of Galectin-1 appeared to involve reversible binding to receptors with affinities in the low M range (42)(43)(44)(45). In addition, Galectin-1 exhibits unique biochemical properties, which make its functional analysis even more complex. This protein contains unpaired cysteine residues in the CRD that, in the absence of carbohydrate binding activity, can form intramolecular disulfide bonds and thereby diminishes its known biological functions (22). Thus, glycan recognition partially protects Galectin-1 from oxidative inactivation and enhances Galectin-1 dimerization (46). Indeed, Galectin-1 exists as a mixture of monomers and dimers at physiological concentrations (45,47). Therefore, the different functions of this lectin can be due to 1) the monomeric or dimeric form of this protein (45,(47)(48)(49)(50), 2) the influence of oxidative versus reducing microenvironments (46), 3) the engagement of Galectin-1 with ligands (51), and 4) the in vivo levels of Galectin-1 in physiological and pathological concentrations.
In this study, we showed that high levels of exogenous Galectin-1 displays, in vitro, a pro-apoptotic effect on endothelial cells (micromolar range concentration), consistent with the pro-apoptotic effect of Galectin-1 described on other cell types. Exogenous Galectin-1 has been previously described to have a biphasic effect on the growth of distinct cell types including endothelial cells. Whereas low concentrations (nanomolar range) induced cell proliferation, high concentrations (micromolar range, equivalent to 20 g/ml) of Galectin-1 seemed to have inhibitory effects (48,(52)(53)(54)(55)(56)(57). Previous studies have shown that endothelial cells secrete Galectin-1 at the ng/ml level upon inflammatory conditions (56). Nevertheless, once secreted, Galectin-1 rapidly binds to the surface of the secreting or neighboring cells or to components of extracellular matrix. As Galectin-1 remains bound to cell or extracellular matrix glycoconjugates, determination of the local concentration of Galectin-1 is basically impossible. Thus, we could hypothesize that, during high inflammatory situation, endothelial cells can be locally exposed to high concentration of Galectin-1.
Beside its role in endothelial cell permeability and angiogenesis, CD146 overexpression has also been associated with sur- vival signals such as Akt phosphorylation and down-modulation of Bad expression (37). Along the same line, Galectin-1 pro-apoptotic effect has already been described for human activated T lymphocytes (36) in which CD45 and CD43 bind to Galectin-1, inducing the activation of AP-1 transcription factor and the down-modulation of bcl-2 proteins (35). However, the mechanisms of Galectin-1 induced apoptosis are still debated since lower and higher amounts of the protein might act on different transmitting receptors, resulting in diverse apoptotic pathways. For instance, apoptosis induced by low levels of Galectin-1, but not high, was inhibited with the pan-caspase inhibitor, zVAD-FMK, which corroborates our results (58).
Interestingly, both Galectin-1 and CD146 endothelial cell expressions are increased in inflammatory conditions (12,59) and we demonstrated that the absence of CD146 renders endothelial cells more susceptible to Galectin-1-induced apoptosis as both down-regulation of CD146 cell surface expression using siRNA and antibody-mediated blockade of CD146 increased Galectin-1-induced endothelial cell apoptosis.
Thus, it is tempting to speculate that the CD146 mucin could be a decoy receptor for Galectin-1, avoiding the Galectin-1 binding to pro-apoptotic receptors. By scavenging Galectin-1, CD146 could also protect Galectin-1 sensitivity to oxidation or enhances Galectin-1 dimer formation, which could display additional or different effects on endothelium. Indeed, dimeric Galectin-1 was shown to induce surface exposure of phosphatidylserine at the plasma membrane of leukemic and promyelocytic cells without inducing apoptosis (36,60). On the other hand, it is conceivable that CD146 cross-linking by dimeric Galectin-1 participates to its role in VEGFR2 signaling involved in survival pathway (33). Altogether, our results identified Galectin-1 as a novel ligand for CD146 and demonstrated that its interaction with Galectin-1 protects endothelium from Galectin-1-induced apoptosis.