Enhanced Binding of the Neural Siglecs, Myelin-associated Glycoprotein and Schwann Cell Myelin Protein, to Chol-1 (α-Series) Gangliosides and Novel Sulfated Chol-1 Analogs*

Extended glycoconjugate binding specificities of three sialic acid-dependent immunoglobulin-like family member lectins (siglecs), myelin-associated glycoprotein (MAG), Schwann cell myelin protein (SMP), and sialoadhesin, were compared by measuring siglec-mediated cell adhesion to immobilized gangliosides. Synthetic gangliosides bearing the α-series determinant (NeuAc α2,6-linked to GalNAc on a gangliotetraose core) were tested, including GD1α (IV3NeuAc, III6NeuAc-Gg4OseCer), GD1α with modified sialic acid residues at the III6-position, and the “Chol-1” gangliosides GT1aα (IV3NeuAc, III6NeuAc, II3NeuAc-Gg4OseCer) and GQ1bα (IV3NeuAc, III6NeuAc, II3(NeuAc)2-Gg4OseCer). The α-series gangliosides displayed enhanced potency for MAG- and SMP-mediated cell adhesion (GQ1bα > GT1aα, GD1α > GT1b, GD1a ≫ GM1 (nonbinding)), whereas sialoadhesin-mediated adhesion was comparable with α-series and non-α-series gangliosides. GD1α derivatives with modified sialic acids (7-, 8-, or 9-deoxy) or sulfate (instead of sialic acid) at the III6-position supported adhesion comparable with that of GD1α. Notably, a novel GT1aα analog with sulfates at two internal sites of sialylation (NeuAcα2,3Galβ1,4GalNAc-6-sulfateβ1, 4Gal3-sulfateβ1,4Glcβ1,1′ceramide) was the most potent siglec-binding structure tested to date (10-fold more potent than GT1aα in supporting MAG and SMP binding). Together with prior studies, these data indicate that MAG and SMP display an extended structural specificity with a requirement for a terminal α2,3-linked NeuAc and great enhancement by nearby precisely spaced anionic charges.

Siglecs (sialic acid-dependent immunoglobulin-like family member lectins) contain an N-terminal V-type Ig domain, a varying number of C-type Ig domains, a single transmembrane domain, and a short cytoplasmic tail (1,2). The six reported siglecs (CD22, CD33, sialoadhesin, MAG, 1 SMP, and siglec 5) share a significant degree of sequence similarity among their V-type domains, which are required for sialic acid binding. Sialoadhesin (siglec 1), which is the largest siglec (17 Ig-like domains), is found on bone marrow macrophages and may play a role in hematopoiesis. The nervous system siglecs, MAG and SMP (siglecs 4a and 4b), each of which has five Ig-like domains, are expressed on myelinating oligodendrocytes and Schwann cells. MAG is involved in myelin maintenance and in myelinaxon interactions that mediate neuronal cytoarchitecture and the inhibition of axon regeneration after injury (3)(4)(5)(6). SMP, which is found only in avian species, may be the avian homologue of MAG (7). Presumably, the biological functions of MAG and other siglecs require their binding to sialoglycoconjugates on apposing cell surfaces (8). Defining the carbohydrate determinants responsible for siglec binding may help reveal their biological target specificities and provide opportunities for the design of carbohydrate mimetics to modulate siglec function.
MAG is expressed on the periaxonal membrane of myelin, directly apposed to the neural membrane, where its ligand is thought to be expressed (19). Since gangliosides carry 75-80% of the sialic acid in the brain (20), and the major brain gangliosides GD1a 2 and GT1b express the preferred NeuAc␣2, 3Gal␤1,3GalNAc terminal target determinant for MAG, we hypothesize that gangliosides are functional MAG ligands. A subset of neurons, those that use acetylcholine as their neurotransmitter, express a unique quantitatively minor family of gangliosides termed "Chol-1" gangliosides, initially defined by their reactivity with a polyclonal antiserum raised against cholinergic neurons (21). Chol-1 gangliosides carry an ␣2,6linked sialic acid on the GalNAc of a gangliotetraose core * This work was supported in part by National Institutes of Health Grant NS37096, a grant from the National Multiple Sclerosis Society, National Science Foundation Grant IBN-9631745, and a grant from the Paralyzed Veterans of America Spinal Cord Research Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 1 The abbreviations used are: MAG, myelin-associated glycoprotein; SMP, Schwann cell myelin protein; PBS, phosphate-buffered saline. 2 Ganglioside nomenclature is that of Svennerholm (46) or as indicated in Fig. 1. cis-GM1 (NeuAc␣2,3Gal␤1,3GalNAc␤1,4Gal␤1,3Glc-␤1,1ЈCer) is also referred to in the literature as "GM1b." (GT1a␣ and GQ1b␣, see Fig. 1), making them part of the "␣-series" ganglioside family (22)(23)(24). Although O-linked glycoproteins can also carry the NeuAc␣2,6GalNAc determinant, brain glycoproteins are not immunoreactive with Chol-1 antisera (25), indicating that gangliosides may express the Chol-1 determinant in a distinctive conformation. The observation that GQ1b␣ is a remarkably high affinity ligand for the neural siglecs, MAG and SMP (17), prompted us to further explore the role of the ␣-series determinant and related structures in extended siglec recognition.
Siglec Expression-cDNAs encoding full-length rat MAG, quail SMP, and mouse sialoadhesin were cloned into the expression vectors pCDM8 (MAG and SMP) or pCDNA1/Amp (sialoadhesin) (7,18,30). Plasmids propagated in either MC1061/P3 (MAG and SMP) or DH5␣ (sialoadhesin) E. coli were purified by polyethylene glycol precipitation. COS-1 cells were propagated in Dulbecco's modified Eagle's medium containing 10% fetal calf serum at 37°C and in a humidified environment of 90% air, 10% CO 2 . Cells were plated at a density of 9 ϫ 10 5 cells/ 100-mm diameter dish, were cultured overnight, and then were transiently transfected by exposure to 4 ml of Dulbecco's modified Eagle's medium supplemented with 2.5% fetal calf serum, 40 g/ml DEAEdextran, 0.1 mM chloroquine, and 3 g of the plasmid of interest. After 3.5 h of transfection, the medium was removed, and cells were treated with 10% Me 2 SO (v/v) in PBS for 5 min and then were returned to growth medium. The cells were cultured for 40 -50 h to allow ample siglec gene expression prior to detachment for use in adhesion assays.
Siglec-mediated Cell Adhesion-Adhesion assays were performed as described previously (16,31). Gangliosides were adsorbed to the bottom of 96-well dishes in an artificial membrane monolayer with phosphatidylcholine and cholesterol. Aliquots (50 l/well) of ethanol/water (1:1) containing 0.5 M phosphatidylcholine, 2.0 M cholesterol, and the indicated amounts of ganglioside were added to wells of a 96-well flat-bottom microwell plate. The plate was left uncovered for 90 min at ambient temperature to allow partial evaporation and lipid immobilization to occur. After adsorption, the wells were washed three times with water and then were preblocked by the addition of 100 l/well Hepes-buffered Dulbecco's modified Eagle's medium containing 2 mg/ml bovine serum albumin. Plates were incubated for 10 min at 37°C prior to the addition of cells.
Transfected cells were harvested by exposure to hypertonic Ca 2ϩ / Mg 2ϩ -free PBS containing 1 mM EDTA as described (18), collected by centrifugation, and resuspended at 10 7 cells/ml in PBS containing 2 mg/ml bovine serum albumin and 10 milliunits/ml of V. cholerae neuraminidase (which enhances adhesion (16)). The cells were incubated with gentle end-over-end agitation for 2 h at 37°C, washed by centrifugation, and finally resuspended in Hepes-buffered Dulbecco's modified Eagle's medium containing 2 mg/ml bovine serum albumin at 2.5 ϫ 10 5 cells/ml. An aliquot of the cell suspension (200 l) was added to each well, and the cells were allowed to settle for 10 min at 4°C prior to incubation for 45 min at 37°C to allow cell adhesion to proceed.
Subsequent to the adhesion incubation, nonadherent cells were removed using carefully controlled centrifugal detachment force. To avoid fluid sheer, the plate was carefully immersed upright in a vat of PBS (at ambient temperature), inverted (while immersed), and placed (inverted) in an immersed custom-designed Plexiglas box, which was sealed with a gasket to exclude air (31). The inverted plate in its fluid-filled chamber was placed in a centrifuge carrier and centrifuged at 110 ϫ g for 10 min to remove nonadherent cells, and the chamber was returned to the vat of PBS. While immersed, the plate was gently removed from the chamber but kept immersed (in the inverted orientation) for 5 min to allow any floating cells to settle away. The plate was then righted (while immersed) and removed from the vat, and excess surface PBS FIG. 1. Gangliosides and ganglioside analogs used in this study. Shorthand ganglioside nomenclature is based on that of Svennerholm (46) (cis-GM1 is also referred to as "GM1b" in the literature). The "␣" designation indicates a NeuAc residue linked ␣2,6 to the GalNAc residue in the gangliotetraose core. Chol-1 gangliosides, antigens of the polyclonal cholinergic-specific Chol-1 antiserum (21), include GT1a␣ and GQ1b␣ (22,23).
was removed by aspiration. At this point, all wells were full of PBS (320 l/well), and only adherent cells remained on the well bottoms. Adherent cells were quantified by lysis and measurement of released lactate dehydrogenase. After completion of the adhesion assay, 10 l of 10% Triton X-100 in PBS were added to each PBS-filled well, adherent cells allowed to lyse for 10 min and then triturated thoroughly with a multichannel micropipettor. An aliquot of lysate (80 l) from each well was transferred to a fresh plate, and 120 l of PBS containing 0.5 mg/ml each of NADH and sodium pyruvate were added. The decrease in UV absorbance at 340 nm was measured simultaneously in 96 wells using a kinetic plate reader (Benchmark Microplate Reader, Bio-Rad). Kinetic rates were compared with those from wells containing aliquots of standard cell suspensions to calculate the percentage of cells added that remained adherent at the end of the experiment. To account for modest variations in transfection efficiency between experiments and between the different transfection vectors, values were normalized to the maximum percentage of cells adhering to positively adherent gangliosides for that lectin in that experiment, adjusted for background adhesion. Over the course of the 14 experiments (39 separate siglec transfections) reported here, the average maximum and background adhesion values were as follows (mean Ϯ S.E.): 75.7 Ϯ 5.2 (maximum) and 12.2 Ϯ 2.4 (background) for MAG; 67.2 Ϯ 4.7 (maximum) and 13.1 Ϯ 1.7 (background) for SMP; and 72.8 Ϯ 5.1 (maximum) and 9.2 Ϯ 0.9 (background) for sialoadhesin. Each value reported in the figures is the average of 3-41 separate determinations and is expressed as mean Ϯ S.E. Gangliosides used in these studies adsorbed comparably to the wells and remained adsorbed equally during incubation with cells (data not shown).

RESULTS
Siglec Adhesion to ␣-Series Gangliosides-In agreement with our previous report (17), MAG and sialoadhesin adhered with moderate potency to the two major brain gangliosides bearing terminal "NeuAc␣2,3Gal␤1,3GalNAc" determinants, GD1a and GT1b (Fig. 2, A and C, and Table I), whereas SMP-mediated adhesion to these gangliosides was less avid (Fig. 2B; SMP-mediated adhesion to weakly supportive gangliosides added at concentrations above 50 pmol/well was often less than maximum levels, perhaps due to charge repulsion (16)). The Chol-1 ganglioside GQ1b␣ was ϳ10-fold more potent in supporting MAG-and SMP-mediated adhesion than was the closely related major brain ganglioside lacking the ␣2,6-NeuAc residue, GT1b (Table I). Gangliosides without the NeuAc␣2,3Gal terminus, such as GM1 (Fig. 2) and GM1␣ (III 6 NeuAc-Gg 4 OseCer, data not shown) did not support adhesion of any of the siglecs.
MAG and SMP displayed enhanced avidity for the other ␣-series gangliosides tested, GD1␣ and GT1a␣ (Fig. 2, A and B; Table I), compared with GD1a and GT1b. In contrast to the neural siglecs, ␣-series gangliosides and GD1a were equally potent in supporting sialoadhesin-mediated cell adhesion (Fig.  2C). A comparison of adhesion to GD1␣ and cis-GM1 (Table I; structures in Fig. 1), indicates that the ␣2,6-NeuAc residue per se enhanced binding of siglecs 3-6-fold.
The Role of Exocyclic Glycerol Chain Hydroxyls of the ␣2,6-Linked NeuAc on Siglec Recognition of GD1␣-The increased potency of Chol-1 gangliosides to support MAG-and SMPmediated adhesion suggests a potential direct interaction between the ␣2,6-NeuAc residue and the neural siglecs. Previously, we reported that MAG required an intact sialic acid exocyclic glycerol chain on the terminal ␣2,3-NeuAc residue for recognition (16). In the current study, we tested the role of the exocyclic glycerol chain hydroxyls on the ␣2,6-NeuAc residue of GD1␣ using synthetic derivatives (32). GD1␣, bearing a 7-, 8-, or 9-deoxysialic acid linked ␣2,6 to GalNAc, supported comparably enhanced siglec-mediated adhesion ( Fig. 3 and Table I), indicating that the hydroxyls on this exocyclic glycerol chain are not strictly required for enhanced siglec recognition. This led us to test the role of the anionic charge at the III 6 -position using sulfate analogs of GD1␣ and GT1a␣.
Replacing the III 6 NeuAc residue of GD1␣ with a sulfate group (cis-GM1 III 6 -sulfate, Fig. 1) had no effect on MAG or sialoadhesin binding (Fig. 4, Table I) and only a modest negative effect on SMP binding, indicating that anionic charge at that position is key to enhanced binding affinity. Notably, an analog of GT1a␣ bearing two sulfate groups (cis-GM1 variant II 3 , III 6 -bissulfate, Fig. 1) demonstrated 10-fold increased binding affinity for MAG and SMP and 2-fold for sialoadhesin, making it the most highly potent target for siglec-mediated cell adhesion tested to date (Fig. 4, Table I).
Siglec-mediated Adhesion to cis-GM1 with Altered Neutral Cores-To investigate extended recognition by siglecs based on FIG. 2. Siglec-mediated cell adhesion to ␣-series and major brain gangliosides. COS cells transiently transfected to express MAG (A), SMP (B), or sialoadhesin (C) were collected from culture dishes; pretreated with neuraminidase to enhance adhesion (16); and placed in microwells previously adsorbed with phosphatidylcholine, cholesterol, and the indicated gangliosides. After incubation, nonadherent cells were removed by centrifugation, and adherent cells were quantified enzymatically (see "Experimental Procedures"). Adhesion is expressed relative to the total number of cells added to each well and represents the mean Ϯ S.E. of 3-41 replicate determinations performed at each concentration. Gangliosides used were as follows: GQ1b␣ (OE), GT1a␣ (E), GD1␣ (q), GD1a (f), GT1b (Ⅺ), and GM1 (‚). the gangliotetraose core, we employed synthetic derivatives of cis-GM1 (29), the minimal gangliotetraose structure bearing the NeuAc␣2,3Gal␤1,3GalNAc terminus preferred by MAG (Fig. 1). MAG bound 3-fold better to cis-GM1 than to a matched structure with a GlcNAc replacing the core GalNAc (Fig. 5). Sialoadhesin, however, recognized both structures equally well. Surprisingly, a novel synthetic structure having the IV-Gal residue in ␤1,6 rather than ␤1,3 linkage (cis-GM1, "Gal ␤1,6-GalNAc" variant) supported siglec adhesion at levels similar to the parent cis-GM1 (Fig. 5). In contrast to cis-GM1, monosialogangliotetraoses GM1 (Fig. 2) and GM1␣ (data not shown) did not support any siglec adhesion. DISCUSSION The current study supports the hypothesis that there is extended recognition of multisialylated gangliosides by the closely related neural siglecs, MAG and SMP. For these siglecs, relative placement of the sialic acids (or sulfates) on the neutral sugar core appears to be an important factor in determining binding affinity. This and prior published data (9, 16 -18) indicate that MAG and SMP require a terminal "NeuAc␣2,3Gal" determinant as the primary structural requirement for binding but that additional (secondary) sialic acids on the same core greatly enhance binding. Relative placement of the secondary sialic acids appears to be key, with the NeuAc␣2,6GalNAc determinant preferred. Thus, GD1a has higher affinity for MAG and SMP than cis-GM1, GD1␣ has yet higher affinity, and GT1a␣ is more potent than GT1b (Table I). Two alternative hypothesis fit these data: (i) MAG and SMP have primary and secondary sialic acid binding sites; or (ii) MAG and SMP are very sensitive to the precise tertiary configuration of the primary determinant (NeuAc␣2, 3Gal␤1, 3GalNAc), which is in turn impacted by nearby sialic acids. NMR studies support the latter hypothesis, in that the core GalNAc and II 3 NeuAc of gangliosides directly interact to restrict the conformation of what would be the primary determinant (33). However, the III 6 NeuAc, which is yet more potent as a secondary sialic acid, may not interact in the same manner to restrict ganglioside conformation, and replacement of the III 6 NeuAc and II 3 NeuAc with sulfates greatly enhanced MAG and SMP binding. In either case, the extended structure of multisialylated ganglio-sides impacts the affinity of MAG and SMP binding. Importantly, the physiological significance of gangliosides bearing extended determinants has recently been established (8). Mice engineered to lack the ganglioside neutral core GalNAc transferase (UDP-GalNAc:GM3/GD3 N-acetylgalactosaminyltransferase) had progressive axon and myelin degeneration similar to the neural deficits found in MAG-deficient mice. Therefore, the extended specificity of MAG for its carbohydrate ligand may have important physiological consequences.
The extended specificity reported here conflicts with data reported recently by Strenge et al. (34), in that the oligosaccharide of GT1a␣ was reported to be no better than that of GM3 (sialyllactose, NeuAc␣2,3Gal␤1,4Glc) in inhibiting MAG d1-3 -Fc binding to human erythrocytes. Other structurefunction differences between the Strenge et. al. study and our studies were also noted, including inhibition of MAG binding  3. Siglec-mediated cell adhesion to GD1␣ derivatives containing modified ␣2,6-linked sialic acid residues. Adhesion of COS cells transiently transfected to express MAG (A), SMP (B), or sialoadhesin (C) to the indicated gangliosides was determined as described under "Experimental Procedures" and the legend to Fig. 2. GD1␣ derivatives carried the following sialic acid linked ␣2,6 to the core GalNAc: NeuAc (q), 7-deoxy-NeuAc (‚), 8-deoxy-NeuAc (Ⅺ), and 9-deoxy-NeuAc (E). A preliminary form of A was published along with the detailed syntheses of the deoxy derivatives (32). by soluble saccharides bearing 7-deoxy-NeuAc or 2-keto-3-deoxy-D-glycero-D-galactonononic acid (KDN) in place of NeuAc and by certain oligosaccharides bearing only terminal ␣2,6linked sialic acids. We agree with Strenge et al. that the differences are probably due to the different assays used. The current study measures binding of full-length cell surface MAG to saccharides oriented on an apposing membrane monolayer, whereas the Strenge et al. study measured the site affinity of soluble saccharides for a soluble MAG-Fc chimera engineered to have only the N-terminal three (of five) Ig-like domains. Presentation, valency, and the different forms of MAG may all contribute to the observed differences. Furthermore, it is difficult to compare the half-maximal inhibitory potency of free oligosaccharides for the MAG-Fc chimera (in the submillimolar range) with the half-maximal potencies reported here for sup-porting MAG adhesion (in the pmol/well range). Notably, genetically engineered mice expressing only truncated gangliosides display neuropathies similar to those in mice lacking MAG, supporting the notion that extended ganglioside specificity is of physiological importance (8).
The two neural siglecs, MAG and SMP, demonstrate the same relative trends in ganglioside binding specificity, although SMP displays characteristically lower binding avidity. These data reflect the close relationship between these two lectins, which share Ͼ70% sequence similarity in the first two N-terminal Ig-like domains (1). Sialoadhesin, which shares ϳ40% sequence similarity with the neural siglecs in its first two N-terminal Ig-like domains, does not demonstrate the same extended specificity (GQ1b␣, GT1a␣, and GD1␣ demonstrated sialoadhesin binding affinities comparable with GD1a). Although extended specificity is suggested by the relatively low affinity of sialoadhesin for cis-GM1 (compared with GD1a or GD1␣), prior results demonstrated moderately high binding affinity of sialoadhesin for GM3 and even GM4, the simplest "NeuAc␣2,3Gal"-bearing ganglioside (17). Together, these results suggest that sialoadhesin is less responsive to multiple sialic acids on the same neutral sugar core.
Sialic acid substructural specificity appears to be relatively stringent for all siglecs, which require an intact exocyclic glycerol side chain and are differentially responsive to the N-acyl moiety (35). This contrasts with selectins, which bind to target glycosides bearing highly modified sialic acids, or the same saccharides with sulfate or carboxylate moieties in place of sialic acid (36 -40). In prior studies, we demonstrated that MAG requires every hydroxyl on its primary sialic acid determinant for binding (17). The data presented here indicate that the structural requirements at the secondary determinant are not as stringent. Synthetic GD1␣ structures lacking the 7-, 8-, or 9-position hydroxyls were comparable with the parent GD1␣ in supporting MAG and SMP binding and were much more potent than cis-GM1 (Table I). Structures bearing sulfate groups in place of the ␣-ganglioside secondary NeuAc residues displayed either equivalent (cis-GM1 III 6 -sulfate) or sharply enhanced affinity (cis-GM1 variant II 3 ,III 6 -bissulfate) compared with their NeuAc-containing analogs (GD1␣ and GT1a␣, respectively). Sulfated/sialylated glycoconjugates have been found as naturally occurring targets for L-selectin (41), and sulfated Lewis blood group oligosaccharide analogs have been touted as potential selectin mimetics (42, 43). Whether sulfated/sialylated glycoconjugates exist in the brain as MAG targets or can be developed as potent MAG antagonists has yet to be determined.
In contrast to the marked effect of multiple sialic acids on the same neutral oligosaccharide core on MAG and SMP binding, changes in the neutral core had only modest effects. The replacement of the gangliotetraose core GalNAc with GlcNAc reduced binding about 3-fold, suggesting a modest impact on recognition. Altering the Gal-GalNAc linkage on cis-GM1 from ␤1,3 to ␤1,6 in the gangliotetraose core, which might be expected to have a large conformational effect, did not alter MAG binding. Furthermore, the potent cis-GM1 bissulfate variant has a ␤1, 4 Gal-GalNAc linkage. Although these data indicate that some neutral saccharide backbone variability is tolerated, interpretation of the conformational significance of these data awaits a high resolution structure of the saccharide binding site(s) of MAG.
Chol-1 gangliosides are minor species identified by their immunoreactivity with a polyclonal antibody raised in sheep against presynaptic plasma membranes of the cholinergic electromotor nerve terminals of Torpedo marmorata (21). The Chol-1 antibody, which specifically recognizes cholinergic synaptosomes from the mammalian brain, was discovered to target only gangliosides (25). GT1a␣ (0.9 mg/kg of brain) and GQ1b␣ (0.5 mg/kg) were defined as the two major Chol-1 targets (22,23), although yet less abundant species (GM1␣, GD1a␣, and GT1b␣) have recently been characterized (24). GD1␣ was discovered on hepatoma cells and tumor cell lines (44), although it was also found in the brain (0.3 mg/kg (45)). All of the "␣-series" gangliosides (bearing an ␣2,6-linked NeuAc residue on a gangliotetraose core) are very minor species compared with GD1a, the major ganglioside of brain (ϳ1200 mg/kg (20)). The biological significance of ␣-series ganglioside recognition by MAG must be considered in light of their relatively low expression. However, the highly restricted expression of ␣-series gangliosides may result in selected neurons with enhanced siglec recognition. Alternatively, yet unappreciated lectins may interact with these unusual gangliosides. Evaluation of their biolog-ical significance may await disruption of the ␣2,6-sialyltransferase responsible for their synthesis. Nevertheless, the ability of the ␣2,6-NeuAc residue, as well as II 3 -and III 6 -sulfates, to enhance MAG and SMP recognition in extended conformation with the primary ␣2,3-NeuAc terminus may provide useful biological and pharmacological tools.