Sialic acid specificity of myelin-associated glycoprotein binding.

Myelin-associated glycoprotein (MAG), a nervous system cell adhesion molecule, is an I-type lectin that binds to sialylated glycoconjugates, including gangliosides bearing characteristic structural determinants (Yang, L. J.-S., Zeller, C. B., Shaper, N. L., Kiso, M., Hasegawa, A., Shapiro, R. E., and Schnaar, R. L. (1996) Proc. Natl. Acad. Sci. U. S. A. 93, 814-818). Two cell adhesion systems, COS-1 monkey kidney fibroblasts transiently transfected to express MAG and Chinese hamster ovary (CHO) cells stably transfected to express MAG, were used to probe the structural specificity of MAG-ganglioside binding. Both cell types bound to the same gangliosides: GQ1bα (IV3NeuAc,III6NeuAc,II3(NeuAc)2-Gg4Cer) > GT1b = GD1a > GM3 > GM1, GD1b, and GQ1b (the latter do not support adhesion). Binding was enhanced by pretreatment of MAG-expressing cells with neuraminidase. MAG-expressing Chinese hamster ovary cells bound directly to gangliosides resolved on thin layer chromatograms, allowing detection of MAG binding species in a mixture. The simplest ganglioside ligand for MAG was GM3 bearing N-acetylneuraminic acid, whereas GM3 bearing N-glycolylneuraminic acid did not support adhesion. Chemical modifications of N-acetylneuraminic acid residues (on GD1a) abrogated MAG binding. Mild periodate oxidation of sialic acids to their corresponding seven-carbon (or eight-carbon) sialic acid aldehydes abolished MAG binding, as did further conversion to the corresponding primary alcohols. Eliminating the anionic charge by ethyl esterification, amidation, or reduction also abolished MAG-mediated cell adhesion. These data demonstrate that MAG-ganglioside binding is highly specific and defines key carbohydrate structural determinants for MAG-mediated cell adhesion to gangliosides.

mains, a single transmembrane domain, and a short cytoplasmic tail (1)(2)(3)(4). It is a minor component of myelin, comprising 1% of central nervous system and 0.1% of peripheral nervous system myelin proteins (5). Based on its periaxonal location (6) and its in vitro adhesion properties (7,8), MAG has been hypothesized to play a role in myelin-axon interactions. Although transgenic MAG-knockout mice are able to myelinate, they have altered periaxonal architecture (9,10).
A second function proposed for MAG is in the inhibition of central nervous system neurite outgrowth (axon regeneration). Neurite outgrowth from neurons plated on a substratum of MAG-expressing Chinese hamster ovary (CHO) cells was significantly inhibited compared with neurons plated on control cells (11). Similarly, NG108 -15 (neuroblastoma-glioma hybrid) cells did not extend neurites onto a substratum adsorbed with myelin or MAG (12). Myelin from MAG-deficient transgenic mice is less inhibitory to NG108 -15 cell neurite outgrowth than is myelin from control mice (13), although the two preparations inhibit neurite outgrowth equally from some peripheral nervous system neurons in vitro. In vivo, transgenic mice lacking the MAG gene demonstrate enhanced axon regeneration in a lesion-induced axon regrowth model (14). Current data are consistent with the presence of multiple myelin inhibitors of neurite outgrowth (15), of which MAG is one. Complementary nerve cell surface ligands for myelin inhibitors of neurite outgrowth have yet to be identified. The discovery that MAG belongs to the family of sialic acid binding lectins termed sialoadhesins (16) or I-type lectins (17) led us to investigate the role of gangliosides, prominent neural cell surface sialoglycoconjugates, as MAG ligands (18).
Gangliosides, sialic acid-containing glycosphingolipids, are highly enriched in the nervous system (19). Their prominent location on the outer leaflet of the plasma membrane and their structural diversity lead to the hypothesis that gangliosides are involved in neural cell adhesion (20,21). If this is the case, gangliosides may interact with complementary binding proteins (lectins) to initiate cell-cell recognition. In support of this hypothesis, endogenous ganglioside receptors have been revealed in the nervous system and elsewhere (22,23). Mammalian ganglioside-binding proteins have distinct structural specificities for their carbohydrate targets. One of the most highly specific ganglioside binding lectins is MAG (18). The current study defines conditions for highly sensitive detection of MAGganglioside adhesion and describes fine structural requirements for MAG binding, with special emphasis on sialic acid substructure.
GD1a Derivatives-The structures of the sialic acid derivatives used in this article are presented in Fig. 2. The glycerol chains of GD1a sialic acids were oxidized to the 7-and 8-aldehyde forms and then reduced to the corresponding truncated primary alcohols as described (31). Briefly, GD1a (1 mg) was treated with 1 ml of ice-cold 150 mM NaCl, 50 mM sodium phosphate, pH 7.4, containing 2.5 mM sodium periodate for 90 min. The aldehyde products were purified by reverse phase chromatography (using a Sep-Pak C 18 cartridge; Millipore Corp., Milford, MA) as described (26). A portion of the resulting GD1a-aldehyde was dissolved in 1 ml of 50 mM sodium bicarbonate containing 10 mM sodium borohydride and incubated for 2 h at 37°C. The resulting products were purified by reverse phase chromatography (26). The sialic acids on 1 mol of GD1a were converted to the corresponding ethyl esters by incubation for 1 h at ambient temperature in 0.6 ml Me 2 SO/iodoethane (5:1) (32). The resulting diester was purified by DEAE-Sepharose (acetate form) chromatography in methanol. The product in the unretained eluent was further purified by reverse phase chromatography (26). One portion of the GD1a diethyl ester was reduced to the corresponding dialcohol in 1% sodium borohydride in methanol for 2 h. The product was purified by DEAE-Sepharose and reverse phase chromatography. Another portion of the GD1a ethyl ester was converted to the corresponding diamide by treatment with 4 M ammonium hydroxide in methanol/water (5:2) for 12 h at ambient temperature. The product was purified by reverse phase and DEAE-Sepharose chromatography. Products were analyzed by TLC and fast atom bombardment-mass spectrometry (33) at the Middle Atlantic Mass Spectrometry Laboratory.
Cell Culture and Transfection-COS-1 cells were maintained in 10% fetal calf serum in Dulbecco's modified Eagle's medium (DMEM) at 37°C in a humidified atmosphere of 90% air/10% CO 2 . Transient COS-1 transfections with an expression plasmid containing cDNA for the long form of MAG (pCDM8-MAG; Ref. 18) and control plasmid (MAG in the reverse orientation) were performed using DEAE-dextran (40 g/ml) as described (34). Cells were harvested for adhesion experiments (see below) 48 h after transfection.
CHO cells were stably transfected using the plasmid pSJL into which cDNA for the long form of MAG was subcloned in the forward (subclone MAG1) or reverse (control) direction as described previously (11). CHO cells were maintained in DMEM supplemented with 10% dialyzed fetal bovine serum, 2 mM glutamine, 347 M proline, 3 M thymidine, and 100 M glycine. Twelve hours before harvesting for adhesion experiments (see below), 100 nM CdCl 2 was added to the CHO growth medium to stimulate expression via an upstream metallothionein promoter.
Microplate Cell Adhesion to Adsorbed Glycolipids-Adhesion was performed as reported previously (18,35). Briefly, a 50-l aliquot of ethanol/water (1:1) containing gangliosides (concentrations as indicated), phosphatidylcholine (0.5 M), and cholesterol (2.0 M) was added to microwells (96-well Serocluster; Costar, Cambridge, MA). Plates were incubated 90 min uncovered at ambient temperature to allow evaporation and lipid adsorption (36,37), after which the wells were washed with water. Wells were preblocked by addition of 100 l/well Hepesbuffered DMEM containing 1.5 mg/ml bovine serum albumin (BSA) when using COS cells or 5 mg/ml BSA when using CHO cells. Plates were covered and incubated 10 min at 37°C prior to cell addition (see below).
Transfected COS cells were harvested using hypertonic Ca 2ϩ -and Mg 2ϩ -free phosphate-buffered saline (PBS) containing 1 mM EDTA as described (18), collected by centrifugation, and resuspended at 10 7 cells/ml in Dulbecco's PBS containing 2 mg/ml BSA. Transfected CHO cells were harvested using 0.53 mM EDTA in Ca 2ϩ -and Mg 2ϩ -free Dulbecco's PBS, collected by centrifugation, and resuspended at 4 ϫ 10 6 cells/ml in PBS containing 2 mg/ml BSA. Aliquots of cells (500 l) were placed in 1.5-ml microfuge tubes, and Vibrio cholerae neuraminidase (Calbiochem) was added (10 milliunits/500 l or as indicated). Suspensions were incubated for 2 h at 37°C with end-over-end mixing. Cells were collected by centrifugation and resuspended at 250,000 cells/ml (for COS cells) or 500,000 cells/ml (for CHO cells) in Hepes-buffered DMEM containing 1.5 mg/ml BSA (for COS cells) or 5 mg/ml BSA (for CHO cells). Aliquots of cell suspension (200 l) were added to preblocked, lipid-adsorbed microwells and incubated at 37°C for 45 min. To remove nonadherent cells after the incubation, the plate was im-  (1) is the sialic acid found on most gangliosides used in this study. Naturally occurring GM3-bearing N-glycolylneuraminic acid (2) was prepared from bovine spleen. Treatment of ganglioside GD1a with mild periodate converted its sialic acids to a mixture of 7-and 8-aldehyde forms (of which only the former is shown; 3). Reduction of these aldehydes resulted in a mixture of the corresponding 7-and 8-alcohols (the former is shown; 4). Treatment of GD1a with iodoethane converted its sialic acids to their ethyl esters (5), which were then converted to their 1-amides (6) or reduced to the corresponding 1-alcohols (7). mersed upright in a vat of PBS, inverted, and placed in an immersed custom-designed Plexiglas box, which was sealed with a gasket to exclude air (35). The inverted plate in its fluid-filled chamber was placed in a centrifuge carrier and centrifuged at 110 ϫ g to gently remove nonadherent cells. The box was again immersed in a vat of PBS, and the plate was removed and righted (while immersed), and surface fluid was removed by aspiration. Adherent cells were lysed by addition of 20 l of 10% Triton X-100, and 80 l/well was removed to a fresh 96-well plate for quantitation. Cell adhesion was quantitated by measuring lactate dehydrogenase activity in the cell lysate after addition of 120 l of 0.1 M potassium phosphate buffer, pH 7.0, containing 0.7 mM NADH and 4.7 mM pyruvate. The decrease in absorbance at 340 nm as a function of time was measured simultaneously in each well using a Molecular Devices UV multiwell kinetic plate reader. Details of the adhesion assay have been described (35).
For comparison with neuraminidase-pretreated cells, some MAGtransfected COS cells were pretreated with sodium periodate under mild conditions, which destroy MAG ligands (see below) but do not result in a loss of cell viability (38). Cells were suspended at 1 ϫ 10 6 cells/ml in 2.5 mM sodium periodate in PBS. After 45 min on ice with intermittent mild agitation, the cells were collected by centrifugation, washed three times with PBS, and finally resuspended at 250,000 cells/ml in Hepes-buffered DMEM containing 1.5 mg/ml BSA (as above). More than 75% of the cells were recovered intact (as measured by cellular lactate dehydrogenase activity) after periodate pretreatment and washing.
Neuraminidase Inhibitor-V. cholerae neuraminidase (0.1 milliunit) in 500 l of Dulbecco's PBS containing 2 mg/ml BSA was incubated with or without 100 M N-acetyl-2,3-dehydro-2-deoxyneuraminic acid (Toronto Research Chemicals, Downsview, Ontario, Canada) for 30 min at 37°C. COS cells transfected with pCDM8-MAG (5 ϫ 10 6 cells) were resuspended in the above solutions and incubated for 2 h at 37°C with end-over-end mixing, collected by centrifugation, and used in adhesion experiments as described above.
Cell Adhesion Directly to Gangliosides Resolved by TLC-Cell adhesion to TLC-resolved glycolipids was performed as described (39). MAGtransfected CHO cells were radiolabeled by addition of 20 Ci of 32 P i to their medium for 16 h. Cells were harvested, neuraminidase-pretreated (see above), and resuspended at 7 ϫ 10 5 cells/ml in Hepes-buffered DMEM containing 5 mg/ml BSA. Gangliosides were applied to duplicate Silica Gel 60 high performance TLC plates (5635; EM Separations, Gibbstown, NJ), and the chromatogram was developed with isopropyl alcohol/0.25% aqueous KCl (3:1). One plate was stained for gangliosides using resorcinol-hydrochloric acid (40). The other was dried, dipped in 100 g/ml polyisobutylmethacrylate (Polysciences, Warrington, PA) in hexanes, blocked in Hepes-buffered DMEM containing 5 mg/ml BSA for 15 min at 37°C, and placed in a Plexiglas TLC adhesion chamber (GlycoTech, Rockville, MD). The cell suspension (14 ml, see above) was pipetted into the chamber, which was incubated at 37°C for 45 min to allow adhesion to occur. The chamber was then inverted, and nonadherent cells were removed by centrifugation at 110 ϫ g for 10 min. The chamber was immersed, the plate was removed while immersed and transferred to a Petri dish without exposing to an air-liquid interface, and cells were fixed in place by addition of 2% glutaraldehyde in Dulbecco's PBS for 30 min at ambient temperature. The fixed plate was washed with PBS, dried, and subjected to PhosphorImager analysis (Fujix BAS 1000, Fuji Photo Film Co., Tokyo, Japan).

Adhesion of MAG-transfected COS and CHO Cells to Ganglioside-adsorbed Microwells-Binding of transiently transfected COS cells and stably transfected CHO cells to control
and ganglioside-adsorbed microwells was compared (Fig. 3). Consistent with our previous report (18), surfaces adsorbed with GT1b supported specific adhesion of MAG-transfected cells, whereas those adsorbed with GM1 did not. Adhesion of both transfected cell lines was greatly enhanced by pretreatment with V. cholera neuraminidase, indicating the presence of inhibitory structures on the transfected cells. COS and CHO cells transfected with control plasmids, with or without pretreatment with neuraminidase, did not specifically adhere to any ganglioside-adsorbed surface (data not shown). Adhesion of neuraminidase-treated, MAG-transfected COS cells to surfaces adsorbed with GT1b was resistant to higher detachment forces than adhesion of untreated cells (data not shown), suggesting that more MAG binding sites were available to support adhesion. Incubation of neuraminidase-pretreated, MAG-transfected COS cells with conformationally specific monoclonal antibody 513 (7, 41) blocked adhesion to GT1b-adsorbed surfaces (data not shown). This is consistent with previous observations (18) and demonstrates that neuraminidase pretreatment unmasks the same MAG binding sites that are apparent on untreated cells.
The relationship between the concentration of neuraminidase used and enhancement of MAG-mediated carbohydrate adhesion is shown in Fig. 4A. As little as 0.02 milliunits/ml neuraminidase significantly increased MAG-mediated cell adhesion, whereas adhesion was maximally enhanced by pretreatment with 2 milliunits/ml enzyme. When N-acetyl-2,3dehydro-2-deoxyneuraminic acid (42) was used to block neuraminidase prior to incubation with cells, MAG-mediated adhesion to GT1b-adsorbed surfaces was reduced to the level of untreated cells (Fig. 4B). These data indicate that removal of endogenous sialic acid residues from the transfected cells is responsible for the enhanced adhesion of MAG-transfected cells to gangliosides.
Carbohydrate Specificity of MAG-mediated Adhesion-A restricted family of structurally related glycolipids supports MAG-mediated cell adhesion (18), the order of potency being (in part) GQ1b␣ Ͼ GT1b ϭ GD1a Ͼ Ͼ GM1, GD1b, and GQ1b (the latter do not support adhesion). MAG-transfected COS and CHO cells, pretreated with neuraminidase to enhance MAG binding, adhered to the same set of gangliosides (Fig. 5). GQ1b␣ was the most potent ganglioside supporting MAG-mediated cell adhesion, whereas GT1b and GD1a supported specific adhesion when adsorbed at higher concentrations. Although GQ1b at high concentrations appeared to support MAG-mediated adhesion of transfected COS and CHO cells, this was due to contaminating GT1b in the commercially obtained sample, as indicated by TLC (see below). All other purified or synthetic gangliosides were single species by TLC (data not shown). These data confirm the requirement for a single terminal 2,3-linked sialic acid to support MAG binding (16,18). This led us to reassess the ganglioside GM3 for binding, using two forms, one carrying only NeuAc (N-acetylneura- minic acid) and the other carrying predominantly NeuGc (Nglycolylneuraminic acid; see Fig. 2). Using neuraminidasepretreated COS cells, adhesion to GM3 (NeuAc) was detected (Fig. 6A), although the potency of GM3 in supporting adhesion was much less than that of GD1a under identical conditions. GM3 (NeuGc), in contrast, did not support MAG-mediated adhesion, demonstrating that the presence of a single additional hydroxyl group on sialic acid can abrogate recognition.
To distinguish whether the enhanced potency of GD1a compared with GM3 was due to steric factors (its longer neutral core) or to additional structural components, two gangliosides bearing an ␣2,3-linked sialic acid moiety on the terminal galactose of a tetraosyl core (GM1b and SLeX) were tested. Using neuraminidase-pretreated COS cells, adhesion to GM1b was relatively poor, and adhesion to SLeX was undetectable (Fig. 6B).
To further test the carbohydrate binding specificity of MAG, a series of soluble oligosaccharides were used in an attempt to inhibit adhesion of neuraminidase-pretreated, MAG-transfected COS cells to gangliosides. Cells were preincubated with various concentrations of potential soluble saccharide inhibitors for 60 min at 37°C with constant end-over-end agitation prior to addition to GT1b-adsorbed microwells. Even at the maximum concentrations tested (25 mM sialic acid, 1.5 mM ␣2,3or ␣2,6-sialyllactose, and 50 M GT1b or GM1 oligosaccharides), none of the soluble saccharides significantly inhibited MAG-transfected COS adhesion (data not shown). In contrast, micellar (intact) GT1b ganglioside (but not GM1
MAG-mediated Cell Adhesion to TLC-resolved Gangliosides-Because purified gangliosides may harbor minor species as contaminants, direct adhesion of MAG-transfected cells to TLC-resolved gangliosides (39,43) was used to determine whether the major species were the active MAG ligands. A partially purified mixture of disialylated gangliosides from bovine brain, purified bovine brain gangliosides GT1b and GQ1b, and synthetic GQ1b␣ were resolved by TLC. The plate was treated with a polymer coating and then incubated with 32 Plabeled, MAG-transfected CHO cells. Nonadherent cells were removed without subjecting adherent cells to an air-liquid interface using a Plexiglas adhesion chamber (GlycoTech; Refs. 39 and 43). Adherent cells were detected with a PhosphorImager, and positions of adhesion were compared with the migration of gangliosides detected on a sister plate by chemical staining (Fig. 7). In the disialoganglioside mixture, cells adhered at the migration position of GD1a, but not to GD1b, even though the sample had more GD1b. Cells adhered at the migration positions of purified GT1b and synthetic GQ1b␣, with the latter supporting avid adhesion, even though it was present at less than 1 ⁄10 of the concentration of GT1b. The cells did not adhere at the migration position of GQ1b. Although cell adhesion was detected in this GQ1b sample (see Fig. 5), adhesion on TLC was at the migration position of GT1b. This particular commercial sample of GQ1b was contaminated with GT1b, as indicated by chemical staining on TLC (Fig. 7). Quantitation by scanning densitometry (27) established that the GQ1b sample contained 27% GT1b (data not shown). Essentially all of the adhesion to microwell-adsorbed GQ1b (Fig. 5) can be accounted for by the presence of GT1b. These data demonstrate the power of combined microwell and TLC binding to confirm the identity of MAG ligands.
Sialic Acid Structural Requirements for MAG Binding-MAG-mediated adhesion to GM3 was absent when the sialic acid was NeuGc rather than NeuAc (Fig. 6). To further assess the fine structural specificity of binding, sialic acids on GD1a (di-NeuAc form) were chemically modified (see Fig. 2). Treatment of GD1a with 2 mM aqueous sodium periodate at 0°C selectively cleaved the sialic acid glycerol side chains to the corresponding 7-or 8-aldehydes (31). Further treatment of the aldehyde products with sodium borohydride reduced the sialic acid aldehydes to their corresponding 7-or 8-position alcohols.
GD1a sialic acid carboxylic acids were ethyl-esterified by treatment with iodoethane, followed by either conversion to the corresponding 1-position amides with ammonia or reduction to the 1-position alcohols with sodium borohydride.
Periodate treatment under these conditions resulted in approximately equal conversion of each sialic acid to either its 7-aldehyde or 8-aldehyde forms, as evidenced by appropriate molecular ions detected by fast atom bombardment-mass spectrometry. Resolution by TLC in isopropyl alcohol/0.25% aqueous KCl (3:1) resulted in two major bands. Reduction of the aldehydes generated the expected mixture of sialic acid 7-alcohol and 8-alcohol forms, confirmed by fast atom bombardmentmass spectrometry and TLC as above. Conversion of GD1a to its diethyl ester resulted in a single species by TLC with the appropriate mass ions by fast atom bombardment-mass spectrometry. Further conversion to the corresponding sialic acid 1-amide or 1-alcohol forms resulted in new single species with appropriate changes in TLC migration.
GD1a and its derivatives were adsorbed to microwells and tested for their ability to support specific adhesion of MAGtransfected COS cells. Neither the truncated glycerol side chain derivatives nor any of the uncharged species (carboxylic acid ethyl esters, 1-position amides, or 1-position alcohols) supported MAG binding (Figs. 8 and 9). All derivatives adsorbed equally to the microwells and remained adsorbed equally during incubation with cells (data not shown). Control transfected COS cells did not adhere to any of the GD1a derivatives (data not shown).
Periodate Treatment of MAG-transfected Cells-The discovery that mild periodate treatment of gangliosides abrogated MAG binding (Fig. 8) provided an experimental method to investigate the mechanism by which neuraminidase treatment of transfected cells enhanced MAG-mediated cell adhesion (Fig.  3). If neuraminidase enhanced MAG-mediated adhesion by destroying endogenous MAG ligands on the transfected cells, periodate would be expected to do the same. In contrast, if neuraminidase enhanced MAG binding by an alternative mechanism, such as reducing the surface anionic charge on transfected cells, periodate would not be expected to enhance binding. The results demonstrate the latter to be the case. Using periodate incubation conditions that destroyed the ability of adsorbed GD1a to support MAG-mediated cell adhesion ( Fig. 10A), periodate pretreatment of intact cells failed to enhance binding (Fig. 10B). Neuraminidase pretreatment, in contrast, sharply enhanced MAG-mediated adhesion. DISCUSSION MAG is a lectin with highly specific binding to terminal sialic acids attached in particular arrays on glycoconjugates (16,18). Optimal conditions for detecting MAG-carbohydrate binding were determined using both transiently and stably transfected cell lines, each expressing full-length MAG as a plasma membrane protein (11,18). Adhesion of the transfected cells to ganglioside-containing phosphatidylcholine and cholesterol monolayers was quantitated in a high throughput microwell adhesion format (35). In these assays, MAG, which is polyvalent on the transfected cell surface, interacts with carbohydrate targets presented as multivalent arrays in an apposing membrane. This system models the potential interaction of natural MAG-expressing cells (oligodendroglia and Schwann cells) with glycoconjugates on neurons or other cells in their environment.
Neuraminidase pretreatment markedly increased MAG-ganglioside adhesion without changing its carbohydrate specificity or anti-MAG monoclonal antibody susceptibility. These data are consistent with at least two possible interpretations. Endogenous sialoglycoconjugates on the transfected cells may bind, in cis configuration, to a significant proportion of the expressed MAG, reducing its availability for interaction with exogenous targets. This mechanism apparently regulates carbohydrate binding by two other I-type lectins, CD22 and CD33 (44 -46). Alternatively, the anionic nature of the cell surface may limit interactions of transfected cells with anionic ganglioside-adsorbed surfaces. In either case, neuraminidase treat-ment would reverse the inhibition. Periodate pretreatment, in contrast, would be expected to destroy endogenous ligands but not to reduce surface anionic charge. Because periodate pretreatment failed to enhance MAG-mediated cell adhesion, neuraminidase is likely to enhance binding by reducing the cell surface anionic charge. This finding may have a physiological correlate. Attenuating charge repulsion to facilitate membrane-membrane interactions is a possible function of a membrane-bound myelin-associated neuraminidase, which is active against myelin glycoproteins and glycolipids (47) and which is involved in the ganglioside "simplification" of mature myelin (48). Whether or not an endogenous myelin neuraminidase facilitates cell-cell interactions, our data indicate that pretreatment of MAG-expressing cells with neuraminidase may reveal more pronounced effects of MAG on neuronal behaviors in vitro (11).
Under optimal binding conditions, using either MAG-transfected COS or CHO cells, MAG binds to a discrete family of gangliosides with the relative potencies: GQ1b␣ Ͼ GT1b ϭ GD1a Ͼ GM3 Ͼ Ͼ GD1b, GM1, and GQ1b (the latter do not detectably bind MAG). Although an ␣2,3-NeuAc bound to the terminal Gal on a neutral sugar core appears to be required for binding (16,18), other moieties enhance or diminish that binding. For example, gangliosides GM3 and GM1b are comparable and much weaker binding ligands for MAG than is GD1a. These data indicate that the internal sialic acid (II 3 NeuAc) on GD1a enhances its interaction with MAG. In addition to the IV 3 NeuAc, the ganglioside GQ1b␣ has additional II 3 NeuAc and III 6 NeuAc residues, which sharply increase binding affinity. Other moieties diminish MAG binding. Whereas MAG binds (albeit weakly) to GM3, it does not bind at all to GM1, which is equivalent to GM3 with a disaccharide constituent on the 4-position of Gal(II). The failure of SLeX, which also has the terminal NeuAc␣2,3Gal group, to support MAG adhesion may be Microwells were adsorbed with phosphatidylcholine/cholesterol monolayers with (q) or without (f) added GD1a (50 pmol/well). After lipid adsorption and washing, the wells were chilled and 100 l/well ice-cold 2.5 mM sodium periodate in PBS were added. At the indicated times the wells were washed with water. Treated microwells were preblocked (10 min, 37°C, Hepes-buffered DMEM containing 1.5 mg/ml BSA), and then 200 l of the same medium containing 50,000 MAG-transfected, neuraminidase-pretreated COS cells were added. Cell adhesion was measured after 45 min at 37°C. Values are expressed relative to the total cells added to each well and are the means Ϯ S.E. (bars) of triplicate determinations. B, effects of neuraminidase and periodate pretreatments on MAG-transfected COS cell adhesion to GD1a. MAGtransfected COS cells were collected from culture dishes, suspended in PBS, and split into three groups. One group (q) was kept on ice in PBS; one group (f) was pretreated with 10 milliunits/ml V. cholerae neuraminidase at 37°C for 75 min; and one group (å) was pretreated with 2.5 mM sodium periodate on ice for 45 min. Cells were washed by centrifugation, resuspended at 250,000 cells/ml in Hepes-buffered DMEM containing 1.5 mg/ml BSA, and added to microwells adsorbed with the indicated concentrations of GD1a. Cell adhesion was measured after 45 min at 37°C. Values are expressed relative to the total cells added to each well and are the means Ϯ S.E. (bars) of quadruplicate determinations. due to steric hindrance by the nearby fucose or to structural differences in the neutral core (neolacto versus ganglio structure).
Monovalent soluble sialo-oligosaccharides, including the oligosaccharide released from GT1b (up to 50 M), failed to inhibit binding of MAG-transfected COS cells to adsorbed gangliosides. This may reflect cooperativity when polyvalent MAG in the transfected cell membrane contacts polyvalent gangliosides on an apposing planar surface (49). Similar polyvalent interactions may take place at the interface between the periaxonal myelin membrane and the axolemma (6). Alternatively, the ceramide moiety of gangliosides may impart conformational restraints on ganglioside oligosaccharides, which enhance binding affinity (50).
The ganglioside binding specificity of MAG may reflect its role in intracellular interactions in vivo. Central nervous system myelin is enriched in the non-MAG binding ganglioside GM1, which comprises up to 90% of adult rat myelin gangliosides (48,51). In contrast, axolemma, which is directly apposed to MAG in vivo, is enriched in the strong MAG-binding gangliosides GT1b and GD1a (52). Gangliosides bearing these structures, as well as GQ1b␣ and related structures, are expressed preferentially on neurons and axons (51)(52)(53) and may act as endogenous MAG ligands. The observations that monoclonal antibody 513 inhibits binding of MAG to both neurons and gangliosides (8,18) and inhibits oligodendrocyte-neuron adhesion in vitro (7) further support the hypothesis that MAG expressed on oligodendrocytes binds to neuronal sialoglycoconjugates, the precise nature of which are unknown. Prior studies suggest that sialoglycoproteins may be relevant ligands for MAG, at least for some neuronal cell types (54). Whether neuronal gangliosides, sialoglycoproteins, or both serve as functional MAG ligands remains to be determined.
As common termini of mammalian glycoconjugates, sialic acids bear a special role in recognition (55). The five I-type lectins (16,17) and three selectins (56) all bind sialic acids, although their precise carbohydrate determinant specificities vary. The sialic acid linkage (␣2,3 or ␣2,6) appears critical for recognition, with CD22 requiring ␣2,6-linked sialic acid and all others requiring ␣2,3 linkages (16). Selectin-mediated adhesion is maintained when critical sialic acid residues are replaced by other anionic functional groups (56,57), a scenario that has not been thoroughly tested with any of the I-type lectins. However, the relatively high specificity of I-type lectins for sialic acid substructure suggests that such replacements would not be fruitful.
Sialic acids are the most complex saccharides in vertebrate glycoconjugates. They constitute a family of structures based on N-acetylneuraminic acid, the most abundant sialic acid and precursor to diverse structures bearing modifications on the glycerol side chain, N-acyl group, and/or hydroxyl at carbon 4 (55). Sialic acid binding lectins vary in their preference or acceptance of alterations in sialic acid substructure. For example, all three selectins tolerate periodate truncation of the glycerol side chain (58 -60), whereas the I-type lectins CD22 (61) and MAG (this study) do not.
Enzymatic hydroxylation of the N-acetyl group of NeuAc forms N-glycolylneuraminic acid (NeuGc), which is rare in humans but common in some tissues of other vertebrates, including rodents. Notably, human CD22 binds NeuAc and NeuGc residues equally (61), whereas mouse CD22 has a marked preference for NeuGc-terminated glycoconjugates (62), which are abundant on murine CD22 targets (lymphocytes and monocytes). In contrast, we report that rodent MAG (95% amino acid identical to human MAG; Ref. 63) binds NeuAc-terminated GM3 but not the corresponding NeuGc form. Since brain gan-gliosides bear predominantly NeuAc residues in both rodents and humans (51), our findings are consistent with a MAG binding preference for its neuronal target glycoconjugates. Murine sialoadhesin, another I-type lectin, has a similar preference for NeuAc-bearing glycoconjugates (62).
Ethyl esterification of the carboxylic acid on NeuAc abrogated MAG binding, and conversion of the ester to the corresponding amide or primary alcohol did not restore binding. All three derivatives were uncharged (did not bind DEAE-Sepharose) and stable during the experimental incubations. Thus, MAG is among the most specific NeuAc-directed lectins, since it does not tolerate changes in either the NeuAc glycerol side chain, its N-acyl group, or its carboxylic acid. Having demonstrated the high specificity of MAG-glycoconjugate binding using naturally occurring, synthetic, and chemically modified gangliosides, the nature of the endogenous neuronal glycoligands for MAG, and their functions in neural cell-cell interactions, await future investigations.