10E4 Antigen of Scrapie Lesions Contains an Unusual Nonsulfated Heparan Motif*

The carbohydrate antigen on heparan sulfate recognized by monoclonal antibody 10E4 is uniquely codis-tributed with the abnormal prion protein, PrP Sc , even in the earliest detectable brain lesions of scrapie-infected mice. Determining the chemical structure of 10E4 antigen is, therefore, an important aspect of structure elu-cidation of scrapie lesions, and a prerequisite for de-signing experiments to understand its role in scrapie pathogenesis. Toward this aim, we have examined preparations of heparan sulfate, with differing sulfate contents, for binding by 10E4 antibody. The highest antigenicity was observed in a preparation (HS-1) with the lowest sulfate content. HS-1 was partially depolymerized with heparin lyase III, and oligosaccharide fragments examined for 10E4 antigen expression by the neoglycolipid technology. An antigen-positive and two antigen-negative tetrasaccharides were isolated and examined by electrospray mass spectrometry. The anti-gen-positive tetrasaccharide sequence on heparan sulfate was thus deduced to contain a unique unsulfated motif that includes an N -unsubstituted glucosamine in the sequence, UA-GlcN-UA-GlcNAc. Antibody binding experiments with neoglycolipids prepared from a series of heparin/heparan sulfate disaccharides, and the trisaccharide derived from the antigen-positive

Scrapie is the prototype of an unusual group of transmissible neurodegenerative diseases characterized by deposition of an alternatively folded, abnormally protease-resistant, isoform of the host prion protein, PrP C 1 , which is proposed to constitute the infective agent (1). There is considerable progress in knowledge of the molecular and cellular biology of prions, but precise details of the mechanism of conversion to the disease-associated form, PrP Sc , remain to be elucidated (2). Much interest has been stimulated in the interactions of PrP with the glycosami-noglycans heparan sulfate and heparin. This has arisen as a result of the finding (3) that amyloid plaques, which occur in the clinically manifest stage of experimental scrapie (as in the human transmissible encephalopathies and Alzheimer's disease) are rich in heparan sulfate proteoglycans (HSPGs). 1 The effects of administering heparin and heparan sulfate and other highly acidic materials have been investigated on the course of experimental scrapie infections (4,5). It has been shown that such acidic compounds can "cure" infected cells or lengthen the survival of scrapie-infected animals. It is not yet known, however, whether these effects are mediated by direct binding to PrP C and PrP Sc , or by indirect mechanisms. Nor is it known whether endogenous glycosaminoglycans or proteoglycans are involved in the normal physiological interactions with prion protein or in pathological interactions as has been suggested (5). It will be important to elucidate at the molecular level any such interactions as a lead to rational therapeutic designs.
Heparan sulfates are major components of the proteoglycans in nervous tissue occurring at the surface of cells and in the extracellular matrix (6 -8). The carbohydrate chains are negatively charged polymers containing disaccharide units of hexuronic acid and N-acetylated or N-sulfated glucosamine. There are diverse patterns of sulfation along the oligosaccharide chains, there being less sulfation overall in heparan sulfates relative to heparin (9). Evidence has been forthcoming for the possible involvement of a specific carbohydrate sequence of heparan sulfate in scrapie pathogenesis. In an immunohistochemical study aimed at examining the temporal relationship between HSPGs and the known sites of PrP pathology, a mouse model of scrapie was selected, which is characterized by both fibrillar amyloid and nonamyloid PrP Sc (10). Six monoclonal antibodies to HSPGs were investigated of which three were directed to core proteins and three to the carbohydrate chains. The presence of one of the carbohydrate antigens, detected with 10E4 antibody (but not with the other antibodies), correlated strikingly with the areas of abnormal PrP deposition, not only in the amyloid plaques, but also in the earliest detectable lesions in the brain of the infected mice. In particular, the presence of the 10E4 antigen in proximity to the diffuse forms of PrP Sc in neuro-anatomically defined target sites, suggests a close relationship between this determinant on heparan sulfate and the formation of the scrapie lesions.
We are investigating the carbohydrate sequence of the 10E4 determinant by the neoglycolipid (NGL) technology (11) in conjunction with electrospray mass spectrometry (ES-MS). A recent advance involving the generation of oligosaccharide probes in the form of fluorescent NGLs (12) has enhanced the power of the technology in pinpointing and isolating bioactive oligosaccharides among highly heterogeneous populations. Here we describe the identification by this means and the assignment of an unusual carbohydrate sequence in the 10E4 antigen.
Depolymerization of HS by Heparin Lyase III and Fractionation of the Fragments-HS-1 was partially depolymerized by limited digestion with heparin lyase III essentially as described (15), and the reaction stopped at around 40% completion. Briefly, the heparan sulfate (150 mg) was incubated with heparin lyase III (300 mU) in 50 mM sodium phosphate buffer (pH 7.1, 6 ml) containing 0.1 M NaCl at 30°C for 8.5 h. The digestion mixture was desalted on a Sephadex G-10 column (1.6 ϫ 36 cm) eluted with water at a flow rate of 20 ml/h, lyophilized, and fractionated on a gel filtration column of Bio-Gel P-6 (1.6 ϫ 90 cm) eluting with 0.2 M NH 4 Cl (pH 3.5) at a flow rate of 15 ml/h. The eluate was monitored on-line by refractive index and UV at 232 nm. The pooled fractions were lyophilized and desalted on a Sephadex G-10 column (1.6 ϫ 36 cm).
HPLC of HS fragments was carried out on an analytical strong anion exchange (SAX) column, S5-SAX (4.6 ϫ 250 mm, Phase Separations Ltd., Clwyd, UK) using a titanium-lined Gilson liquid chromatograph system fitted with a variable wavelength UV detector monitor at 232 nm. Elution was carried out with a gradient of NaCl (solvent A: 0.2 M NaCl and solvent B: 1.5 M NaCl; pH 3.5): 0 -5% B in 15 min followed by 5-10% B in 15 min at a flow rate of 1 ml/min. The major fractions were collected, desalted on a Sephadex G10 column, and lyophilized.
Removal of ⌬UA from Tetrasaccharides-The 4,5-unsaturated hexuronic residue (⌬UA) at the nonreducing terminus of the HS-1 tetrasaccharides was removed by oxymercuration (16). In brief, to a solution of 25 nmol of tetrasaccharide in 18 l H 2 O was added 2 l of Hg reagent (20 mM mercuric acetate in 130 mM sodium acetate, pH 5.0). The reaction mixture was kept at room temperature for 30 min and immediately lyophilized. Conjugation of the lyophilized reaction product to ADHP was essentially as described above, but the incubation time was limited to 24 h. Preparative TLC (11) was performed to isolate the major NGL products from the two reaction mixtures, and to remove the Hg reagent salts, excess lipid and side reaction products.
Chromatogram Binding Assays-NGLs were resolved by HPTLC using as solvent chloroform/methanol/water, 60:35:8, by volume. Binding of monoclonal antibody 10E4 to NGLs chromatographed on aluminum-backed HPTLC plates (Merck) was assayed essentially as described previously (17) except that Plexigum was not used. In brief, after blocking nonspecific binding sites with 1% casein (Pierce) in Trisbuffered saline, 10 mM Tris-HCl buffer, 150 mM NaCl, pH 8.0 (TBS), the chromatogram was overlaid with 10E4 antibody (10 g/ml) in 1% casein; antibody binding was detected using protein-LA-peroxidase (1/500 of stock) in 1% casein followed by FAST TM 3,3Ј-diaminobenzidine peroxidase substrate. For comparing intensities of antibody binding, immunostained bands were scanned, and the derived peak areas scored using a Shimadzu CS-9000 scanner (Shimadzu Corporation, Kyoto, Microwell Binding Assays-Microwell binding assays were performed essentially as described (18) except that Immulon 4 microwells (Dynex Technologies, Billingshurst, UK) were first precoated by applying to each, 50 l of a 50 g/ml solution of poly-L-lysine in phosphatebuffered saline, 10 mM phosphate buffer, 137 mM NaCl, pH 7.4 (PBS) and incubating for 2 h at 37°C. Glycosaminoglycans in PBS were added at the indicated concentrations and allowed to dry at 37°C for 16 h. After blocking with 3% w/v bovine serum albumin in PBS, binding of 10E4 (10 g/ml) was detected using peroxidase-labeled rabbit antibodies to mouse immunoglobulins (25 g/ml) followed by 2,2Ј-azino-bis(3ethylbenzthiazoline-sulfonic acid) peroxidase substrate. Absorbance was read at 405 nm.
Mass Spectrometry of NGLs-Negative-ion ES-MS and collision-induced dissociation (CID) MS/MS were carried out with a Q-Tof instrument (Micromass Ltd., Altrincham, UK). NGLs were dissolved in chloroform/methanol/water 25:25:8, by volume, typically at a concentration of 10 to 20 pmol/l, of which 5 l was loop-injected. Mobile phase (acetonitrile/0.5 mM aq. NH 4 HCO 3 1:1, v/v) was delivered by a syringe pump at a flow rate of 5 l/min. Cone voltage was maintained at 30 V whereas capillary voltage was at 3 kV. CID spectra were obtained using argon as the collision gas at a pressure of 1.7 bar. The source temperature was kept at 80°C. The collision energy was optimized for maximal structural information, ranging between 25 and 35 V.
Analysis of NGLs resolved on TLC plates was carried out in situ by liquid secondary-ion mass spectrometry on a VG ZAB2-E mass spectrometer (Micromass) as described previously (19).

RESULTS
Initial binding experiments revealed that the preparation of heparan sulfate, HS-1, that had the lowest sulfate content elicited the greatest binding signal with 10E4 antibody (Fig. 1,  inset). No binding signals were detected with chondroitin sulfate A, which has a sulfate content similar to that of HS-1. The HS-1 preparation was selected for further investigation. Partial depolymerization of HS-1 was carried out using heparin lyase III, and this was followed by gel filtration (Fig. 1). Pooled fractions containing the smaller fragments, fractions 1 to 6, were subjected to ES-MS analysis (Table I), and thereby found to be di-to octasaccharides with various N-acetyl and sulfate contents.

10E4 Antibody Binding to HS-1 Fragments
HS-1 oligosaccharide fractions 3 to 6 were converted to fluorescent NGLs, resolved by HPTLC, and examined for 10E4 antibody binding (Fig. 2, A and AЈ). Although little or no inhibitory activity could be detected in the free oligosaccharides at the highest concentrations tested (3 mg/ml, results not shown), antigen-positive components were detected among the NGLs of fractions 3, 5, and 6, but not in fraction 4. Those in fractions 5 and 6 were clearly minor components as they were barely detectable by fluorescence. Fraction 3, which contained tetrasaccharides (Table I) and in which the antigen-positive components appeared relatively more abundant ( Fig. 2A), was selected for further study. The nonderivatized fraction 3 oligosaccharides were further resolved by SAX HPLC (Fig. 3A) and the five most abundant subfractions, 3-1, -2, -3, -5, and -8, were converted into fluorescent NGLs and examined for 10E4 antibody binding (Fig. 2, B and BЈ). A duplex of strongly immunoreactive components was revealed in fraction 3-1, but no immunoreactivity was detected in the four other fractions tested. The mixture of NGLs in fraction 3-1 was resolved by HPLC (Fig. 3B), and peaks designated a and b were found by chromatogram binding (not shown) to correspond to the respective immunoreactive components a and b in Fig. 2  presence of a GlcNAc residue linked to the lipid moiety. Thus, the sequence of the antigen-positive NGL, a, is Knowing that the predominant cleavage site of the highly purified recombinant heparin lyase III is at a glucuronic acid

10E4 Antibody Binding to Disaccharides from Heparin and Heparan Sulfate
There was immunoreactivity in regions of lanes 3, 5, and 6 corresponding to NGLs of disaccharides. As these were minor components, we examined for 10E4 immunoreactivity the NGLs of a series of twelve structurally defined heparin/heparan sulfate disaccharides (Table II and Fig. 6). Among these, the NGL derived from ⌬UA-GlcN gave by far the strongest binding signal. Weak binding (8 -9-fold less intense in this semi-quantitative assay) was recorded to ⌬UA-GlcN(6S), ⌬UA-GlcNS, and ⌬UA-GlcNAc(6S). It should be recalled that, in the NGLs, the inner monosaccharide is in a reduced, ring-opened form. The results are, nevertheless, in accord with the importance of the UA-GlcN sequence in 10E4 antigenicity. Other tentative interpretations of these data are (a) that 6-O-sulfated or N-sulfated glucosamine, or N-acetylglucosamine, but not a disulfated glucosamine, in the penultimate position may weakly cross-react with the nonsubstituted glucosamine, and (b) the antibody cannot accommodate an outer 2-O-sulfated hexuronic acid.

Investigation of the Requirement of the Nonreducing Terminal Hexuronic Acid in 10E4 Antigenicity
The major NGL product obtained after oxymercuration treatment to remove the terminal unsaturated hexuronic acid in the 10E4 antigen-positive tetrasaccharide fraction 3-1 (Fig.  6A, lane 4) gave, by TLC-liquid secondary ion-MS, a [MϪH] Ϫ ion at m/z 1451 and two fragment ions at m/z 1290 and m/z 1114 (not shown) corresponding to the expected sequence, GlcN-UA-GlcNAc-ADHP. The liquid secondary ion mass spectrum of the major NGL product of the similarly treated antigen-negative tetrasaccharide fraction 3-3 (Fig. 7A, lane 3) corresponded to that for GlcNAc-UA-GlcNAc-ADHP, ([MϪH] Ϫ ion at m/ z 1493 and fragment ions at m/z 1290 and m/z 1114).
The lipid-linked trisaccharides derived from the antigenpositive and antigen-negative tetrasaccharides, together with the "parent" tetrasaccharides were examined for 10E4 antigenicity by TLC overlay (Fig. 7B). Immunoreactivity was detected, as before, in the 3-1 tetrasaccharide but not in the trisaccharides. Thus the nonreducing end hexuronic acid in the 3-1 tetrasaccharide is required for 10E4 antigenicity.

DISCUSSION
The 10E4 antigen-active tetrasaccharide is, to our knowledge, the first carbohydrate antigen sequence to have been assigned on heparan sulfate. Unexpectedly, the sequence lacks sulfate and constitutes a unique motif with an internal Nunsubstituted glucosamine. This residue is important for antigenicity of the tetrasaccharide as shown by the lack of antibody binding to the two tetrasaccharide analogs investigated, one of which has a N-acetylated, and the other a N-sulfated glucosamine at this position. The importance of N-unsubstituted glucosamine for eliciting strong 10E4 antibody binding is corroborated by the binding experiments with the series of NGLs derived from heparin/heparan sulfate disaccharides; only the disaccharide IV-H, containing the N-unsubstituted glucosamine shows strong binding. The weak 10E4 binding to the NGLs of the three disaccharides, III-H, IV-S, and II-A, raises the possibility that, in the heparan sulfate polysaccharide, certain of the monosulfated glucosamine and N-acetylglucosamine residues may contribute to antigenicity. From the binding data with the tri-and disaccharides, a further conclusion is drawn, namely that a hexuronic acid on the nonreducing side of the glucosamine is essential for 10E4 antigenicity and that it must be nonsulfated. Collectively our results are in   The sequence of the tetrasaccharide was deduced from the CID MS/MS ions (Fig. 4A) that arose from fragmentation of the NGL. The second immunoreactive neoglycolipid, with a measured molecular mass of 1592.6 Da, was deduced to be a dehydrated analog in which the elements of water are lost from the reducing end GlcNAc or lipid tail (11) during neoglycolipid formation.
accord with the preferential binding of 10E4 to heparan sulfate polysaccharide that is undersulfated.
It is interesting that N-unsubstituted glucosamine was identified by van den Born et al. (24) to be a part of the heparan sulfate antigen recognized by another monoclonal antibody, JM-403. This antibody had been included by McBride et al. (10) in the immunohistochemical studies of scrapie lesions but was found not to overlap at all with 10E4 immunoreactivity. The JM-403 antigen sequence is clearly different from that of the 10E4 antigen, and the complete characterization of the two antigens will be important for identifying the definitive scrapie-associated heparan sulfate sequence.
The occurrence of N-unsubstituted glucosamine as a natural component on HSPGs was considered doubtful until relatively recently. This is because the well documented steps in the biosynthesis of the sugar chains involve the formation of a polymer of repeated -GlcA-GlcNAc-units; this is followed by the action of a bifunctional enzyme that sequentially achieves N-deacetylation and N-sulfation of N-acetylglucosamine residues (25). This appears to be a critical cue for further modifications of the carbohydrate chains, which include the epimerization of glucuronic to iduronic acid, and O-sulfation of the monosaccharides at various positions (25). A sulfatase that would desulfate the glucosamine has not yet been identified. Nevertheless, nonsulfated N-unsubstituted glucosamine has been detected by chemical methods in various preparations of heparan sulfates. A preparation of endothelial HSPGs that was bound by L-selectin was reported to contain such N-unsubstituted glucosamine (26). In quantitative studies (22), N-unsubstituted glucosamine was detected as a variable minor component (1-2 in every 50 monosaccharide residues) in heparan sulfates isolated from different sources. Structural studies with oligosaccharide fragments isolated from heparan sulfates have suggested that such glucosamine residues may be flanked by a glucuronic acid on the reducing and nonreducing side, and located in the boundary between low and high sulfated regions (22). It has been suggested that they may also be interspersed in highly sulfated domains that contain iduronic acid residues (27). The possibility that N-unsubstituted glucosamine occurs naturally on HSPGs, rather than being an artifact of isolation or storage of the glycosaminoglycans has been strengthened by the finding that oligosaccharides containing N-unsubstituted glucosamine can serve as substrates for a newly characterized heparan sulfate O-sulfotransferase (27,28).
Arguably the strongest evidence for the natural occurrence of N-unsubstituted glucosamine is that it has been identified as being a dominant part of the antigens recognized by two monoclonal antibodies, 10E4 and JM-403, each of which immunostains fresh frozen tissues with a distinct pattern (10,13,24,29). Analyses of embryonic and adult tissues revealed that the expression of 10E4 antigen varies markedly during organogenesis and development (13). Thus 10E4 antigen, like a number of monoclonal antibody-defined carbohydrate antigens, is a differentiation antigen, which by analogy with those of the blood group Lewis x and Lewis a series (30 -33) can be predicted to have biological functions. There is strong precedence for assignments of roles to heparan sulfates, for example, as ligands and modulators of the activities of growth factors (34). Just as exogenously added polyamines can deplete PrP Sc from certain scrapie-infected cells (35), we propose that the 10E4 determinant, with its content of a free amine (and in multivalent display on heparan sulfate), is a candidate endogenous ligand for a protein or proteins in the pathway of PrP C to PrP Sc conversion.
Thus it will be important to determine the size and complete sequence of the 10E4 antigen, and investigate its role in the normal and pathological behavior of prion protein. In particular, the immunoreactive oligosaccharide(s) visualized as minor components in our chromatogram binding experiments with hexasaccharide fragments of HS-1 should be characterized, as the initial studies of David et al. (13) indicated the 10E4 antigen contains glucosamine that is N-sulfated. It can be predicted, therefore, that the complete 10E4 antigen encompasses the tetrasaccharide sequence we have identified, flanked on one or other side by additional monosaccharide residues including N-sulfated glucosamine. The sulfation issue in regard to 10E4 antigen sequence will be important to clarify, as it has been suggested (5) that certain endogenous glycosaminoglycans or proteoglycans influence the in vivo behavior of PrP. Moreover, cell biological experiments have indicated that endocytosis of PrP is a prerequisite for the formation of PrP Sc (2) and that negatively charged polysaccharides increase the rate of endocytosis of PrP C (36). Further work is under way to characterize hexasaccharides and longer oligosaccharides for 10E4 antigen activity.