The ST6Gal I Sialyltransferase Selectively ModifiesN-Glycans on CD45 to Negatively Regulate Galectin-1-induced CD45 Clustering, Phosphatase Modulation, and T Cell Death*

The addition of sialic acid to T cell surface glycoproteins influences essential T cell functions such as selection in the thymus and homing in the peripheral circulation. Sialylation of glycoproteins can be regulated by expression of specific sialyltransferases that transfer sialic acid in a specific linkage to defined saccharide acceptor substrates and by expression of particular glycoproteins bearing saccharide acceptors preferentially recognized by different sialyltransferases. Addition of α2,6-linked sialic acid to the Galβ1,4GlcNAc sequence, the preferred ligand for galectin-1, inhibits recognition of this saccharide ligand by galectin-1. SAα2,6Gal sequences, created by the ST6Gal I enzyme, are present on medullary thymocytes resistant to galectin-1-induced death but not on galectin-1-susceptible cortical thymocytes. To determine whether addition of α2,6-linked sialic acid to lactosamine sequences on T cell glycoproteins inhibits galectin-1 death, we expressed the ST6Gal I enzyme in a galectin-1-sensitive murine T cell line. ST6Gal I expression reduced galectin-1 binding to the cells and reduced susceptibility of the cells to galectin-1-induced cell death. Because the ST6Gal I preferentially utilizes N-glycans as acceptor substrates, we determined that N-glycans are essential for galectin-1-induced T cell death. Expression of the ST6Gal I specifically resulted in increased sialylation of N-glycans on CD45, a receptor tyrosine phosphatase that is a T cell receptor for galectin-1. ST6Gal I expression abrogated the reduction in CD45 tyrosine phosphatase activity that results from galectin-1 binding. Sialylation of CD45 by the ST6Gal I also prevented galectin-1-induced clustering of CD45 on the T cell surface, an initial step in galectin-1 cell death. Thus, regulation of glycoprotein sialylation may control susceptibility to cell death at specific points during T cell development and peripheral activation.

The addition of sialic acid to T cell surface glycoproteins influences essential T cell functions such as selection in the thymus and homing in the peripheral circulation. Sialylation of glycoproteins can be regulated by expression of specific sialyltransferases that transfer sialic acid in a specific linkage to defined saccharide acceptor substrates and by expression of particular glycoproteins bearing saccharide acceptors preferentially recognized by different sialyltransferases. Addition of ␣2,6-linked sialic acid to the Gal␤1,4GlcNAc sequence, the preferred ligand for galectin-1, inhibits recognition of this saccharide ligand by galectin-1. SA␣2,6Gal sequences, created by the ST6Gal I enzyme, are present on medullary thymocytes resistant to galectin-1-induced death but not on galectin-1-susceptible cortical thymocytes. To determine whether addition of ␣2,6-linked sialic acid to lactosamine sequences on T cell glycoproteins inhibits galectin-1 death, we expressed the ST6Gal I enzyme in a galectin-1-sensitive murine T cell line. ST6Gal I expression reduced galectin-1 binding to the cells and reduced susceptibility of the cells to galectin-1-induced cell death. Because the ST6Gal I preferentially utilizes N-glycans as acceptor substrates, we determined that N-glycans are essential for galectin-1induced T cell death. Expression of the ST6Gal I specifically resulted in increased sialylation of N-glycans on CD45, a receptor tyrosine phosphatase that is a T cell receptor for galectin-1. ST6Gal I expression abrogated the reduction in CD45 tyrosine phosphatase activity that results from galectin-1 binding. Sialylation of CD45 by the ST6Gal I also prevented galectin-1-induced clustering of CD45 on the T cell surface, an initial step in galectin-1 cell death. Thus, regulation of glycoprotein sialylation may control susceptibility to cell death at specific points during T cell development and peripheral activation.
The role of glycosylation in these functions is specific, i.e. the different functions require specific sugars on specific glycoprotein acceptors. Regulated glycosylation of specific acceptor substrates can affect immune function by creating or masking ligands for endogenous lectins. For example, modification of cell surface oligosaccharides by the C2GnT and Fuc TVII glycosyltransferases results in specific selectin-mediated trafficking patterns for Th1 and Th2 subsets (3). Similarly, modification of CD45 by the C2GnT glycosyltransferase regulates thymocyte susceptibility to cell death induced by galectin-1 (10).
During T cell development, expression of several glycosyltransferases is temporally and spatially controlled (9,11,12). In the human thymus, different members of the sialyltransferase family are expressed in distinct anatomic compartments, so cells in those compartments bear unique complements of sialylated oligosaccharides. For example, the SA␣2,6Gal sequence, the product of the ST6Gal I sialyltransferase, is detected only on mature medullary thymocytes (12). Intriguingly, mature medullary thymocytes displaying SA␣2,6Gal sequences are resistant to galectin-1-induced cell death (13,14). Because the addition of sialic acid in the ␣2,6 linkage to galactose could mask terminal galactose residues required for galectin-1 binding to T cell glycoproteins (15), we asked whether expression of the ST6Gal I would control susceptibility of T cells to galectin-1-induced death.
Transfection-Rat ST6Gal I cDNA in the plasmid ST Tyr -Myc-pcDNA 3.1 (17, 18) (gift of Dr. Karen Colley, University of Illinois, Chicago, IL) or vector alone were transfected into Pha R 2.1 and T200 Ϫ cells as described (9). Following selection in G418, positive Pha R 2.1 clones were identified by SNA flow cytometry. Positive T200 Ϫ clones were identified by RT-PCR, performed essentially according to the protocol provided in the Super Script TM One-Step RT-PCR with Platinum Taq (Invitrogen, Carlsbad, CA), using the primers 94 sense (TATGAGGCCCCTTACAC-TG) and 943A antisense (GCCGGAGGATGGGGGATTTGG) (18).
Galectin-1 Binding Assay-5 ϫ 10 5 cells were suspended in PBS containing the indicated amount of biotinylated galectin-1 (16) at 4°C for 1 h. After washing, the cells were incubated with streptavidinfluorescein isothiocyanate (5 g/ml) (Jackson Immunoresearch Laboratories, West Grove, PA) for 45 min at 4°C. After washing, the cells were analyzed by flow cytometry.
Galectin-1 Cell Death Assays-Galectin death assays were performed as described (16) with the following modifications. 10 5 cells were incubated with 20 M galectin-1 in 1.6 mM dithiothreitol/Dulbecco's modified Eagle's medium or in 1.6 mM dithiothreitol/Dulbecco's modified Eagle's medium alone as a control for 4 -6 h at 37°C. 0.1 M ␤-lactose (final concentration) was added to dissociate galectin-1, and the cells were washed with PBS. Apoptotic cells were identified using annexin V and propidium iodide as previously described (10).
Precipitation and Western Blot Analysis-The cell lysates from 4 -9 ϫ 10 6 cells were prepared as described (12). To precipitate SNAbinding glycoproteins, the lysates were precleared for 1 h with biotinylated bovine serum albumin (0.25 g/225 l cell lysate) and Immuno-Pure Immobilized Streptavidin (Pierce). After centrifugation to remove insoluble material, the supernatants were incubated with SNA-biotin (5 g/300 l cell lysate) and ImmunoPure Immobilized Streptavidin overnight. The precipitates were washed four times with lysis buffer prior to SDS-PAGE. All of the steps were performed at 4°C. To precipitate CD45, the supernatants were precleared with purified rat IgG2 b, (0.25 g/300 l cell lysate) (Pharmingen, San Diego, CA) and ImmunoPure Immobilized Protein G (Pierce), and CD45 was precipitated with monoclonal antibody 30-F11 (3 g/300 l cell lysate) (Pharmingen) and ImmunoPure Immobilized Protein G overnight.
CD45 or SNA precipitates were separated by SDS-PAGE, blotted to nitrocellulose, and probed with polyclonal goat anti-mouse CD45 (0.2 g/ml) (Research Diagnostics Inc., Flanders, NJ) or SNA-biotin (1 g/ ml). Bound reagent was detected with horseradish peroxidase-labeled rabbit anti-goat IgG (Bio-Rad) or streptavidin-horseradish peroxidase, respectively, and visualized by ECL (Amersham Biosciences). ST6Gal I immunoblotting of whole cell lysates was performed as described in Ref. 12, with rabbit anti-rat ST6Gal I antiserum (gift of Dr. K. Colley).
PNGase F Digestion of CD45-Cell lysates (10 6 cells) were separated by SDS-PAGE, blotted to nitrocellulose and probed with polyclonal goat anti-mouse CD45 (M-20) (Santa Cruz Biotechnology, Santa Cruz, CA). The band corresponding to CD45 was excised from the nitrocellulose, and the bound antibody was stripped with Restore buffer (Pierce). After washing two times with 25 mM Tris, 150 mM NaCl, 0.05% Tween, pH 7.5 (TBS-T) followed by two washes with 50 mM sodium phosphate, pH 7.5, the membrane was incubated with 1.5 ml of 50 mM sodium phosphate containing 10,000 units of PNGase F (New England BioLabs, Beverly, MA) overnight at 37°C with rocking. The enzyme-treated membrane was washed with TBS-T and probed with SNA-biotin, as described above.
CD45 Segregation Analysis-The cells were treated with or without galectin-1 as in the death assays. After treatment, the cells were washed with PBS and fixed with 2% paraformaldehyde for 30 min at 4°C. The reaction was quenched with 0.2 M glycine for 5 min at 4°C, and the cells were blocked with in 10% goat serum for 1.5 h at room temperature. The cells were washed with PBS and incubated with polyclonal goat anti-mouse CD45 conjugated to fluorescein isothiocyanate (Pharmingen) in 2% goat serum for 1.5 h at room temperature in the dark. After washing, the cells were mounted on slides with Prolong Anti-fade medium (Molecular Probes, Eugene OR). CD45 cell surface localization was analyzed on a Fluoview laser scanning confocal microscope (Olympus America Inc, Melville, NY), at 100ϫ. The number of cells demonstrating CD45 segregation or clustering and the total number of cells were counted for six randomly selected fields for each experiment. Approximately 50 cells were counted in six fields. Percent CD45 segregation was calculated as [100 ϫ (number of CD45 segregated cells/total number of cells)].
Protein-tyrosine Phosphatase Activity Assay-4 ϫ 10 6 cells were incubated at 37°C in 400 l of medium containing PBS as a control (0 min) or 30 g of galectin-1 for 1, 5, 15, or 30 min. At the indicated times, the cells were cooled on ice, washed in PBS at 4°C, and lysed (12), and the protein concentrations of the cell lysates were determined (protein assay kit; Bio-Rad). Protein-tyrosine phosphatase (PTP) activity was measured using p-nitrophenyl phosphate (Calbiochem) as a substrate in the presence of okadaic acid to inhibit protein Ser/Thr phosphatases. Cell lysate (20 g/25 l) was incubated at room temperature for 4 h in 475 l of PTP assay buffer (100 mM Hepes, pH 7.2, 150 mM NaCl, 1 mM EDTA, 1 mM dithiothreitol, 10 mM p-nitrophenyl phosphate, 50 nM okadaic acid) either in the absence or presence of 50 M bpV (phen) (potassium bisperoxo (1,10-phenanthraline)oxo-vanadate(v)) as a specific PTP inhibitor (10). The reaction was stopped by adding 500 l of 1 N NaOH, and released p-nitrophenol was measured at A 415 against appropriate blanks.

N-Glycans Are Essential for Galectin-1 Death-Galectin-1
preferentially recognizes Gal␤1,4GlcNAc (LacNAc) sequences that can be presented on N-or O-linked glycans (15). Although prior work from our lab demonstrated that O-glycans participate in galectin-1 T cell death (9), the role of N-glycans in galectin-1 cell death is not clear. In addition, the ST6Gal I enzyme preferentially sialylates terminal galactose residues on N-glycans (12,17); if the ST6Gal I participated in regulating galectin-1 cell death in vivo, it would likely occur through the modification of N-glycans.
To determine whether N-glycans are necessary for galectin-1 induced death, human and murine T cell lines were treated with the mannosidase I inhibitor DMNJ, to block trimming of terminal mannose residues and subsequent elongation of Nglycans with LacNAc sequences. The effectiveness of DMNJ treatment was determined by analyzing treated cells with the PHA, because inhibition of mannosidase I activity would prevent elongation of the N-glycan chain recognized by PHA (19). Cells treated with DMNJ showed a marked reduction in PHA binding compared with cells cultured in medium alone; importantly, DMNJ treatment did not affect the level of cell surface expression of galectin-1 receptors CD43 or CD45, as determined by flow cytometric analysis using the relevant antibodies (data not shown).
DMNJ-treated cells were examined for susceptibility to galectin-1-induced cell death (Fig. 1). The Pha R 2.1, CEM, and MOLT-4 cell lines are all susceptible to galectin-1-induced cell death, whereas the BW5147 cells are resistant to galectin-1 because of the lack of core 2 O-glycans on cell surface glycoproteins (9,10,16). DMNJ treatment resulted in a dramatic reduction in galectin-1-induced cell death of the galectin-1-susceptible murine (Pha R 2.1) and human (CEM, MOLT-4) T cell lines. Although previous studies demonstrated that the glycosylation inhibitors benzyl-␣-GalNAc and swainsonine reduced T cell susceptibility to galectin-1 (9, 16), neither benzyl-␣-Gal-NAc nor swainsonine had the dramatic inhibitory effect on cell death that we observed with DNMJ treatment. These results demonstrated that N-linked glycans are essential for galectin-1-mediated T cell death.
Expression of the ST6Gal I Reduces Galectin-1 Binding to T Cells-The addition of terminal ␣2,6-linked sialic acid can block galectin-1 binding to the preferred saccharide ligand LacNAc. This has been demonstrated for individual LacNAc units and for poly-LacNAc chains (15, 20 -22). The ability of a terminal SA␣2,6Gal sequence to block galectin-1 binding suggested that the addition of ␣2,6-linked sialic acid to T cell surface glycoproteins (12) could regulate the susceptibility of thymocytes and T cells to galectin-1.
To directly examine whether addition of ␣2,6-linked sialic acid would affect susceptibility to galectin-1, we expressed the ST6Gal I in the galectin-1-susceptible murine T cell line Pha R 2.1. The plant lectin SNA recognizes the SA␣2,6Gal sequence (12,23). We used SNA binding, detected by flow cytometry, to screen for clones expressing the ST6Gal I. Two SNA ϩ clones, SNA.1 and SNA.9, demonstrated increased SNA binding compared with a control clone transfected with vector alone (C.2) ( Fig. 2A). Both RT-PCR and immunoblot analysis with anti-ST6Gal I serum demonstrated abundant expression of ST6Gal I mRNA and protein in SNA.9 cells (Fig. 2B) and SNA.1 cells (data not shown), whereas no reactivity was observed in control C.2 cells (Fig. 2B). ST6Gal I expression did not affect the level of expression of galectin-1 receptors CD43 or CD45 on the SNA.1 and SNA.9 cells, determined by flow cytometric analysis, nor the level of CD7 expression, detected by immunoblotting (data not shown).
We then determined whether addition of cell surface sialic acid by the ST6Gal I enzyme would reduce galectin-1 binding to T cells. As shown in Fig. 2C, galectin-1 binding to the SNA.9 cells (closed circles) was markedly reduced compared with the level of binding observed for the C.2 control cells transfected with vector alone (open circles). However, the reduced, but not absent, binding of galectin-1 to SNA.9 cells indicated that some of the potential binding sites on these cells were not modified by the ST6Gal I enzyme. For both SNA.9 and C.2 cells, galectin-1 binding was completely inhibited in the presence of 100 mM lactose (squares), demonstrating that binding was saccharide-dependent.
Expression of the ST6Gal I Reduces Susceptibility to Galectin-1-The SNA.1, SNA.9, and C.2 cells were examined for susceptibility to galectin-1-induced death. As shown in Fig. 2D, the C.2 cells transfected with vector alone were susceptible to galectin-1; ϳ50% of the cells underwent cell death, determined by annexin V binding and PI uptake. In contrast, the SNA.1 and SNA.9 cells demonstrated only 10 and 18% galectin-1induced death, respectively. The resistance of the SNA.1 and SNA.9 cells to galectin-1 did not appear to result from a complete block of galectin-1 binding, as demonstrated by the binding curve in Fig. 2C; in addition, all of the clones demonstrated cell agglutination when galectin-1 was added, and the cell agglutinates were dispersable by the addition of lactose (data not shown). These data demonstrated that expression of the ST6Gal I, resulting in creation of SA␣2,6Gal sequences on cell surface glycoproteins, reduced susceptibility to galectin-1-induced T cell death and suggested that sialylation of specific glycoproteins was responsible for resistance to galectin-1-induced cell death.
ST6Gal I Expression Results in Increased Sialylation of CD45-Our laboratory has demonstrated that the T cell surface glycoproteins CD7, CD43, and CD45 are receptors for galectin-1 (24). To determine whether these glycoproteins were specifically modified by the ST6Gal I, we examined CD7, CD43, and CD45 for increased sialylation of N-glycans by SNA and antibody precipitation.
Total SNA-binding proteins were precipitated and probed with SNA. As shown in Fig. 3A, there was only one significant difference in the pattern of SNA-binding glycoproteins precipitated from cells expressing the ST6Gal I (SNA.1) compared with vector-transfected controls (C.2, C.4). In extracts of SNA.1 cells, there was an obvious increase in SNA binding to a band of approximate molecular mass of 200 kDa. Other SNA reactive bands of various sizes were occasionally seen in different experiments (data not shown), but these other bands were not consistently observed. In contrast, the 200-kDa SNA ϩ band was consistently observed in ST6Gal I-expressing clones. The relative mobility of the 200-kDa band suggested that it could be CD45, a highly glycosylated protein that is known to bear SA␣2,6Gal sequences on both murine and human T cells (12,24).
To specifically determine whether the band exhibiting increased SNA binding was CD45, both SNA and CD45 antibody were used to precipitate material from vector transfected (C.4) and SNA.1 cells, and the precipitates were probed with CD45 (Fig. 3B). The 200-kDa band exhibiting increased SNA binding reacted with CD45 antibody, demonstrating that CD45 was selectively hypersialylated in the SNA.1 cells. In addition, this band migrated with the same mass as immunoprecipitated CD45. To determine whether the increased sialylation of CD45 occurred on N-glycans, the preferred glycan acceptor for the ST6Gal I, SNA.1 cells were pretreated with DMNJ prior to SNA or CD45 precipitation. DMNJ treatment reduced SNA binding to protein precipitated from SNA.1 cells to the level observed for control cells (C.4) transfected with vector alone (Fig. 3B). PNGase F treatment confirmed that, in cells overexpressing the ST6Gal I, sialic acid addition to CD45 occurred on N-glycans. Whole cell lysates of SNA.9 cells were probed with antibody to CD45. The CD45 bands were excised from the blot and incubated with or without PNGase F, and the bands were reprobed with SNA. As shown in Fig. 3C, PNGase F dramatically reduced SNA binding to CD45 from SNA.9 cells. Thus, the increased SNA binding to CD45 on cells expressing the ST6Gal I resulted from the specific addition of ␣2,6-linked sialic acid to N-glycans on CD45. The background level of binding of SNA to CD45 on control cells and on DMNJ-treated cells may reflect SNA recognition of SA␣2,6GalNAc sequences on O-glycans on CD45 (23).
As mentioned above, the three primary receptors for galectin-1 on T cells are CD7, CD43, and CD45 (24). We specifically precipitated CD7 and CD43 from SNA.9, SNA.1, and C.2 cells and saw no difference in SNA binding to CD7 or CD43 (data not shown), indicating that the inhibitory effect on galectin-1 cell death was not due to sialylation of CD7 or CD43. We also did not detect CD7 or CD43 by immunoblotting SNA precipitates with the respective antibodies (data not shown). To further examine the acceptor substrate preference of the ST6Gal I, we expressed the ST6Gal I in the murine T200 Ϫ cell line, a mutant of the BW5147 line that does not express CD45. Despite repeated attempts, we could not isolate SNA ϩ clones from T200 Ϫ cells transfected with ST6Gal I cDNA (SNA.T1), nor could we detect any increase in SNA binding to whole cell lysates of SNA.T1 cells (Fig. 4, A and B). Although RT-PCR analysis demonstrated that the ST6Gal I mRNA was present in nine FIG. 4. A, expression of the ST6Gal I in the T200 Ϫ cell line, which does not express CD45, did not result in increased SNA binding. Nine clones expressing ST6Gal I mRNA were examined, but none demonstrated increased SNA binding by flow cytometry; clone SNA.T1 is shown for example. C.T1 is one of nine control clones transfected with vector alone. B, SNA blotting of whole cell extracts of C.T1 and SNA.T1 cells did not demonstrate any differences in staining between the two clones. C, T200 Ϫ clones express ST6Gal I mRNA and protein. RT-PCR and immunoblot analysis of nine clones demonstrated ST6Gal I expression, as shown for the SNA.T1 clone, with no ST6Gal I expression in any of the controls, as shown for the C.T1 clone. The samples are representative of all 18 clones examined. The expressed protein is enzymatically active, as demonstrated by the addition of sialic acid to asialofetuin. Asialofetuin was incubated with lysates of C.T1 or SNA.T1 cells and precipitated with anti-fetuin, and ␣2,6-linked sialic acid was detected by SNA blotting. Weak SNA reactivity of fetuin incubated with extract of C.T1 cells may reflect the addition of ␣2,6-linked sialic acid to O-glycans, because no SNA reactivity was detected with the asialofetuin acceptor substrate alone (not shown). Densitometric analysis of the SNA-binding bands was performed; the ratio of SNA binding to fetuin incubated with SNA.T1 extract compared with C.T1 extract was 6.3:1.

FIG. 3. The ST6Gal I preferentially sialylates N-glycans on CD45.
A, total SNA-binding glycoproteins were precipitated from control clones transfected with vector alone (lanes C.2 and C.4) or from the SNA.1 clone expressing the ST6Gal I. Precipitated glycoproteins were probed with biotinylated SNA. The only significant difference in the profile of SNA binding glycoproteins was an increase in a band with a mass of ϳ200 kDa (arrow). B, the SNA reactive band is CD45. The cells were cultured in 2 mM DMNJ, as above, or in medium alone. The cell lysates were precipitated with CD45 antibody or SNA (indicated below) and probed with CD45 antibody. The band with increased SNA staining reacts with both SNA and antibody to CD45. In addition, the increased SNA binding to CD45 is abolished by pretreatment with DMNJ, which blocks synthesis of complex N-glycans. In both blots, the width of the CD45 band is diminished in DMNJ-treated cells compared with cells expressing the ST6Gal I, as a result of decreased complexity of glycosylation. C, increased SNA binding to CD45 results from sialylation of N-glycans. CD45 was detected in whole cell lysates of SNA.9 cells by immunoblotting (top panel). The CD45 bands were excised and incubated with or without PNGase F, as indicated, and reprobed with SNA-biotin. Removal of N-glycans from CD45 by PNGase F treatment reduced SNA binding. independent clones of ST6Gal I transfected T200 Ϫ cells (Fig.  4C), every clone was SNA Ϫ by flow cytometry (Fig. 4A). In addition, we detected ST6Gal I protein by immunoblotting in the ST6Gal I transfected T200 Ϫ cells (Fig. 4C), although the cells were SNA Ϫ . Finally, to confirm that the ST6Gal I expressed in T200 Ϫ cells was enzymatically active, we used asialofetuin as an acceptor substrate to assay sialyltransferase activity. After incubation with cell lysate from either control cells (C.T1) or SNA.T1 cells, fetuin was immunoprecipitated and subjected to blotting with SNA-biotin to detect ␣2,6-linked sialic acid. As shown in Fig. 4C, there was a significant increase in SNA binding to fetuin incubated with SNA.T1 extract, compared with C.T1 extract. We performed densitometric analysis of the SNA reactive bands; the ratio of SNA binding to fetuin incubated with SNA.T1 cell extract compared with C.T1 cell extract was 6.3. This ratio was comparable with the ratio we observed when asialofetuin was incubated with SNA.9 cell extract compared with C.4 cell extract, 5.0 (data not shown). These data demonstrated that equivalent ST6Gal I activity was present in the SNA.9 and SNA.T1 cells, although only the SNA.9 cells that express CD45 became SNA ϩ by flow cytometry. Thus, CD45 is the primary glycoprotein acceptor substrate for the ST6Gal I in these T cells, and in the absence of CD45, there was no detectable sialylation of other potential acceptors by the ST6Gal I in T200 Ϫ cells.
ST6Gal I Expression Inhibits CD45 Segregation on Galectin-1-treated Cells-We have demonstrated that galectin-1 binding to T cells results in reorganization of the glycoprotein receptors CD45, CD43, and CD7 into novel membrane microdomains (24). Specifically, CD45 segregates from CD43 and CD7 and localizes to membrane blebs on dying cells. The segregation of CD45 caused by galectin-1 binding is regulated in part by expression of the C2GnT glycosyltransferase that creates branches on O-glycans bearing the LacNAc sequences recognized by galectin-1. Cells that do not express the C2GnT do not demonstrate CD45 segregation after galectin-1 and are not susceptible to galectin-1-induced cell death (10).
We examined the effects of ST6Gal I expression on CD45 segregation after galectin-1 binding (Fig. 5). On C.4 cells transfected with vector alone, galectin-1 binding resulted in the segregation of CD45 to membrane blebs on dying cells, exactly as previously described (24). In contrast, galectin-1 binding to SNA.9 cells did not result in any detectable segregation of CD45. The diffuse distribution of CD45 on the cell surface was identical for SNA.9 cells treated with either galectin-1 or buffer control. Thus, expression of the ST6Gal I inhibited galectin-1induced CD45 segregation on the plasma membrane (Fig. 5A), as well as inhibiting galectin-1-induced cell death (Fig. 2D). A comparison of the effects of ST6Gal I expression on galectin-1induced CD45 segregation and on galectin-1-induced cell death is shown in Fig. 5B.
ST6Gal I Expression Abrogates Galectin-1-mediated Inhibition of PTP Activity-Previous work has demonstrated that binding of galectin-1 to CD45 reduces the PTP activity of CD45 (25,26). We asked whether ST6Gal I expression would modify the galectin-1-mediated effect on PTP activity. In human cell lines, the galectin-1 effect on immunoprecipitated CD45 has been examined (25,26). However, because all of the murine CD45 antibodies that we tested would not bind CD45 in the presence of galectin-1, we measured the PTP activity of whole cell lysates. The Pha R 2.1 cell line, the parental line of the SNA.1, SNA.9, and C.2 cells, demonstrated robust PTP activity (Fig. 6A). In contrast, the T200 Ϫ cell line derived from the same precursor line as the Pha R 2.1 cells does not express CD45 and has significantly reduced PTP activity (Fig. 6A). These results indicate that CD45 accounts for the majority of PTP activity in the Pha R 2.1 cells.
To assess the effect of ST6Gal I expression on PTP activity, we examined the SNA.9 and C.2 cells at the indicated time points after galectin-1 binding. As shown in Fig. 6B, galectin-1 binding to C.2 control cells resulted in a rapid and sustained decrease in PTP activity (open circles). However, this effect was not seen when galectin-1 was added to SNA.9 cells (closed circles); the PTP activity in lysates of SNA.9 cells treated with galectin-1 did not differ appreciably from that observed for cells treated with buffer alone (100%). All of the measurable pnitrophenol release was due to tyrosine phosphatase activity, because release was completely inhibited by the addition of bpV (phen), a tyrosine phosphatase inhibitor (10). DISCUSSION Regulated expression of glycosyltransferases affects many cell fate decisions. Altered glycosylation can directly modulate cellular responses by creating or masking ligands for endogenous lectins. For example, expression of specific glycosyltransferases creates potential selectin ligands on peripheral T cells migrating to sites of inflammation (3). Altered glycosylation can also indirectly modulate cellular responses by affecting glycoprotein conformation or by controlling intermolecular interactions. Expression of the GnT V enzyme controls the amplitude of the T cell response to antigen (6), and sialylation of cell surface glycoproteins regulates binding of MHC class I molecules to thymocytes (4,5).
Previous work from our group demonstrated that O-glycans are involved in galectin-1 induced cell death; specifically, addition of core 2 O-glycans on CD45 was required for galectin-1-mediated clustering of CD45, an initial step in the death pathway (9,10). The present work demonstrates that N-glycans are also essential for galectin-1-induced cell death, because treatment of murine and human T cells with the mannosidase I inhibitor DMNJ, which blocks all complex N-glycosylation, virtually abolished susceptibility to galectin-1 (Fig. 1). The dramatic inhibition of cell death seen with DMNJ treatment expands our previous work (16), demonstrating that treatment of T cells with the mannosidase II inhibitor swainsonine, which prevents branching of N-glycans by the GnT V enzyme, only partially inhibited galectin-1-induced cell death. Indeed, the murine Pha R 2.1 cell line used in this study is highly susceptible to galectin-1, although this cell line does not express the GnT V (9). Thus, although the GnT V branch may augment galectin-1 susceptibility, other LacNAc sequences on N-glycans are sufficient for galectin-1 binding to trigger the death signal.
In the T cell lines examined in this study, the preferred acceptor substrate for the ST6Gal I was CD45. Preferred utilization of CD45 as an acceptor substrate for the ST6Gal I is supported by our finding that, despite expression of the ST6Gal I in the T200 Ϫ cell line that lacks CD45, we detected no increase in SNA binding to these cells. Increased SNA binding to the other major galectin-1 receptors, CD43 or CD7, was not detected in the Pha R 2.1 or the T200 Ϫ cell lines. CD45 may be a preferred substrate because of accessibility of CD45 glycans to the ST6Gal I during synthesis or to recognition of peptide or conformational determinants on the CD45 backbone by the ST6Gal I enzyme.
Developmentally regulated changes in CD45 isoform expression may also control recognition by or accessibility to the ST6Gal I during glycoprotein synthesis. In human thymus, the SA␣2,6Gal sequence was only detected on the CD45RA isoform on mature thymocytes (12). In murine thymus, CD45 on mature thymocytes also appears to be a preferred acceptor for the ST6Gal I, because only mature thymocytes bound CD22, a lectin that preferentially recognizes SA␣2,6Gal (27). Recent work by Xu and Weiss (28) has also demonstrated preferential sialylation of high molecular weight isoforms of CD45, compared with the smallest CD45RO isoform. Few examples of this degree of preferential acceptor substrate recognition by sialyltransferases in vivo have been reported. For example, polysialyltransferase enzymes are expressed in a range of tissues, but polysialic acid is detected primarily on the neural cell adhesion molecule NCAM (29,30). Thus, tissue specificity in both glycosyltransferase expression and in glycoprotein acceptor substrate expression can control cell surface glycosylation. Because galectin-1 is abundantly expressed throughout a variety of tissues, T cells will encounter galectin-1 in many organs and at many points during T cell development and peripheral activation. Thus, it is likely that the T cell response to galectin-1 will be controlled at the level of the T cell, i.e. by regulating glycosylation to control susceptibility to cell death (31).
Glycosylation of CD45 depends on a number of factors, including lymphocyte subset and stage of maturation or activation. Differential glycosylation of CD45 is controlled in part by the repertoire of glycosyltransferase enzymes expressed by the cell at each stage in T cell development (11,12,(32)(33)(34)(35). Regulated expression of different complements of glycosyltransferases during T cell maturation and activation implies that different glycoforms of CD45 will interact with different endogenous lectins, such as CD22, the cysteine-rich domain of the mannose receptor, or galectin-1 (10,27,36).
Sialylation of CD45 has recently been shown to regulate homodimerization of CD45 on the T cell surface (28). CD45 homodimerization is one mechanism to down-modulate the PTP activity of the CD45 cytoplasmic domains, an effect that would reduce T cell responsiveness to antigen. Although CD45 homodimerization has been proposed to occur spontaneously (28), we and others have shown that galectin-1 binding clusters CD45 and reduces PTP activity (Figs. 5 and 6 and Refs. 25 and 26), and the data presented here demonstrate that this effect is negatively regulated by expression of the ST6Gal I and sialylation of CD45. Galectin-1 clustering of cell surface receptors has also been demonstrated to reduce T cell responsiveness to antigen (37), suggesting that galectin-1 binding to CD45 and regulation of CD45 PTP may contribute to the observed antiinflammatory properties of galectin-1 in a number of animal models (reviewed in Ref. 31). The addition of SA␣2,6Gal sequences to CD45 may be a mechanism to finely tune immune FIG. 6. ST6Gal I expression inhibits galectin-1-induced modulation of CD45 protein-tyrosine phosphatase activity. A, CD45 is the major PTP in Pha R 2.1 cells. Whole cell lysates of the CD45 ϩ parental cell line, Pha R 2.1, and the CD45 Ϫ derivative T200 Ϫ , were assayed for PTP activity in the presence (solid bar) or absence (open bar) of the PTP inhibitor bp V (phen). PTP activity was measured by the release of p-nitrophenol, detected at 415 nm. B, ST6Gal I expression abrogates the decrease in PTP activity triggered by binding of galectin-1. C.2 cells (open symbols) and SNA.9 cells (closed symbols) were incubated with 30 g of galectin-1 for the indicated times at 37°C. At the indicated times, the cells were lysed, and PTP activity in whole cell lysates was measured as described under "Experimental Procedures," in the presence (squares) or absence (circles) of the PTP inhibitor bpV (phen). C.2 cells demonstrate a 40% reduction in PTP activity 1 min after galectin-1 binding. SNA.9 cells demonstrate no change in PTP activity after galectin-1 binding. regulation by galectin-1 and to prevent galectin-1-induced apoptosis of specific populations, e.g. mature thymocytes.
How does ST6Gal I expression inhibit galectin-1-induced cell death? One possibility is that galectin-1 binds to LacNAc sequences on CD45 glycans; the addition of ␣2,6-linked sialic acid directly masks LacNAc sequences, inhibiting galectin-1 binding to and clustering of CD45 and initiation of cell death. Alternatively, the addition of sialic acid to CD45 would also impart additional negative charge. Galectin-1 may bind to other LacNAc sequences on CD45 that are not modified by sialic acid addition, but charge repulsion could prevent close packing of CD45 required to initiate cell death. On the cell surface, both direct masking of galectin-1 ligands on CD45 glycans and increased charge repulsion among CD45 molecules may contribute to inhibition of galectin-1-induced clustering, reduced PTP modulation, and resistance to death.
We are beginning to elucidate the critical roles played by specific glycosyltransferases in lymphocyte development and function. C2GnT transgenic mice demonstrated reduced T cell responses to antigen (2), whereas GnT V null mice demonstrated increased T cell responses to antigen (6). In ST3Gal I null mice, Marth and co-workers (7) found increased apoptosis of peripheral CD8 cells. This group also found profound defects in B cell function in ST6Gal I mice, although no defects in T cell development or function were reported (38). However, it is likely that complex interactions of glycosyltransferases may govern thymocyte susceptibility to galectin-1. For example, in C2GnT transgenic mice, we found increased susceptibility of immature double-positive thymocytes to galectin-1 but no increase in galectin-1 susceptibility of mature, single positive thymocytes (9). Based on our prior observation that medullary thymocytes expressed the SA␣2,6Gal sequence, we suggested that sialylation of single positive thymocytes could protect those cells from death.
It is increasingly apparent that post-translational modifications such as glycosylation are essential in regulating cellular signaling in the immune system, and that glycosylation is dynamically regulated during immune system activation (39). Understanding the functions of specific glycans, identifying the enzymes required to create or modify the glycans and characterizing the glycoprotein substrates that bear the glycans will provide new approaches to controlling lymphocyte development and survival.