Restricted Receptor Segregation into Membrane Microdomains Occurs on Human T Cells During Apoptosis Induced by Galectin-1¹

Apoptosis is a critical process regulating T cell development in the thymus and in controlling the immune response in the periphery. Galectin-1, a homodimeric member of a family of conserved lectins, induces apoptosis of human thymocytes, activated T cells, and T lymphoblastoid cell lines (1, 2, 3, 4).

In the thymus, galectin-1 is produced by thymic epithelial cells (5). Galectin-1 induces apoptosis of CD3⁻ and CD3^dim immature thymocytes, but not of mature CD3^bright thymocytes (3). The phenotype of the susceptible thymocytes implies that galectin-1 may participate in both nonselection, i.e., death of cells that are not positively selected, as well as negative selection (3, 6, 7). In the periphery, galectin-1 appears to eliminate Ag-specific T cells; in animal models of myasthenia gravis (8) and experimental autoimmune encephalitis (9), administration of galectin-1 abrogated the autoimmune response, and Ag-specific T cells could not be isolated from the lymph nodes of treated animals (9).

Elucidation of the mechanism of galectin-1-induced apoptosis is critical for understanding the role of galectin-1 in immune development and in the immune response. We have previously determined that galectin-1-induced apoptosis occurs via a mechanism distinct from that used in Fas or CD3-mediated apoptosis (4). Moreover, we have shown that the dimeric form of galectin-1 is required for induction of apoptosis (4), implying that cross-linking of cell surface receptors is involved in transducing the death signal. In the present study, we demonstrate that the initial steps in galectin-1 signaling involve binding to and segregation of specific T cell surface glycoproteins into discrete membrane microdomains. The composition of these microdomains is very different from that observed during signaling through the TCR (10, 11, 12, 13, 14, 15, 16). In addition, while mammalian lectins are known to participate in cell adhesion (17) and in phagocytosis of microorganisms and apoptotic cells (18, 19), induction of T cell apoptosis represents one of the first signaling functions identified for a mammalian lectin.

The specificity of glycoprotein receptor binding and segregation during galectin-1 signaling raises the question of how galectin-1 can discriminate among potential glycoprotein receptors, because the disaccharide ligands preferentially recognized by galectin-1 (Galβ1,3GlcNAc and Galβ1,4GlcNAc,³ type I and II lactosamine, respectively) are found on all complex N-linked and many O-linked glycoproteins (20, 21). Our results demonstrate that galectin-1 can discriminate among T cell surface glycoproteins and specifically bind to three T cell surface receptors, CD45, CD43, and CD7. Galectin-1 binding to CD45, CD43, and CD7 results in novel interactions among these glycoproteins during apoptosis.

The human T lymphoblastoid cell lines MOLT-4, CEM, and ARR (gift of Dr. Ken Weinberg, Los Angeles Children’s Hospital, Los Angeles, CA) were maintained in RPMI 1640 medium (Life Technologies, Gaithersburg, MD), 10% fetal bovine serum, and 10 mM HEPES. mAbs were obtained from the following sources: DFT1 (CD43), 2B11 (CD45), LCA (CD45), UCHL1 (CD45R0), 4KB5 (CD45RA), PD7/26 (CD45RB), UCHT1 (CD3), MT310 (CD4), DK25 (CD8), MHM24 (CD11a), DAK-G01 (mouse IgG1), and rabbit anti-mouse IgG1 from Dako (Carpinteria, CA); BL5 (CD18) and J-173 (CD44) from AMAC (Westbrook, ME); LT27 (CD27), 2H10 (CD120a), and 4D1B10 (CD120b) from Caltag Laboratories (South San Francisco, CA); FN50 (CD69), M-T701 (CD7), and DX2 (CD95) from PharMingen (San Diego, CA); T1 (CD5) from Coulter (Miami, FL); Leu-28 (CD28) from Becton Dickinson (San Jose, CA); LT7 (CD7) from Advanced Immunochemical (Long Beach, CA); DFT1-biotin (biotin-labeled CD43) from Ancell (Bayport, MN); FS-7 (CD95) from Upstate Biotechnology (Lake Placid, NY); fluorescein-conjugated goat anti-mouse IgG1 and Texas red-conjugated goat anti-mouse IgG2a from Southern Biotechnology Associates (Birmingham, AL); 9.4 (CD45) from the American Type Culture Collection (Manassas, VA); T305 (CD43), a gift from R. Fox (Scripps Clinic and Research Foundation, La Jolla, CA); PA2.6 (MHC I), a gift from A. Jewett and B. Bonavida (University of California, Los Angeles, CA); and rabbit polyclonal galectin-1 anti-serum, a gift from Incyte Pharmaceuticals (Palo Alto, CA). Saccharides were obtained from the following sources: β-lactose, α-D(+) fucose, and D(+)-mannose from Sigma (St. Louis, MO); and NeuNAcα2–6Galβ1–4GlcNAc and NeuNAcα2–3Galβ1–4GlcNAc from Oxford Glycosystems (Rosedale, NY).

The competent Escherichia coli line BL21DE3 was transformed with the plasmid which encodes galectin-1, pT7IML-1 (gift from Incyte Pharmaceuticals), and recombinant galectin-1 was produced and isolated as previously described (3). Before use, galectin-1 was dialyzed into 10 mM phosphate buffer, 140 mM NaCl (pH 7.4) (PBS), and 4 mM DTT and stored at −20°C.

Soluble, recombinant, biotin-labeled galectin-1 was preincubated with a competitive saccharide diluted in PBS, 1% BSA, and 1.2 mM DTT in a 96-well U-bottomed plate for 30 min on ice. MOLT-4 cells were washed and added to the plate for 1 h on ice. The final concentrations of biotinylated galectin-1 was 0.7 μM and saccharide inhibitor was 1 mM. The figure of 1 mM was chosen because the IC₅₀ value for lactose was determined to be 1 mM using this assay. After washing, cells were incubated with HRP-conjugated streptavidin (Zymed Laboratories, South San Francisco, CA) for 1 h on ice. Following washing, cell-associated peroxidase activity was detected using the colorimetric substrate o-phenylenediamine dihydrochloride (Sigma). The plates were read at 492 nm using a Bio-Rad (Richmond, CA) ELISA reader with data reduction software.

mAbs specific for T cell surface glycoproteins were diluted in ice cold PBS and 1% BSA (pH 7.4) (PBS-BSA) and incubated with 5 × 10⁶ MOLT-4 cells in a 96-well U-bottomed plate for 30 min on ice. Following washing, biotin-labeled galectin-1 was added to the cells for a final concentration of 0.7 μM and incubated for 1 h on ice. Unbound galectin-1 was removed and the cells were incubated with HRP-conjugated streptavidin for 1 h on ice. Following washing, cell-associated peroxidase activity was detected as described above.

Cell surface glycoproteins were biotin-labeled as previously described (22) and membrane proteins were isolated (23). Isolated membrane proteins were solubilized in 0.1% Triton X-100. Solubilized membrane proteins were applied to a galectin-1 affinity column that was prepared by coupling galectin-1 to cyanogen bromide activated Sepharose 4B (Sigma) with a final concentration of 10 mg galectin-1/ml Sepharose 4B (24). After washing with PBS, 0.1% Triton X-100, 0.02% sodium azide (wash buffer), bound proteins were eluted with 0.1 M β-lactose in wash buffer. Samples were resolved by 10% SDS-PAGE under reducing conditions, transferred to nitrocellulose, and Western blotted with HRP-conjugated streptavidin (Zymed Laboratories). Proteins were visualized using Enhanced Chemiluminescence (Amersham, Arlington Heights, IL).

Immunoprecipitations were performed as previously described, with minor modifications (25). Fractions constituting the peak of the eluate from the galectin-1 affinity column were pooled. Depending on the Ab used, the samples were precleared twice with either protein A-Sepharose or with protein A-Sepharose saturated with rabbit anti-mouse IgG1 (Sigma) for 2 h, with rotating at 4°C. The precleared supernatant was collected and incubated with 5 μg mAb, recognizing the T cell surface glycoprotein of interest, or 5 μg of an isotype matched control, overnight at 4°C. To capture the Ab-protein complexes, protein A-Sepharose or protein A-Sepharose saturated with rabbit anti-mouse IgG1 was added and incubated for 2 h at 4°C with rotating. The beads were collected by centrifugation, washed, and boiled in SDS-PAGE sample buffer for 5 min. The samples were separated by SDS-PAGE under reducing conditions, transferred to nitrocellulose, and Western blotted as described above.

A total of 1 × 10⁷ MOLT-4 cells were washed and treated with 20 μM galectin-1, 1.2 mM DTT, or 1.2 mM DTT alone for 10 min on ice and 20 min in a 37°C water bath and cooled on ice. The cells were solubilized with 0.1% digitonin in PBS with protease inhibitors (solubilization buffer) for 1 h on ice. Insoluble extract was removed by centrifugation at 1500 rpm for 10 min. The supernatants were precleared twice with protein A-Sepharose for 2 h at 4°C with gentle rocking and incubated with 25 μl of galectin-1 antiserum for 2 h at 4°C with gentle rocking. The beads were collected and washed three times with solubilization buffer. Proteins bound to galectin-1 were dissociated with 0.1 M β-lactose/solubilization buffer and immunoprecipitated using Abs against the glycoproteins of interest as described above.

A total of 1 × 10⁷ MOLT-4 cells or human thymocytes were incubated with 20 μM galectin-1 and 1.2 mM DTT, or 1.2 mM DTT alone as a control, for 10 min on ice followed by 20 min in a 37°C water bath to allow migration of counterreceptors on the cell surface. Cells were cooled to 4°C, and bound galectin-1 was dissociated with ice cold 0.1 M β-lactose. Paraformaldehyde (2%) was added for 30 min on ice to fix the cells, and the reaction was quenched with 0.2 M glycine for 5 min on ice. The cells were stained by incubating for 1.5 h at 25°C with 15 μg/ml each of the indicated combination of two Abs: 9.4 (CD45, mouse IgG2a), DFT1 (CD43, mouse IgG1), DFT1-biotin (CD43, mouse IgG1-biotin conjugated), M-T701 (CD7, mouse IgG1), LT7 (CD7, mouse IgG2a), or UCHT1 (CD3, mouse IgG1). The cells were washed and incubated at 25°C for 1.5 h with the appropriate secondary reagents: fluorescein-conjugated goat anti-mouse IgG1, fluorescein-conjugated goat anti-mouse IgG2a, Texas red-conjugated goat anti-mouse IgG2a, fluorescein-conjugated streptavidin, Texas red conjugated-streptavidin, or goat anti-mouse IgG1. The cells were washed and, when required, a tertiary reagent, fluorescein-conjugated rabbit anti-goat IgG, was added for 1.5 h at 25°C. The cells were washed, dropped on glass microscope slides, and mounted using Prolong Anti-fade mounting media (Molecular Probes, Eugene, OR). The fluorescently labeled cells were analyzed using the ×100 objective on a Leica (Deerfield, IL) CLSM confocal laser scanning microscope. The cells were scanned by dual excitation of fluorescein (green) and Texas red (red) fluorescence. Dual emission fluorescent images were collected in separate channels at 0.5 μm optical slices. The images were processed on a Sun Workstation using AVS (Advanced Visualization Systems, Waltham, MA) image processing software. Areas of red and green overlapping fluorescence were represented with a yellow signal. The confocal images were printed on a Fuji (Tokyo, Japan) Pictography 3000 printer.

A total of 1 × 10⁷ MOLT-4 or CEM cells were treated with 20 μM galectin-1 and 1.2 mM DTT, or 1.2 mM DTT as a control or 0.5 μg/ml FS-7 (CD95, IgM) for 20 min at 37°C as described above. Before fixation, the cells were resuspended in 1 ml binding buffer A (10 mM HEPES (pH 7.4), 150 mM NaCl, 2.5 mM CaCl₂, and 1 mM MgCl₂) and 5 μg biotin-conjugated Annexin V (Calbiochem, Cambridge, MA) and incubated on ice for 10 min. Paraformaldehyde (2%) was added for 30 min on ice, followed by quenching with 0.2 M glycine and calcium-free PBS. All washes used calcium-free PBS to assure that any unbound Annexin V could not bind to the cells during subsequent steps because Annexin V binding requires calcium. The mAb 9.4 (CD45, mouse IgG2a) was added for a 1.5-h incubation, 25°C. The cells were washed and incubated with fluorescein-conjugated anti-biotin Ab (mouse IgG1) for 1.5 h at 25°C. Following washing, the cells were incubated for 1.5 h with Texas red-conjugated goat anti-mouse IgG2a. Following a wash with calcium-free PBS, the cells were dropped on a slide and mounted with Prolong Anti-fade mounting medium. The fluorescently labeled cells were processed and analyzed as described above.

A total of 1 × 10⁷ MOLT-4 cells were treated with galectin-1 as described above and incubated at 37°C for indicated time points. Following incubation, the cells were washed with 0.1 M β-lactose to dissociate cell aggregates. The samples were resuspended in 1 ml binding buffer B (10 mM HEPES (pH 7.4), 150 mM NaCl, 5 mM KCl, 1 mM MgCl_2, and 1.8 mM CaCl₂) containing 5 μg fluorescein-conjugated Annexin V and 2.5 × 10⁻⁴ μg propidium iodide (R&D Systems, Minneapolis, MN), and incubated on ice for 10 min. The samples were immediately analyzed by flow cytometry. Flow cytometry data was acquired using a Becton Dickinson FACScan and analyzed using CellQuest software.

To identify candidate glycoprotein receptors that bind galectin-1 on the T cell surface, we screened a large panel of mAbs to identify Abs that would block galectin-1 binding to MOLT-4 human T cells. A limited screen had previously determined that Abs recognizing CD45 and CD43 inhibited galectin-1 binding to T cells (4, 5). In this experiment, we included Abs to adhesion and homing receptors, developmentally regulated T cell surface molecules, costimulatory molecules, and members of the TNF receptor family. Table I demonstrates that Abs recognizing CD45, CD43, CD8, CD7, CD4, and CD3 inhibited galectin-1 binding to MOLT-4 cells. Abs to the adhesion molecules CD44, CD11a, and CD18 did not inhibit galectin-1 binding, nor did Abs to the TNF receptor family members CD95, CD120a, CD120b, or CD154 (Table I).

To isolate galectin-1 binding glycoproteins directly, membrane proteins from MOLT-4 cells were isolated and extracted with the detergent Triton X-100; this detergent was chosen to dissociate complexes of membrane glycoproteins, so only glycoproteins that bound directly to galectin-1 would be isolated. The membrane extract was subjected to galectin-1 affinity chromatography, and bound glycoproteins were isolated by elution with lactose. Five major bands (200, 180, 95, 60 and 45 kDa) and four minor bands (40, 28, 20, 16 kDa) were isolated (Fig. 1, A and C). Based on the relative mobilities of the bands and the Ab inhibition data described above (Table I), the relevant Abs were used to immunoprecipitate and identify the eluted glycoproteins.

The 200- and 180-kDa bands were identified as isoforms of CD45 (CD45RA and CD45R0, respectively). The 95-kDa band was identified as CD43, and the 60-kDa band as CD4 (Fig. 1,A). The 40-kDa band was identified as CD7 by immunoprecipitation (Fig. 1,B). The 45-kDa band was identified as actin by Western blotting (data not shown); galectin-1 has previously been shown to bind actin (26). Three minor bands of 28, 20, and 16 kDa were also isolated. The size of these three bands corresponded to the size of components of the CD3 complex (27), and an Ab to CD3 inhibited galectin-1 binding to MOLT-4 cells (Table I), suggesting that these bands were the components of the CD3 complex. However, we could not immunoprecipitate these bands from the galectin-1 column eluate because the CD3 Ab did not recognize the dissociated complex. To confirm that these bands represented the members of the CD3 complex, galectin-1 binding proteins were isolated from another T cell line, ARR, which does not express CD3 (data not shown). We compared the pattern of bands isolated from the MOLT-4 and ARR cell lines. Fig. 1 C demonstrates that the 28-, 20-, and 16-kDa bands were not isolated from the CD3⁻ ARR cells, indicating that the low m.w. bands isolated from MOLT-4 cells were components of the CD3 complex. Thus, all of the galectin-1 binding proteins isolated from MOLT-4 cells were identified, using Abs that blocked galectin-1 binding to the cell surface.

Although an Ab to CD8 also inhibited galectin-1 binding (Table I), CD8 was not isolated by galectin-1 affinity chromatography. A possible explanation for this discrepancy is that CD8 associates with CD45 on the T cell surface (27), so that steric hindrance by the CD8 Ab may have blocked galectin-1 binding to CD45 on the cell surface.

Although six potential receptors were identified by Ab blocking experiments, and five glycoproteins bound to a galectin-1 affinity column, previous data demonstrated that all of these glycoproteins are not required for galectin-1-induced apoptosis (3, 4). We have demonstrated that the CD3⁻CD4⁻CD8⁻ ARR cell line is susceptible to galectin-1 (4), as are CD3⁻ human thymocytes (3). Only three galectin-1 binding proteins were isolated from ARR cells, CD45R0, CD43, and CD7 (Fig. 1 D). Because the ARR cell line is susceptible to galectin-1, these data suggested that CD4, CD3, and CD8 are not involved in the death signal and, at most, CD45, CD43, and CD7 are directly involved in delivering the death signal following galectin-1 binding.

Soluble glycoproteins may be more accessible to galectin-1 on an affinity matrix than on the cell surface. Therefore, to identify the glycoproteins that bound galectin-1 on the cell surface, we coimmunoprecipitated galectin-1 and the candidate receptors that had bound galectin-1 from MOLT-4 cells. Only CD45, CD43, and CD7 coimmunoprecipitated with galectin-1 (Fig. 1,E), whereas CD3, CD4 (Fig. 1,E) and CD8 (data not shown) did not (Fig. 1 E). Although Abs recognizing CD3 and CD4 inhibited galectin-1 binding to the cell surface in the Ab inhibition assay, these glycoproteins did not coimmunoprecipitate with galectin-1 from the cell surface. As mentioned above for CD8, CD4 and CD3 also associate with CD45 on the cell surface (28), so that steric hindrance by Abs recognizing CD4 and CD3 may have blocked galectin-1 binding to CD45. In addition, although CD4 and CD3 may have oligosaccharide chains that are recognized by galectin-1 in solution, these oligosaccharides may not be accessible to galectin-1 on cell surface CD4 and CD3. Cell surface CD45 and CD43 are large, extended molecules that may be more accessible to galectin-1 (29, 30). The coimmunoprecipitation experiment indicated that CD45, CD43, and CD7 are the either the primary or the most accessible glycoprotein receptors for galectin-1 on the cell surface, implying that galectin-1 death involves one or more of these three glycoproteins.

The identification of three different glycoproteins that bound galectin-1 on the cell surface raised the question of how these glycoproteins interact to deliver the death signal. We previously determined that receptor cross-linking by dimeric galectin-1 was required to deliver the apoptotic signal (4). The different glycoproteins could be cross-linked via galectin-1 into a heterotypic complex, or galectin-1 could cross-link only identical glycoprotein receptors into a homotypic complex (31, 32). To examine the interaction of the glycoproteins following galectin-1 binding, cells were treated with galectin-1 for 20 min at 37°C. This length of time was chosen because we have previously demonstrated that the death signal is irreversible after this period (4), even if galectin-1 is removed from the cell surface (4). After 20 min, the cells were cooled to prevent further changes, galectin-1 was removed from the cell surface with lactose, and the cells were fixed. Because our results indicated that CD45, CD43, and CD7 were important for delivering the apoptotic signal, Abs to CD45, CD43, CD7 were added, and the cell surface localization of the glycoproteins was determined by confocal microscopy. In addition, although CD3 is not required for galectin-1-induced apoptosis, CD3 cross-linking augments galectin-1-induced death (3, 33), so the localization of CD3 was examined as well.

As shown in Fig. 2, galectin-1 binding induced a dramatic redistribution of CD45 and CD43 on the surface of MOLT-4 T cells. Before galectin-1 binding, there was a uniform distribution of CD45 (red) and CD43 (green) on the cell surface, with occasional colocalization (yellow) (Fig. 2,A). On galectin-1-treated cells, CD45 localized into one or two large islands on the cell surface, virtually excluding CD43. CD43 aggregated into multiple small clusters that segregated away from CD45 (Fig. 2). Thus, the binding of galectin-1 to T cells induced redistribution and segregation of CD45 and CD43, suggesting that this process was important for inducing apoptosis.

Because CD7 also bound galectin-1, it was possible that CD7 associated with CD45 or CD43, or that CD7 segregated from the other two glycoproteins following galectin-1 binding. Fig. 3,A demonstrates that, before galectin-1 treatment, CD45 (red) and CD7 (green) were also uniformly distributed over the cell surface, with some regions of colocalization (yellow). Following galectin-1 treatment, CD45 again segregated into one or two large islands on each cell, and CD7 aggregated into multiple small clusters that segregated away from CD45. This pattern was identical to that observed for CD45 and CD43. Indeed, before galectin-1 treatment, CD7 (green) colocalized with CD43 (red) (Fig. 3,B). Following galectin-1 treatment, CD43 and CD7 remained associated and moved into larger aggregates (Fig. 3 B). These data indicate that CD43 and CD7 may act as a complex during the delivery of the galectin-1 death signal. It is important to note that images of galectin-1-treated and control cells were collected by confocal microscopy under exactly the same conditions, although the staining of CD43 and CD7 appears brighter on the galectin-1-treated cells. This is likely due to the movement of CD43 and CD7 into larger and subsequently brighter clusters.

As described above, although CD3 did not appear to directly bind galectin-1 on the cell surface, CD3 signaling augments the galectin-1 death signal in both human and murine thymocytes (3, 33). In addition, we have previously found that only immature CD3⁻ and CD3^dim thymocytes, but not mature CD3^bright thymocytes, are susceptible to galectin-1 (3). Therefore, the localization of CD3 was examined on human thymocytes before and after galectin-1 treatment, to determine the localization of CD3 after galectin-1 binding, and to determine whether the redistribution and segregation of glycoproteins that we observed correlated with susceptibility to galectin-1.

Galectin-1 binding to thymocytes resulted in the same pattern of receptor segregation that we have observed with MOLT-4 T cells. We first examined the localization of CD3 and CD45 on both mature CD3^bright and immature CD3^dim thymocytes. The differences in fluorescence intensity of CD3 on these two cell populations were readily distinguishable by confocal microscopy. In addition, the number of CD3^bright cells determined by confocal microscopy was approximately equivalent to the number of CD3^bright or single positive cells determined by flow cyometry (3). On mature CD3^bright thymocytes, CD3 (green) and CD45 (red) were colocalized and uniformly distributed over the cell surface before galectin-1 binding (Fig. 4 A). On immature CD3^dim thymocytes, the weak CD3 staining that was detectable was also distributed around the cell surface and largely colocalized with CD45. A striking difference was observed between CD3^dim and CD3^bright thymocytes after galectin-1 treatment. On the immature CD3^dim cells, galectin-1 binding resulted in the movement of all CD45 and CD3 into one or two islands on the cell surface, exactly as we observed with MOLT-4 cells. However, on CD3^bright cells, the mature thymocyte subset that is not susceptible to galectin-1 (3), no significant movement of CD45 or CD3 was seen.

A similar pattern was seen when the localization of CD7 (red) and CD3 (green) was examined on CD3^dim and CD3^bright thymocytes (Fig. 4 B). Before galectin-1 binding, immature CD3^dim thymocytes had a patchy distribution of CD3 and CD7 on the cell surface (bottom two cells in field), while mature CD3^bright thymocytes showed an intense, uniform distribution of CD3 and small clusters of CD7 (top cell in field). Following galectin-1 binding, CD3 and CD7 segregated on the immature CD3^dim cells, with the CD3 moving into one large island and the CD7 aggregating into small clusters. However, on the mature CD3^bright cells, no significant movement of CD3 and CD7 was seen. These data demonstrate that galectin-1 induces the same type of receptor segregation on thymocytes as we observed with a human T cell line. Importantly, this receptor segregation occurred only on the immature CD3^dim subset of thymocytes that we have previously shown to be susceptible to galectin-1 (3) and not on the mature CD3^bright subset.

The absence of receptor segregation on mature thymocytes following galectin-1 binding suggested that alterations in the saccharide profile of these cells may have prevented galectin-1 binding. We have demonstrated that CD45 on mature human thymocytes differs from CD45 on immature human thymocytes by being terminally substituted with a SAα2,6Galβ1,4GlcNAc sequence (34, 35). Previous studies of galectin-1 binding specificity for soluble oligosaccharides have shown that galectin-1 can bind lactosamine terminally substituted with sialic acid in an α2,3 linkage, but not in an α2,6 linkage (36); structural studies of galectin-1 revealed that α2,6-linked sialic acid would distort the binding pocket and prevent binding of the lactosamine sequence (36). We examined whether α2,6-linked sialic acid would prevent the binding of galectin-1 to the T cell surface. Table II shows that α2,3 sialyllactose (SAα2,3Galβ1,4GlcNAc) effectively competed for the binding of galectin-1 to MOLT-4 cells, with a potency equivalent to lactose. However α2,6 sialyllactose (SAα2,6Galβ1,4GlcNAc) failed to compete for galectin-1 binding, suggesting that the presence of α2,6-linked sialic acid on CD45 of mature thymocytes would not allow galectin-1 binding and may account for the resistance of these cells to galectin-1.

Although the images in Fig. 4 demonstrated that receptor segregation occurred only on thymocyte subsets that are susceptible to galectin-1, these results did not demonstrate that cell death occurred during this process. To verify that cells that underwent receptor redistribution and segregation also underwent apoptosis, we treated MOLT-4 cells with galectin-1 for 20 min at 37°C and examined the movement of CD45 and the binding of Annexin V, which binds to externalized phosphatidylserine on apoptotic cells. By confocal microscopy, ∼70–80% of cells treated with galectin-1 at 37°C for only 20 min bound Annexin V. Importantly, Annexin V bound only to galectin-1-treated cells that demonstrated movement and patching of CD45 (Fig. 5,A). Galectin-1-treated cells that did not display movement and patching of CD45 did not bind Annexin V (Fig. 5,A). We also examined the localization of CD45 on CEM cells that were triggered to undergo apoptosis by incubation with anti-Fas Ab for an identical time period. CEM cells are susceptible to both Fas and galectin-1-induced apoptosis, whereas MOLT-4 cells are not (4). As we observed with MOLT-4 cells, galectin-1 treatment of CEM cells resulted in the redistribution of CD45 into large patches on the surface of apoptotic cells (Fig. 5,B). In contrast, on anti-Fas treated cells, CD45 remained uniformly distributed on the surface of both apoptotic and nonapoptotic cells (Fig. 5 C).

To our surprise, Annexin V bound only to the large patches of CD45 on the surface of galectin-1 treated MOLT-4 and CEM apoptotic cells (Fig. 5 A and B). These data revealed that movement of phosphatidylserine to the external plasma membrane leaflet occurred in discrete microdomains on the cell surface, rather than occurring globally during the process of galectin-1-induced apoptosis. A similar morphologic appearance of phosphatidylserine exposure on apoptotic blebs has been observed to occur on HeLa cells induced to undergo apoptosis by nutrient deprivation for 16 h (37). The exposure of phosphatidylserine in discrete microdomains on apoptotic blebs suggests that the scramblase, which has been shown to be responsible for exposure of phosphatidylserine exposure on apoptotic lymphocytes (38, 39), may be locally activated.

Annexin V binding was readily visible by confocal microscopy on ∼70–80% of cells that had been treated with galectin-1 for only 20 min at 37°C (Fig. 5, A and B). These findings indicated that the local externalization of phosphatidylserine may be a relatively early event in the process of galectin-1-induced apoptosis. This period of time is very brief relative to the time required to detect Annexin V binding by flow cytometry in apoptosis triggered by other agents (40). However, a similar rapid time course of phosphatidylserine exposure has been reported in cells that express high levels of a phospholipid scramblase (39). To determine the duration of galectin-1 treatment that was required to detect phosphatidylserine exposure by flow cytometry, we performed a kinetic analysis (Fig. 6). Incubation of MOLT-4 cells with galectin-1 for only 20 min at 37°C resulted in Annexin V binding to the majority of cells (Fig. 6), correlating with Annexin V binding on 70–80% of cells by confocal microscopy. The incremental shift in Annexin V binding intensity with increased galectin-1 incubation time may be due to the initial exposure of phosphatidylserine on discrete membrane microdomains, followed by a more global or uniform exposure of phosphatidylserine at later time points.

To further examine the domains where CD45 and Annexin V colocalized, we performed three-dimensional image reconstruction of galectin-1 treated cells (Fig. 7,A). These images revealed that the islands of CD45 were localized on large membrane blebs protruding from the cell surface. Little CD45 was visible on other areas of the cell surface. The formation of membrane blebs is a morphologic hallmark of apoptosis. Fluorescent microscopy of galectin-1-treated cells also demonstrated that Annexin V bound to blebs protruding from the cell surface, indicating that phosphatidylserine exposure happened first on membrane blebs during galectin-1-induced apoptosis (Fig. 7 B). Thus, galectin-1 binding to T cells results in the rapid redistribution and segregation of specific membrane glycoproteins, with a complex of CD45 and CD3 localized to the apoptotic blebs, and a complex of CD43 and CD7 localized in regions excluded from the apoptotic blebs. Moreover, we have observed rapid colocalization of CD45 with externalized phosphatidylserine during galectin-1-induced apoptosis, although it is not yet clear if CD45 relocalization and phosphatidylserine exposure occur in a specific temporal sequence.

Galectin-1 binds CD45, CD43, and CD7. Identification of these receptors and characterization of their interactions is critical for understanding the novel mechanism of apoptosis induction triggered by galectin-1. In addition, interaction of these receptors in apoptosis has not previously been described. Importantly, the ability of galectin-1 to specifically recognize CD45, CD43, and CD7 among all of the lactosamine-bearing glycoproteins on the T cell surface demonstrates that receptor recognition by galectin-1 relies on both the presence of the lactosamine ligand and on the presentation of lactosamine to galectin-1 by a specific protein backbone. Proper presentation and clustering of oligosaccharide ligands on multiple carbohydrate side chains would increase the binding avidity of galectin-1 for a particular glycoprotein (41, 42). Because CD45 and CD43 possess numerous carbohydrate side chains spaced closely together on the protein backbone (29, 30), these two glycoproteins would present many clustered patches of lactosamine sequences to galectin-1, allowing preferential binding of galectin-1 to these two glycoproteins. Restricted specificity for glycoprotein receptors has been found for several mammalian lectins. In placenta, galectin-1 binds to fibronectin and laminin (43), and galectin-3 binds to lysosomal-associated membrane proteins 1 and 2 on human melanoma cells (44). The selectins, another lectin family that recognizes a common saccharide ligand, also recognize a restricted set of specific glycoproteins that bear clustered oligosaccharide chains on target cells (42).

We have shown that cross-linking of galectin-1 receptors is essential for triggering galectin-1-induced apoptosis, as monomeric galectin-1 does not have apoptotic activity (4). In the present study, galectin-1 binding to human thymocytes resulted in the redistribution and segregation of CD45 together with CD3, and of CD43 together with CD7. These results demonstrated that galectin-1 cross-linking could separate a mixed population of glycoproteins on the cell surface into unique complexes in specific membrane microdomains. Brewer and coworkers (31, 32, 45, 46) have demonstrated that bovine galectin-1 can form ordered homotypic complexes composed of identical glycoproteins out of a mixture of glycoproteins in solution. The formation of homotypic complexes is thermodynamically driven, suggesting that the formation of such homotypic matrices may also occur on the surface of cells (31). In solution studies, the ability of galectin-1 cross-linking to form homotypic or heterotypic complexes of soluble glycoproteins relied solely on the number of oligosaccharide ligands per glycoprotein that could bind galectin-1, termed the valency of the glycoprotein. However, on the cell surface, the glycoprotein valency is not the only factor that will control the interaction and movement of galectin-1 receptors. The structure of the glycoprotein, protein-protein interactions of glycoproteins above and below the plasma membrane, and association with other cytoplasmic proteins will also determine how glycoproteins interact following galectin-1 binding.

As mentioned earlier, protein structure and conformation will contribute to the presentation of clustered lactosamine residues, promoting galectin-1 binding. In addition, protein structure will influence the ability of glycoproteins to pack into an ordered cross-linked lattice. In solution studies, the formation of lattice structures required identical repeating units of glycoproteins (31, 45, 47). The extended rod-like conformation of both CD43 and CD7, created by the high density of O-glycans on the peptide backbone (47, 48), may facilitate the packing of these two glycoproteins into a lattice on the cell surface.

Protein-protein interactions may also regulate the movement and segregation of glycoproteins by galectin-1. Before galectin-1 binding, CD45 and CD3 (Fig. 4,A) and CD43 and CD7 (Fig. 3,B) were colocalized on the cell surface. The complex of CD45 and CD3 may act as a single glycoprotein receptor, associating via protein-protein interactions before galectin-1 cross-linking. Galectin-1 binding would cross-link and segregate repeating units of CD45 associated with CD3 into a “homotypic” complex (Fig. 8). The same may be true for CD43 and CD7 (Fig. 8). In contrast, CD7 and CD3 co-localize before galectin-1 binding (Fig. 4,B; Ref. 49). However, CD7 and CD3 separate into apparently more stable complexes with CD43 and CD45, respectively, after galectin-1 binding. The same was seen for CD45 and CD7 (Fig. 3 A), which are known to associate on the T cell surface in the absence of galectin-1 (50). These findings demonstrate that galectin-1 binding results in selective maintenance of some pre-existing glycoprotein associations and disruption of others. Moreover, galectin-1 binding results in the association of glycoproteins, such as CD3 and CD45, that are segregated during TCR interactions with an Ag-MHC complex (10, 12). This demonstrates that galectin-1 binding creates a unique set of membrane microdomains distinct from those created during other T cell signaling events (10, 11, 12, 13, 14, 15, 16).

Interactions of the intracellular domains of the galectin-1 receptors with other molecules may promote or restrict the ability of different glycoproteins to cross-link and segregate. The cytoplasmic tail of CD45 associates with the cytoskeletal protein fodrin (50), whereas CD43 is associated with the cytoskeletal protein radixin (51) (Fig. 8). Cytoskeletal changes during death may influence the movement and segregation of CD45 and CD43 after galectin-1 binding. Alternatively, disruption of cytoskeletal integrity resulting from the movement of CD45 and CD43 may be involved in triggering galectin-1 death, as has been described for endothelial cells (52). Alterations in both fodrin and radixin have been noted in apoptotic cells (53, 54), suggesting that these proteins may play a role in the apoptotic pathway mediated by galectin-1.

It is not yet clear how CD45, CD43, and CD7 are involved in galectin-1-induced apoptosis. Both CD45 and CD43 have been implicated in triggering T cell apoptosis (55, 56, 57). Ab cross-linking of a specific glycoform of CD43 resulted in apoptosis of the Jurkat cell line (55). This apoptosis-inducing Ab specifically recognized a GalNAc-Ser containing epitope (55); because this monosaccharide is not a ligand for galectin-1, the relationship between this Ab and galectin-1 in triggering cell death is not yet clear. We are currently determining the role of CD43 in galectin-1-induced apoptosis. Although Ab cross-linking of CD45 has been shown to induce a pro-apoptotic signal in human T cells (56, 57), CD45 has also been shown to act as a negative regulator of T cell apoptosis (58). This may be relevant to galectin-1-induced apoptosis. Walzel et al. (59) have demonstrated that galectin-1 treatment of Jurkat T cells resulted in decreased phosphatase activity of immunoprecipitated CD45. In addition, we have recently found that protein tyrosine phosphatase inhibitors that block CD45 protein tyrosine phosphatase activity enhanced galectin-1-induced T cell death (J. T. Nguyen and L. G. Baum, unpublished data). These data support a model in which the galectin-1-induced CD45 cross-linking and sequestration that we have observed would functionally remove and/or inhibit the protein tyrosine phosphatase domain of CD45 from a pro-apoptotic signaling complex. Because galectin-1 binding to T cells results in tyrosine phosphorylation (33), segregation of the CD45 phosphatase may allow a kinase-dependent signal to be transduced.

CD7 is an intriguing candidate for a role in galectin-1-induced apoptosis. Galectin-1 is the first ligand to be identified for CD7, although previous studies have suggested that the CD7 ligand may be a lectin-like molecule (60). CD7 associates with the PI3 kinase, and ligation of CD7 results in tyrosine phosphorylation (49), as does galectin-1 binding (33). Moreover, CD7 is highly expressed on human thymocytes and activated T cells, populations that are susceptible to galectin-1 (3, 4). We have found that a CD7⁻ human T cell line, HUT78, is resistant to galectin-1-induced apoptosis (data not shown), suggesting that CD7 is a necessary component of the galectin-1 death pathway. The galectin-1 death pathway may depend on concomitant signals from CD45, CD43 and CD7 to trigger death, signals that may result directly from the redistribution and segregation of these glycoproteins induced by galectin-1 binding.

The segregation of CD45, CD3, CD43, and CD7 into discrete complexes on the T cell surface may represent a paradigm for a variety of complex signaling pathways that involve matrix formation of cell surface receptors, as has been shown to occur during T cell activation (10, 11, 12, 13, 14, 15, 16). In addition to inducing apoptosis, galectin-1 is capable of exerting a mitogenic or cytostatic effect on cells, depending on the type and functional state of the target cells (61, 62, 63, 64). Subtle changes in receptor expression, glycosylation, stoichiometry, or cross-linking may determine which of these various effects are exerted by galectin-1.

We thank James Chang for assistance with the confocal microscope; O. Witte, M. Pang, M. Galvan, and T. Prigozy for helpful discussions; and F. Brewer, M. Kronenberg, and R. Modlin for critical reading of the manuscript.