Sialylation of the Fas Death Receptor by ST6Gal-I Provides Protection against Fas-mediated Apoptosis in Colon Carcinoma Cells*

The glycosyltransferase, ST6Gal-I, adds sialic acid in an α2–6 linkage to the N-glycans of membrane and secreted glycoproteins. Up-regulation of ST6Gal-I occurs in many cancers, including colon carcinoma, and correlates with metastasis and poor prognosis. However, mechanisms by which ST6Gal-I facilitates tumor progression remain poorly understood due to limited knowledge of enzyme substrates. Herein we identify the death receptor, Fas (CD95), as an ST6Gal-I substrate, and show that α2–6 sialylation of Fas confers protection against Fas-mediated apoptosis. Intriguingly, differences in ST6Gal-I activity do not affect the function of DR4 or DR5 death receptors upon treatment with TRAIL, implicating a selective effect of ST6Gal-I on the Fas receptor. Using ST6Gal-I knockdown and forced overexpression colon carcinoma cell models, we find that α2–6 sialylation of Fas prevents apoptosis stimulated by FasL as well as the Fas-activating antibody, CH11, as evidenced by decreased activation of caspases 8 and 3. We also show that α2–6 sialylation of Fas does not alter the binding of CH11, but rather inhibits the capacity of Fas to induce apoptosis by blocking the association of FADD with Fas cytoplasmic tails, an event that initiates death-inducing signaling complex formation. Furthermore, α2–6 sialylation of Fas inhibits Fas internalization, which is required for apoptotic signaling. Although dysregulated Fas activity is a well known mechanism through which tumors evade apoptosis, the current study is the first to link Fas insensitivity to the actions of a specific sialyltransferase. This finding establishes a new paradigm by which death receptor function is impaired for the self-protection of tumors against apoptosis.

The ability to evade apoptosis is one of the defining characteristics of a malignant tumor cell (1). Escape from cell death can be accomplished through alterations in various cellular components, including dysregulation of oncogenes and tumor suppressors, and mutations in apoptotic and anti-apoptotic signaling machinery. The TNF family of death receptors (TNFRs), 3 including TNFR1, DR4, DR5, and Fas (CD95), rep-resents one category of signaling molecules that is commonly disrupted in human tumors and has been strongly implicated in tumor cell survival (2,3).
The Fas death receptor, like other TNFRs, is a homotrimeric transmembrane receptor that activates multiple intracellular signaling cascades, one of which directs apoptosis. Upon association with activating ligands, Fas undergoes higher order clustering, which facilitates the binding of cytosolic proteins to the Fas cytoplasmic tails. The first protein recruited to the Fas tails is FADD, which binds to Fas through a region known as the death domain. Several other proteins, including procaspase 8 and procaspase 10, are then recruited to the Fas/FADD complex, and together these proteins form the death-inducing signaling complex (DISC). This complex is internalized through clathrin-mediated endocytosis and allows for further DISC formation required for apoptotic signaling (4). This enhanced formation of the DISC leads to the autolytic cleavage and activation of procaspase 8, which goes on to cleave the effector caspase, caspase 3, ultimately resulting in apoptotic endpoints such as membrane blebbing and DNA fragmentation (2).
Diminished Fas expression and activity are well established as mechanisms responsible for the apoptotic resistance of tumor cells. A myriad of studies have reported alterations in the expression of pro-and anti-apoptotic components involved in the Fas pathway such as c-FLIP, BAX, and BCL-2 (5-7). Additionally, in many tumor types there is up-regulation of the endogenous ligand for the Fas receptor, FasL, which is thought to provide a mechanism of self-conservation for tumor cells through FasL-directed killing of tumor-invading immune cells (8). However, despite the extensive research focused on changes in the expression level of Fas and associated signaling molecules, very few studies have investigated molecular mechanisms that alter Fas function independent of variant protein expression, for example, post-translational modifications such as glycosylation.
The Fas receptor is modified with both O-and N-linked glycans (9), although the functional significance of these glycoconjugates in Fas signaling has received minimal attention. It has been known for decades that tumor cells express an altered profile of cell surface oligosaccharides, and in fact there is a specific subset of glycosylating enzymes exhibiting aberrant activity in human cancers. The ST6Gal-I sialyltransferase is one of the enzymes up-regulated in multiple types of cancer (10 -15), and high ST6Gal-I levels are associated with increased metastatic potential (12,(15)(16)(17). ST6Gal-I is responsible for the addition of the negatively charged sugar, sialic acid, in an ␣2-6 linkage to the N-linked glycans of cell surface or secreted glycoproteins. In the current study we have identified Fas as an ST6Gal-I substrate and further determined that ␣2-6 sialylation of Fas inhibits Fas-mediated cell death. These collective results elucidate a novel mechanism by which tumor cells evade apoptosis.

EXPERIMENTAL PROCEDURES
Cell Culture-HD3 colon epithelial cells expressing oncogenic ras were developed as previously reported (18). These HD3 cells were then transduced with lentivirus (purchased from Sigma) containing either shRNA sequence against ST6Gal1 (HD3.sh) or an empty vector (HD3.ev). A pooled population of clones stably expressing shRNA was generated by puromycin selection. Down-regulation of ST6Gal-I expression was confirmed by Western blot (19). SW48 cells were purchased from ATCC. These cells were transduced with either lentivirus containing rat liver ST6Gal-I cDNA (SW.ST6) or lentivirus containing an empty vector (SW.ev). The original ST6Gal-I plasmid was obtained from Dr. Karen Colley (University of Illinois, Chicago), and lentiviral vectors developed from this plasmid were constructed by Dr. John Wakefield (Open-Biosystems, Inc.). SW.ST6 cells represent a pooled population of stably transduced clones, isolated by puromycin selection. Confirmation of expression and functionality of ST6Gal-I has been published previously (20). HD3 cells were maintained in low glucose (1 g/liter) DMEM with 7% FBS and 1% antibiotic/ antifungal containing penicillin G, streptomycin sulfate, and amphotericin B (Invitrogen). SW48 cells were maintained in 1X-L15 medium with 10% FBS and 1% antibiotic/antifungal.
Immunofluorescent Staining-Cells were grown on chamber slides and treated with the Fas-activating antibody, CH11 (Millipore) at 0.5 mg/ml or TRAIL (Biomol International) at 2 mg/ml for 18 h. The reaction was then stopped with ice-cold PBS. The protocol for the Caspases 3 and 7 FLICA Apoptosis Detection Kit was followed as recommended by the manufacturer (Immunochemistry Technologies). The slides were imaged by band-pass filter with excitation at 550 nm and emission at 580 nm for the red fluorescence of cleaved caspases 3 and 7, and UV-filter with excitation at 365 nm and emission at 480 nm for Hoescht staining of the nuclei. For quantification, the percentage of apoptosis was calculated as the number of FLICA-positive cells compared with the total number of cells in three independent fields of view. Statistical analysis was performed by repeated-measure analysis of variance, with p Ͻ 0.05 considered statistically significant.
Lectin Affinity Analysis-Cell lysate protein (1000 g) was incubated overnight with 50 l of immobilized SNA-1 or MAA lectin (EY Laboratories) with rotation at 4°C. The ␣2-6 sialylated proteins complexed with SNA or ␣2-3 sialylated proteins complexed with MAA lectin were collected by brief centrifugation and washed. Sialylated proteins were released from the complexes by boiling in SDS-PAGE sample buffer. The proteins were then resolved by SDS-PAGE and immunoblotted to detect Fas.
For the first dimension, IPG strips were focused using Amersham Biosciences IPGphor II isoelectric focusing unit at 50 A per strip at 20°C. For the second dimension, strips were equilibrated in DTT equilibration buffer containing 6 M urea, 20% glycerol, 50 mM Tris-HCl, pH 8.8, 2% SDS, 65 mM DTT, and trace bromphenol blue for 30 min and in IA equilibration buffer containing 6 M urea, 20% glycerol, 50 mM Tris-HCl, pH 8.8, 2% SDS, 2.5% iodoacetamide, and trace bromphenol blue for 15 min. Strips were then electrophoresed on 12.5% criterion gels (Bio-Rad) at a 100-V constant in the Bio-Rad Criterion gel box. Proteins were then transferred to PVDF membranes and immunoblotted for Fas (Santa Cruz Biotechnology).
Flow Cytometry-Cells were harvested non-enzymatically from tissue culture dishes using Cell Stripper solution (Cellgro). Cells were then resuspended into ice-cold PBS containing 0.2% heat-denatured BSA. For the antibody binding assay: cells were treated with CH11 (Millipore) at 20 g/ml or IgM control for 1 h at 4°C. Cells were then incubated with FITC-tagged secondary antibodies. For the CH11 binding curve, the same experiment was performed using a range of CH11 concentrations, including 1, 10, 20, or 40 g/ml. For Fas expression: cells were treated with FITC-conjugated anti-human CD95 (BC Pharmingen) for 1 h at 4°C. After labeling, cells were washed in PBS/BSA and then analyzed with a FACSCalibur (BD Biosciences) at the University of Alabama at Birmingham Arthritis and Musculoskeletal Center Analytic and Preparative Core Facility.
DISC Immunoprecipitation-DISC components were immunoprecipitated by treating 8.0 ϫ 10 6 cells with 1.0 g/ml CH11 (Millipore) in media at 4°C (control) or 37°C for 30 min. Treatment was stopped by the addition of ice-cold PBS. Cells were centrifuged to remove unbound antibody and washed with ice-cold PBS. Cells were then subjected to lysis buffer containing 50 mM Tris-HCl (pH 7.4) with 1% Triton X-100 and protease inhibitors (Roche Applied Bioscience) on ice for 15 min. Samples were centrifuged, and supernatant was collected and rotated overnight with 40 l of prewashed anti-IgM-conjugated agarose beads (Sigma). Precipitated proteins were released from the complexes by boiling in SDS-PAGE sample buffer. Proteins were resolved by SDS-PAGE and immunoblotted for the Fas-binding protein, FADD (Cell Signaling Technology). The PVDF membrane was then stripped with Restore Western Blot Stripping Buffer (Thermo Scientific) and reprobed with Fas (Santa Cruz Biotechnology) as previously described.
Internalization Assay-Cells were lifted from the tissue culture plastic with trypsin, and trypsin activity was stopped with trypsin inhibitor. Cells were resuspended in media with 1.0 g/ml CH11 (Millipore) at 4°C (control) or were warmed to 37°C for 30 min to activate internalization. Cells were washed and resuspended in 3.7% formaldehyde in PBS and fixed for 15 min at room temperature. Cells were then washed and resuspended in media with 1.25 g/ml anti-mouse IgM Alexa 488 (Invitrogen) at 4°C for 1 h. Stained cells were then washed and resuspended in Vectashield (Vector Laboratories, Burlingame, CA) and plated on a glass microscope slide. Slides were imaged with the Zeiss LSM 710 laser confocal scanning microscope.

␣2-6 Sialylation Confers Protection against Fas-mediated Apoptosis, but Does Not Affect DR4/DR5
Signaling-To evaluate the role of ST6Gal-I-mediated sialylation in regulating tumor cell behavior, we previously developed a cell model system with variant levels of ST6Gal-I expression. HD3 colon carcinoma cells express high levels of endogenous ST6Gal-I as a secondary consequence of oncogenic ras activity (18). This cell line was subsequently stably transduced with shRNA to force down-regulation of ST6Gal-I (19). In the current study, the parental HD3 cell line expressing high endogenous ST6Gal-I (HD3.par) and the shRNA knockdown of ST6Gal-I (HD3.sh) were treated with either the ligand for DR4 and DR5, TRAIL, or the Fas-activating antibody, CH11. Levels of apoptosis were then evaluated by immunoblotting for the apoptotic marker, cleaved caspase 3. As shown in Fig. 1A, when treated with CH11, HD3.sh cells exhibited a markedly greater degree of cleaved caspase 3 than HD3.par cells, suggesting that low levels of ST6Gal-I-mediated sialylation were associated with enhanced Fas-mediated apoptosis. In contrast, equally high levels of activated caspase 3 were observed in cells treated with TRAIL. Thus, variant levels of ST6Gal-I sialylation appeared to modulate apoptosis induced by Fas receptors, but not DR4 or DR5. To ensure that ST6Gal-I knockdown did not alter the expression of death receptors, Western blots of Fas, DR4, and DR5 were performed, which showed equivalent levels in the HD3.par and HD3.sh cells (Fig. 1B).
To further investigate the selective protection by ST6Gal-I against Fas-mediated apoptosis, we again treated the cells with TRAIL or CH11 and monitored apoptosis through immunofluorescence staining with the FLICA reagent, which binds to activated caspases 3 and 7 (cells were counterstained with Hoechst to reveal cell nuclei). HD3.par cells treated with CH11 displayed a small apparent increase in FLICA staining relative to untreated HD3.par cells, however extensive FLICA staining was observed in the CH11-treated HD3.sh cells ( Fig. 2A). In addition, CH11-treated HD3.sh cells exhibited condensed nuclei, and a disruption in the integrity of cell colonies, both indicative of apoptotic activity (Fig. 2B). In contrast to the effects of CH11, treatment with TRAIL induced very strong caspase activation in both the HD3.par and HD3.sh cells (Fig.  2C). Analysis of the percentage of apoptosis can be seen in Fig.  2D. These data suggest that sialylation by ST6Gal-I protects HD3 cells from apoptosis induced by the Fas receptor, but not DR4/DR5, and importantly, that ST6Gal-I knockdown can sensitize cells to Fas-mediated cell death.
␣2-6 Sialylation Confers Protection against Ligand-induced Fas-mediated Apoptosis in Two Cell Models-Although the CH11 antibody is commonly used to stimulate Fas-mediated apoptosis, it was important to examine the effects of ST6Gal-I on apoptosis induced by the biologic ligand for the Fas receptor, FasL. HD3.par, HD3.sh, and additionally, an empty vector control cell line (HD3.ev) were treated with either CH11 or FasL and immunoblotted for cleaved caspase 3. As shown in Fig. 3A, HD3.sh cells displayed higher levels of cleaved caspase 3 in response to both FasL and CH11 as compared with HD3.par and HD3.ev cells, confirming that diminished ␣2-6 sialylation renders cells more susceptible to two independent Fas activators.
Having shown that ST6Gal-I knockdown enhances Fas-mediated apoptosis, we next evaluated whether overexpression of the enzyme in cells with low endogenous ST6Gal-I would inhibit Fas-induced apoptosis. To this end we monitored cell death in the SW48 colon epithelial cell model, which lacks any detectable expression of ST6Gal-I (21). Previously we generated an SW48 cell line that stably expresses ST6Gal-I and verified protein expression and enzyme activity (20). SW48 parental cells (SW.par), which have no ST6Gal-I, SW48 cells with forced expression of ST6Gal-I (SW.ST6), and an SW48 cell line transduced with an empty vector lentivirus (SW.ev) were treated with CH11 or FasL and surveyed for apoptosis (Fig. 3B). These experiments revealed substantially lower levels of activated caspase 3 in SW.ST6 cells compared with SW.par and

ST6Gal-I Confers Protection against Fas-mediated Apoptosis
SW.ev cells. Hence, the combined results in Fig. 3 establish that ST6Gal-I activity confers protection against apoptosis induced by both the Fas-activating antibody, CH11, and also the endogenous ligand, FasL, in two distinct colon carcinoma cell models.
Fas Is a Target for ST6Gal-I ␣2-6 Sialylation-We next sought to determine if the Fas receptor is a direct target for ␣2-6 sialylation by ST6Gal-I, given that this receptor is known to have two possible N-linked glycosylation sites (22,23). To address this, cell lysates were incubated with agarose-conjugated SNA, a lectin that binds specifically to ␣2-6-linked sialic acids. Samples were centrifuged to selectively precipitate ␣2-6 sialylated proteins, and ␣2-6 sialylated proteins were then immunoblotted for the Fas receptor. As depicted in Fig. 4A, the band representing Fas in the SNA precipitates from HD3.par cells is denser than that noted in precipitates from HD3.sh cells, indicating more sialylated Fas in the presence of high ST6Gal-I expression. However, no differences were observed in the amount of Fas present in whole cell lysates (representing total Fas protein), indicating that variant ST6Gal-I expression alters Fas sialylation but not Fas protein expression. Also, MAA lectin precipitation (specific for ␣2-3 sialic acids) revealed that, despite a decrease in ␣2-6 sialylation, there was no change in ␣2-3 sialylation of the Fas receptor (Fig. 4A). SNA precipitation analyses performed with SW48 cells yielded similar results; Fas was found to be heavily sialylated in SW48 cells with forced expression of ST6Gal-I (SW.ST6 cells), whereas the Fas receptor expressed by parental and empty-vector SW48 cells lacked ␣2-6 sialylation, as can be seen in Fig. 4B. (MAA precipitation of SW48 cell lysate was not performed as these cells have no ␣2-3 sialyltransferases (21).) As with HD3 cells, variant expression of ST6Gal-I did not alter total Fas protein levels.
To further confirm Fas as an ST6Gal-I substrate, whole cell lysates from HD3.ev and HD3.sh cells were resolved using twodimensional electrophoresis to separate proteins by both molecular weight and isoelectric point, and then immunoblotting was performed to detect Fas isoforms (Fig. 4C). Several Fas isoforms were revealed by this assay; however, a higher molecular weight, more negatively charged band was missing from the HD3.sh lysates. These results are consistent with the loss of a sialylated Fas isoform in cells with ST6Gal-I knockdown. We also used flow cytometry to show that variant Fas sialylation did not change the levels of cell surface Fas expression (Fig. 4D).
␣2-6 Sialylation of Fas Inhibits DISC Formation-One possible explanation for the decrease in apoptotic signaling from sialylated Fas receptors was that sialic acids might sterically block the binding of Fas activators to the ligand-binding domain. We therefore examined the binding of CH11 antibody to the Fas receptor using flow cytometry. As shown in Fig. 5A,

ST6Gal-I Confers Protection against Fas-mediated Apoptosis
JULY 1, 2011 • VOLUME 286 • NUMBER 26 there was substantial, and equivalent, binding of CH11 to Fas expressed by the HD3.par and HD3.sh cell lines. To further investigate Fas-activator binding, we examined CH11 binding at a range of concentrations from 1 to 40 g/ml. HD3.par and HD3.sh cell lines showed comparable mean fluorescent intensity at every examined concentration (Fig. 5B). Thus, the strong inhibitory effect of ␣2-6 sialylation on Fas-dependent apoptosis cannot be attributed to diminished binding of the Fas-activating antibody. These data point to a role for sialylation in modulating some aspect of Fas receptor activation rather than ligand binding.
To examine the effect of ST6Gal-I sialylation on Fas signaling, HD3 cells were treated with CH11, and whole cell lysates were immunoblotted for cleaved caspase 8. Caspase 8, an initiator caspase, was evaluated because of its early recruitment to the DISC after Fas activation. As shown in Fig. 5C, the amount of cleaved caspase 8 was dramatically increased in CH11treated HD3.sh cells as compared with the HD3.par cells, suggesting that Fas sialylation alters signaling at some step upstream of caspase 8 activation.
FADD is the initial protein recruited to the DISC that binds directly to the cytoplasmic tail of the Fas receptor after activation. Therefore we examined the amount of FADD associated with the Fas receptor tails by using co-immunoprecipitation experiments. HD3 cells were treated with CH11 at 37°C to activate DISC formation, and then the Fas receptor and the associated DISC complex were immunoprecipitated. As a control, cells were incubated with CH11 antibody at 4°C, a treatment that does not induce DISC formation. Immunoprecipitates were then blotted for associated FADD (Fig. 5D). These experiments showed that a basal amount of FADD was bound to the Fas cytoplasmic tails in control cells, and no apparent increase in Fas/FADD association was observed in the HD3.par cells upon activation by CH11. In marked contrast, CH11 treatment of HD3.sh cells at 37°C induced a substantial increase in FADD binding to the Fas cytoplasmic tails, indicative of DISC formation. These results suggest that sialylation of Fas somehow alters the accessibility of Fas cytoplasmic domains for binding to FADD and consequently regulates the first step in DISC formation.
␣2-6 Sialylation Inhibits Fas Receptor Internalization-The internalization of Fas after receptor activation is necessary for

. Fas is a substrate for ST6Gal-I sialylation.
A, knockdown of ST6Gal-I reduces the amount of ␣2-6, but not ␣2-3, sialylation on Fas. HD3 cell lysates were incubated with either SNA lectin (specific for ␣2-6 sialic acids) or MAA lectin (specific for ␣2-3 sialic acids) each conjugated to agarose beads. Sialylated proteins were precipitated by centrifugation, and, after extensive washing, precipitates were resolved by SDS-PAGE, and immunoblotted for Fas. As shown, ␣2-6 sialylated Fas is present at a much higher level in the HD3.par as compared with the HD3.sh cells, whereas levels of ␣2-3 sialylation were very similar in the two cell lines. Whole cell lysates (i.e. not subjected to lectin precipitation) were also immunoblotted for Fas to reveal total levels of Fas protein. B, forced expression of ST6Gal-I induces ␣2-6 sialylation of Fas. SW48 cells were subjected to SNA precipitation as described in A. Whole cell lysates were also immunoblotted to detect total Fas levels (MAA precipitation was not performed with SW48 cells due to the known lack of endogenous sialyltransferases in this cell line.) In A and B, the levels of ␣2-6 sialylated Fas present are correlated with ST6Gal-I expression, suggesting Fas to be a target for ST6Gal-I sialylation. C, loss of a higher molecular weight, more acidic Fas isoform in cells with ST6Gal-I knockdown. Whole cell lysates were subjected to resolution by two-dimensional SDS-PAGE. Proteins were transferred to PVDF membranes and immunoblotted for Fas. The more acidic, higher molecular weight isoform of Fas is not present upon the knockdown of ST6Gal-I (HD3.sh). D, surface expression of Fas is not altered by ␣2-6 sialylation. Cells were stained with FITC-conjugated anti-Human Fas or FITC-conjugated isotype IgG control and characterized by flow cytometry. Surface expression of Fas is not altered as a consequence of differential ST6Gal-Imediated sialylation.

ST6Gal-I Confers Protection against Fas-mediated Apoptosis
Fas-mediated apoptotic signaling (4). Therefore, to further characterize the effects of receptor sialylation, we examined Fas internalization after treatment with CH11. We treated cells with CH11 at either 4°C (as a control) or 37°C (to allow for signaling and internalization), fixed the cells, and then used anti-mouse IgM Alexa-fluor 488 (Invitrogen) to visualize the Fas receptor remaining on the cell surface of non-permeabilized cells. As shown in Fig. 6, no major differences were noted in surface Fas levels upon CH11 treatment at 4°C; however, upon receptor activation at 37°C, substantially more Fas was internalized in HD3.sh cells. These data suggest that ␣2-6 sialylation of Fas inhibits receptor internalization, thus limiting Fas-dependent apoptosis.

DISCUSSION
There are multiple sialyltransferases that add sialic acid in an ␣2-3 linkage to N-glycans, however ST6Gal-I is the predominant enzyme that elaborates the ␣2-6 linkage of sialic acid to FIGURE 5. ␣2-6 sialylation of Fas inhibits DISC formation. A, Fas-activating antibody binds equally well to sialylated and unsialylated Fas receptor. HD3.par and HD3.sh cells were treated with CH11, and antibody binding was measured by flow cytometry (cells were treated at 4°C to prevent Fas activation). The binding levels for both cell lines are very similar. B, CH11 binding curve. Flow cytometric analyses of cells treated with CH11 at 1, 10, 20, or 40 g/ml revealed comparable mean fluorescent intensities (MFI) at all concentrations. Values for MFI were normalized to the IgM control. C, caspase 8 cleavage is affected by variant Fas sialylation. Cells were treated with CH11, and lysates were resolved by SDS-PAGE and immunoblotted for cleaved caspase 8. The level of cleaved caspase 8 is higher in the HD3.sh as compared with HD3.par upon treatment with CH11, indicating enhanced cleavage of caspase 8 when ST6Gal-I is down-regulated. D, DISC formation enhanced by knockdown of ST6Gal-I. DISC components were isolated by treating cells at 4°C (control) or 37°C for 30 min with CH11 and precipitating through anti-IgM-conjugated agarose beads. Proteins were resolved by SDS-PAGE, and immunoblotted for the Fas-binding protein, FADD. As shown, there is more FADD in HD3.sh after Fas activation with CH11 at 37°C, indicating more DISC formation in ST6Gal-I deficient cells. The PVDF membrane was stripped and reprobed for the presence of Fas to verify equal immunoprecipitation. Densitometry was used to quantify the FADD and Fas bands, and the FADD:Fas ratio was calculated. Values shown represent the averages from two independent experiments. N-glycosylated proteins (24,25). Another ␣2-6 sialyltransferase, ST6Gal-II, has been identified, however this enzyme is localized primarily to the brain and preferentially sialylates oligosaccharides rather than glycoproteins (26,27). There are also several ST6GalNAc enzymes that add ␣2-6 sialic acid to the GalNAc residue of O-linked glycans or gangliosides (reviewed in Ref. 28). The ST6Gal-I enzyme is overexpressed in at least 13 different types of cancers, including colon, breast, esophageal, oral, ovarian, cervical, leukemias, and brain tumors (10 -15, 29 -31), and high ST6Gal-I levels are associated with metastasis and poor patient prognosis (15)(16)(17). The functional contribution of ST6Gal-I to tumor progression has not been widely investigated, however increased ␣2-6 sialylation has been linked to enhanced tumor cell migration and invasion (12, 18 -20, 32). We and others have reported that ST6Gal-I expression is increased by oncogenic ras (reviewed in Ref. 33), which is found in over 30% of human cancers, and Piller's group showed this up-regulation takes place through the ralGEF pathway (34). Studies of ST6Gal-I in colon carcinoma have revealed that 90% of colon tumors screened had up-regulated ST6Gal-I expression (11) and 70% of colorectal cancers were positive for the ␣2-6 sialic acid modification added by ST6Gal-I (35). Animal studies also support a role for ST6Gal-I up-regulation in tumor progression. Human (16) and murine (36) cancer cells with high levels of ␣2-6 sialylation were more metastatic to liver following splenic injection in nude mice and enzymatic removal of sialylation from tumor cells prior to injection inhibited metastasis. Furthermore, Varki's group reported that ST6Gal-I-null mice bred to a spontaneous breast cancer model displayed tumors that were more differentiated than tumors from wild-type mice, suggesting that ST6Gal-I activity contributes to the poorly differentiated phenotype of more advanced cancers (37).
Despite these compelling results, there is still a limited understanding of the mechanism by which ST6Gal-I-directed sialylation regulates tumor cell behavior, because of the sparse knowledge of ST6Gal-I substrates, as well as the lack of information regarding the effects of ␣2-6 sialylation on the function of specific proteins. The current investigation provides a significant advance toward defining the role of ST6Gal-I in tumor progression by showing that the Fas receptor is ␣2-6 sialylated by ST6Gal-I. Furthermore, we show that ␣2-6 sialylation of Fas inhibits Fas signaling through both 1) blocking the binding of FADD to the Fas cytoplasmic tail, which is the first step in DISC formation, and 2) inhibiting internalization of the stimulated Fas receptor, which is necessary for Fas apoptotic signaling (4). Sialylation can alter receptor function through several mechanisms, including conformational alteration (38), clustering (39), and differential internalization rate, depending on the specific receptor. Consistent with our work, the CD45 and PECAM receptors have been shown to be targets for ␣2-6 sialylation by ST6Gal-I, and this sialylation affects internalization in both cases (39 -41).
The role of ␣2-6 sialylation in tumor progression has most often been associated with effects on tumor cell migration and invasion, however it is emerging that ST6Gal-I may be a major regulator of tumor cell survival. This sialylation-dependent survival benefit is likely mediated through multiple molecular pathways. Studies by our group and others have shown that ST6Gal-I-directed ␣2-6 sialylation of selected receptors serves as a key negative regulator of galectin-induced apoptosis (39,42,43). Additionally, the diminished internalization of PECAM due to ␣2-6 sialylation allows anti-apoptotic signaling from PECAM for a longer time interval (40). It has also been determined that ST6Gal-I levels are up-regulated after radiation treatment in mice (44), and these higher levels confer protection against radiation-induced apoptosis. Our novel finding that ␣2-6 sialylation inhibits Fas signaling adds to the growing body of literature suggesting that ST6Gal-I modifies a select group of substrates to regulate a multiplicity of apoptotic signaling cascades, thus providing a strong selective advantage for tumor cells.
Changes in expression of the Fas receptor and associated signaling molecules have long been regarded as a mechanism by which tumors evade apoptosis. Several types of cancers are known to down-regulate the Fas receptor as a protective measure, including colon (45), testicular (46), and hepatoma (47). In addition, increased FasL expression on the tumor cell surface has been reported in several cancers (8) and is thought to play a role in tumor evasion from the immune system. Expression levels of downstream effectors in the Fas-induced apoptotic cascade are also altered in tumor tissue, providing additional mechanisms for protection against Fas-mediated apoptosis. Alterations include up-regulation of the anti-apoptotic Bcl and c-FLIP and down-regulation of pro-apoptotic Bax (5-7). Interestingly, Fas is highly expressed on the surface of many tumor types that are not susceptible to Fas-mediated apoptosis (48 -51). Results presented herein implicate ␣2-6 sialylation as a newly identified mechanism by which tumor cells disable Fas signaling. A role for sialylation in regulating the Fas receptor was previously suggested by studies in lymphocytes (52,53). Keppler et al. reported that highly sialylated subclones of the BJA-B B lymphoma cell line were less susceptible to Fas-mediated apoptosis, although ␣2-3 and ␣2-6 sialic acids were not distinguished in this study (52). Consistent with these results, Peter et al. determined that enzymatic removal of all cell sialic FIGURE 6. ␣2-6 sialylation of Fas decreases Fas internalization. Cells were suspended and treated at 4°C (control) or 37°C for 30 min with 1 g/ml CH11. Cells were then fixed, and secondary anti-mouse IgM Alexa-fluor 488 (Invitrogen) was used to visualize the remaining Fas receptors on the surface of non-permeabilized cells. As can be seen in the HD3.sh cells (bottom panels), there is extensive internalization of the Fas receptor at 37°C, which is not seen in the HD3.par cells (top panels).

ST6Gal-I Confers Protection against Fas-mediated Apoptosis
acids (i.e. both ␣2-3 and ␣2-6 sialylation) increased the vulnerability of B and T cells to Fas-mediated apoptosis (53). Although these prior studies clearly linked sialylation to Fas activity, they provided limited information regarding the mechanism given that there are more than 20 different sialyltransferases. In the current study we show that Fas activity is regulated through ␣2-6 sialylation by a distinct sialyltransferase, ST6Gal-I, and of equal importance, this type of sialic acid modification has physiologic relevance given the known up-regulation of ST6Gal-I in cancer. Indeed we previously reported that another ST6Gal-I substrate, the ␤1 integrin, expresses elevated ␣2-6 sialylation in 100% of human colon tumors (20). It is also noteworthy that the total levels of sialylation on the two cell model systems evaluated in this study are quite different; HD3 cells express ␣2-3 sialyltransferases (19), whereas SW48 cells express no endogenous sialyltransferases (21). Thus, ST6Gal-I-mediated sialylation of Fas blocks Fas signaling in cells with both extensive ␣2-3 cell surface sialylation, including ␣2-3 sialylation of the Fas receptor (HD3), and cells with no surface sialylation other than that directed by exogenously expressed ST6Gal-I (SW48). These results imply a unique functionality imparted by the ␣2-6 sialic acid modification added by ST6Gal-I. Further highlighting the specificity of this novel molecular pathway, ST6Gal-I activity appears to have no effect on apoptotic signaling by the DR4 or DR5 death receptors. Intriguingly, work by Ashkenazi's group has shown O-glycosylation of the DR5 receptor regulates sensitivity to the DR5 ligand, TRAIL, but does not affect signaling by Fas (54). Moreover, the sites for O-glycosylation on DR5 are not conserved in the Fas receptor, and correspondingly, the DR5 receptor does not contain consensus sequences for N-glycosylation. A fundamental concept highlighted by these observations is that there is specificity in the effects of certain glycan structures on the function of distinct death receptors.
In conclusion, the current study is the first to demonstrate that a specific sialic acid modification, elaborated by an enzyme known to be up-regulated in cancer, inactivates signaling through the Fas receptor. These results have several important translational implications, including the potential for ␣2-6 sialylation to serve as a biomarker for Fas insensitivity. In addition, the finding that forced down-regulation of Fas ␣2-6 sialylation sensitizes tumor cells to apoptosis, even in cells expressing the powerful ras oncogene, suggests that ST6Gal-I may be a promising therapeutic target. Finally, ST6Gal-I-dependent regulation of Fas suggests a new paradigm in death receptor signaling, and it more broadly highlights an emerging role for ST6Gal-I as a critical mediator of multiple cell survival pathways.