N-unsubstituted glucosamine in heparan sulfate of recycling glypican-1 from suramin-treated and nitrite-deprived endothelial cells. mapping of nitric oxide/nitrite-susceptible glucosamine residues to clustered sites near the core protein.

We have analyzed the content of N-unsubstituted glucosamine in heparan sulfate from glypican-1 synthesized by endothelial cells during inhibition of (a) intracellular progression by brefeldin A, (b) heparan sulfate degradation by suramin, and/or (c) endogenous nitrite formation. Glypican-1 from brefeldin A-treated cells carried heparan sulfate chains that were extensively degraded by nitrous acid at pH 3.9, indicating the presence of glucosamines with free amino groups. Chains with such residues were rare in glypican-1 isolated from unperturbed cells and from cells treated with suramin and, surprisingly, when nitrite-deprived. However, when nitrite-deprived cells were simultaneously treated with suramin, such glucosamine residues were more prevalent. To locate these residues, chains were first cleaved at linkages to sulfated l-iduronic acid by heparin lyase and released fragments were separated from core protein carrying heparan sulfate stubs. These stubs were then cleaved off at sites linking N-substituted glucosamines to d-glucuronic acid. These fragments were extensively degraded by nitrous acid at pH 3.9. When purified proteoglycan isolated from brefeldin A-treated cells was incubated with intact cells, endoheparanase-catalyzed degradation generated a core protein with heparan sulfate stubs that were similarly sensitive to nitrous acid. We conclude that there is a concentration of N-unsubstituted glucosamines to the reducing side of the endoheparanase cleavage site in the transition region between unmodified and modified chain segments near the linkage region to the protein. Both sites as well as the heparin lyase-sensitive sites seem to be in close proximity to one another.

teins are noncovalently inserted into the plasma membrane via penetrating peptide segments or covalently attached via a Cterminal glycosyl-phosphatidyl-inositol anchor as in the glypican family (GLP). To date, six distinct GLP gene products have been demonstrated in mammals and two in Drosophila (1). 2 Mammalian GLP-1 is the only member that is widely expressed in the vascular system (2). The three HS attachment sites in GLP-1 are situated close together and near the membrane anchor within the sequence DDGSGSGSD. Biosynthesis of the HS chains proceeds in a stepwise manner. The serine residues are first substituted with Xyl, and then the common linkageregion, GlcUA-Gal-Gal-Xyl is formed. HS assembly is initiated by a unique ␣GlcNAc-transferase that adds the first GlcNAc. By the alternating addition of GlcUA and GlcNAc, catalyzed by HS-copolymerases, the extended, linear heparan backbone is formed. A unique step in HS biosynthesis is the patchwise exchange of N-acetyl for sulfate on GlcNAc catalyzed by various isoforms of N-deacetylase/sulfotransferase. An intermediate in this reaction is GlcNH 2 . Further modifications of the HS precursor chain then take place, including epimerization of GlcUA to iduronic acid (IdoUA) and O-sulfations at various positions. A vast structural variation is generated by the presence of different enzyme isoforms with a more or less extensive action in specific structural contexts (3).
Early structural studies on HS using degradations with nitrous acid or specific enzymes revealed a common basic domain structure whereby blocks of uninterrupted, consecutive GlcUA-GlcNAc repeats alternated with more modified regions, containing both highly sulfated short segments of IdoUA(SO 4 )-GlcNSO 3 as well as mixed or alternating arrangements (4,5). By using periodate oxidation and alkaline scission of GlcUA in consecutive GlcUA-GlcNAc repeats, we isolated oligosaccharides derived from the mixed regions, which seemed to contain some GlcNH 2 moieties (6). More recent studies using a monoclonal antibody directed against epitopes containing GlcNH 2 have shown that these are located in regions of HS composed of mixed N-sulfated and N-acetylated disaccharide repeats (7). Because GlcNH 2 -GlcUA bonds appear to be resistant to HS lyase degradation (8,9), GlcNH 2 can be localized in the enzyme degradation products. Thus, GlcNH 2 moieties have been found both flanked by GlcUA residues (8) and at the reducing side of IdoUA(2-SO 4 ) (9). Most HS chains analyzed to date seem to have the originally proposed common basic design where Nsulfated regions are separated by intervening unmodified regions (see Fig. 1). Variations in HS structure are due mainly to differences in the content and pattern of O-sulfation (1, 3, 4 -6, 10). Sulfated IdoUA residues (boldface S in Fig. 1) are rare in HS from human vascular endothelial cells (11), but the number of such residues as well as the length of the modified regions increase toward the nonreducing end of the chain (12). It has been proposed (8) that GlcNH 2 moieties are preferentially located on the proximal border (relative to the linkage region) between high sulfated and low sulfated regions (see Fig. 1).
Turnover of HSPG involves partial degradation of HS by endoheparanase, which usually cleaves ␤-D-glucuronidic linkages and is inhibited by suramin (Refs. 13-17 and references therein). There are indications for different cleavage sites; some are near the reducing side, others are near the nonreducing side of the highly modified regions (Fig. 1). The structural features of a cleavage site (2) have been defined as -GlcNR-GlcUA-2-GlcNR-HexUA(2-SO 4 )-GlcNR-, where R is an unspecified substituent and HexUA could be IdoUA or GlcUA (14,17).
We have recently shown (18) that the human vascular endothelial cell line ECV 304 expresses a HSPG with a core protein of 60 -70 kDa, which is recognized by an anti-GLP-1 monoclonal antibody and by a polyclonal antiserum raised against recombinant GLP-1 protein. By using metabolic radiolabeling and compounds that inhibit intracellular progression of secretory proteins or degradation of HS, effects on the turnover of different GLP-1 glycoforms were studied. As summarized in Fig. 2 we have proposed a recycling of GLP-1 (18). In unperturbed cells, most of the radiolabeled GLP-1 carries truncated HS chains and is accompanied by HS oligosaccharides. In BFAtreated cells, a large GLP-1 PG with long HS chains accumulates and the oligosaccharides disappear. The long HS chains contain scattered and clustered GlcNH 2 moieties. However, very little metabolic radiolabeling of the core protein of this PG could be obtained. Results of pulse-chase experiments indicated that the large-size PG is the precursor of a partly HSdegraded PG form that was captured by suramin treatment. When cells pretreated with suramin were chased in the presence of BFA, formation of large PG with long HS chains is resumed with increased capacity. The latter process is retarded by continued presence of suramin and completely blocked in cells deprived of nitrite. Resynthesis of HS is restored when an NO donor was supplied. Radiolabeling of GLP-1 core protein could only be detected after pulse-labeling of protein in nitrite-deprived cells followed by chase-labeling of the HS chains in BFA-treated cells provided with an NO donor.
Endogenous nitrite is formed by oxidation of NO (19), and endothelial cells generate sufficient amounts to degrade exogenously supplied HS at moderately acidic pH values (20). Cleavage at GlcNH 2 moieties of endogenously formed HS during degradation and turnover is thus a distinct possibility. In the present study we set out to examine whether the content of GlcNH 2 in the HS chains/stubs of various GLP-1 glycoforms was affected by nitrite deprivation and where in the HS chains these susceptible moieties were preferentially located.

EXPERIMENTAL PROCEDURES
Materials-Cells, culture media, sera, antiserum to glypican-1, the various enzyme inhibitors and other inhibitory compounds, radioactive precursors, enzymes, prepacked columns, and other media or chemicals were the same as described previously (18). In some experiments D-  H]glucosamine with a specific activity of 60 Ci/mmol (Amersham Pharmacia Biotech) was used, and prepacked columns of Superdex peptide HR 10/30 and PD-10 were also obtained from Amersham Pharmacia Biotech, Sweden. Centriplus 30 was from Millipore, Sweden. Gal␤1-3GalNAcO[ 3 H] was a gift from the former BioCarb Company.
Cells, Media, and Radiolabeling-ECV 304 cells were maintained in DMEM as described previously (18) and preincubated with the appropriate medium before radiolabeling. [  Treatments with BFA, suramin, NOS inhibitors, neocuproine, and sulfamate were carried out as described previously (18).
Extraction and Isolation of PG-Cells were extracted with phosphate-buffered saline containing 0.1% (w/v) SDS, 0.5% (v/v) Triton X-100, 0.5% (w/v) sodium deoxycholate (RIPA) followed by immunoisolation of GLP-1 glycoforms using anti-GLP-1 antiserum in three consecutive steps performed as described previously (18). Regularly, PG and PG-derived material not reacting with the antiserum were recovered from the supernatants obtained after immunoisolation by passage over DEAE-cellulose as described (18). Separation into PG and degradation products was achieved by gel-permeation chromatography on Superose 6, by further purification of PG material by ion exchange chromatography on MonoQ (gradient elution), and sometimes by hydrophobic interaction chromatography on octyl-Sepharose (12,18). In some experiments, material extracted with only Triton X-100-contain-  10 -17 repeats from the linkage region (--, GlcUA-Gal-Gal-Xyl) and then alternate with regions containing GlcNSO 3 moieties, either in consecutive repeats (solid) or mixed ones (hatched). The total length is generally from 50 to 100 disaccharides, but precursor forms may contain up to 200. Repeats containing IdoUA(-SO 4 ) are only found centrally within the modified block regions (marked with boldface S). Proposed sites for GlcNH 2 are marked with ϩ. Endoheparanase cleavage sites are marked with arrows (open for those acting on the nonreducing side, filled for those acting on the reducing side). Sites for cleavage by HS lyase should be abundant in the unmodified regions (except for those marked ϩ) and also sometimes in the mixed ones. Heparin lyase cleaves at boldface S (see also Refs. 10 -17).
FIG. 2. Proposed pathway for degradation and recycling of GLP-1 in ECV cells. A PG comprising GLP-1 with long HS side chains that contain N-unsubstituted glucosamine residues (GlcNH 2 ) accumulates in cells treated with BFA (top left). Degradation of HS by endoheparanase commencing at peripheral sites (top right) results in the formation of HS-chain fragments and a core protein with truncated side chains. Degradation appears to proceed until the GlcNH 2 residues are near the nonreducing end of the stubs (bottom right). We propose that the nonreducing portion comprising the GlcNH 2 residues is removed by endogenously produced nitrite (bottom left). Resynthesis of a large-size PG occurs by extension of the HS chain stubs, and new GlcNH 2 residues are introduced (18).
ing buffers (18) was directly subjected to the biochemical isolation procedure. Cells were sometimes released by trypsinization before extraction with Triton X-100. Trypsin digestion was carried out using 20 g/ml for 5 min at 37°C. The reaction was terminated by the addition of serum (10%, v/v) followed by centrifugation at 3000 rpm for 5 min. Cells and supernatants were analyzed separately.
Degradations and Separations-HS chains and chain stubs were released from the core protein by treatment with alkaline borohydride. Enzymatic digestions of HS were performed with HS or heparin lyase and deaminative cleavage with HNO 2 either at pH 1.5 or 3.9. PG, HS, and degradation products were separated by gel-permeation chromatography. All procedures have been described in detail elsewhere (12,18). Buffer changes, concentration, and recovery of material were performed using Centriplus 30 or PD-10 columns as described by the manufacturer or by precipitations with 10 volumes of ethanol/0.4% (w/v) sodium acetate. As carriers (25-100 g/ml) we used HS chains (18) and/or an oligosaccharide mixture consisting of equal parts of HS degraded with HNO 2 at pH 1.5, raffinose, sucrose, and glucose. Radioactivity was measured by ␤-scintillation.

Biosynthesis and Characterization of GLP-1 with Long and
Short HS Chains-BFA can inhibit both transport of newly synthesized proteins to the Golgi (21) as well as endocytosis, recycling, and reprocessing of resident membrane proteins in polarized cells (18,22). Therefore, we first assessed the proportions of extracellular and intracellular PG material in untreated or BFA-treated cells. ECV cells were incubated with [ 35 S]sulfate for 24 h in the absence or presence of BFA. They were released from the culture plates by trypsinization and separated from solubilized material by centrifugation. The cell pellet was lysed in Triton X-100, and PG and PG-derived material were recovered separately from the trypsinate and the cell lysate by passage over DEAE-cellulose. In the case of untreated cells, 22% of total radiolabeled material was released by trypsin, and most of this was eluted as a small PG on Superose 6, whereas the inaccessible material consisted of HSoligosaccharides (data not shown). In brefeldin A-treated cells, no radiolabeled material was released by trypsin. All of the radiolabeled material recovered from the cell pellet consisted of a large PG (18).
Previous studies have shown that the PG from BFA-treated cells (comprising GLP-1) is substituted with long HS sidechains, whereas the predominant glycoform from untreated cells has shorter HS chains (18). In this study, we investigated whether the content of GlcNH 2 residues in the PG chains varied with the nitrite level in the cells producing the PG. To obtain radiolabeled chains, [ 3 H]glucosamine-and [ 35 S]sulfatelabeled GLP-1 glycoforms were isolated on a preparative scale by three consecutive immunoisolations followed by gel-permeation chromatography on Superose 6. In the case of untreated cells (Fig. 3), it was estimated that at least 75% of all GLP-1 glycoforms had been isolated (A-C). These profiles were also GLP-1 antiserum. Three consecutive immunoisolations were made, and the samples were separately subjected to gelpermeation chromatography on Superose 6 (A, B, and C, respectively). The same aliquots were taken for counting of radioactivity in A-C. The materials in the major peak (see bars in A-C) were pooled, degraded with alkaline borohydride, and rechromatographed (D). The released glycan chains were pooled (see bar in D) and again rechromatographed before (E) and after (F) treatment with HNO 2 at pH 3.9.
similar to those obtained previously with total radiolabeled polyanionic material isolated from Triton X-100 cell extracts by ion-exchange chromatography (18), except that no accompanying HS chain fragments and oligosaccharides were obtained (see dashed line without symbols in C). The HS degradation products were recovered from the supernatant after immunoisolation (data not shown). The HS chains of the immunoisolated small GLP-1 glycoform from untreated cells (see bars in Fig. 3, A-C) were released by alkali and chromatographed on Superose 6 (Fig. 3D). The chain size of the major pool (see bar in D) was estimated to be 30 -50 kDa (23). Some smaller chain stubs were also present (fractions 45-55).
The large-size radiolabeled PG from BFA-treated cells was much less reactive with the antiglypican antiserum. Only 5-10% was recovered after three consecutive immunoisolations. However, a 66-kDa-core protein, derived from the PG of BFA-treated cells by digestion with HS lyase, reacted strongly with monoclonal antibodies to GLP-1 upon SDS-polyacrylamide gel electrophoresis and immunoblotting (18). It is thus possible that the long HS side chains could shield epitopes in the core protein. Unfortunately, the PG from BFA-treated cells incorporates very little radioactivity from radiolabeled amino acids, because it appears to be primarily made from resident, unlabeled core protein precursors or HS-chain-truncated PG (18).
Because the yield of immunoreactive intact PG from BFAtreated cells was low, the HS chains from immunoisolated as well as unreactive PG material were analyzed separately. After gel-permeation chromatography on Superose 6 ( Fig. 4) largesize PG material, eluting in the void volume, was obtained in both cases (pool I in A and a corresponding pool in B, see bar). Also a low molecular size [ 3 H]glucosamine-and [ 35 S]sulfatelabeled GLP-1 glycoform was obtained from the immunoisolated material (pool II in A) but not from the unreactive material (B). The peaks at 29 -32 and 40 -43 in Fig. 4B are probably glycoprotein contaminants.
The elution profiles of the alkali-released HS chains from the large immunoreactive GLP-1 glycoform (pool I in Fig. 4A) and from corresponding material that was unreactive to the antiserum (see bar in Fig. 4B) as well as that of HS-chains from the small immunoreactive GLP-1 glycoform (pool II in Fig. 4A) are shown in Fig. 4, C, D, and G, respectively. The large PG forms contained long HS side chains in both cases (C and D) with an estimated average size of approximately 100 kDa (23). The  (18). Radiolabeled GLP-1 was immunoisolated from a detergent extract (RIPA) of the cells and chromatographed on Superose 6 (A). Radiolabeled polyanionic material was also recovered from the supernatant of the immunoisolates by passage over DEAE-cellulose and chromatographed on Superose 6 (B). Pooled high molecular weight proteoglycan material (pool I in A and corresponding pool in B, see bars) was degraded with alkali and rechromatographed (C and D, respectively). The released glycan chains were pooled (see bars), treated with HNO 2 at pH 3.9, and again rechromatographed (E and F, respectively). The low molecular weight immunoisolated material (pool II in A) was also degraded with alkali and rechromatographed (G). 3 small immunoreactive GLP-1 glycoform yielded a broad-size distribution on Superose 6 ( Fig. 4A), suggesting a polydisperse HS chain composition. The chains from this PG separated into three major and one or two minor-size pools as judged from the distribution of the [ 3 H]glucosamine label, which represents the backbone structure of the chains (Fig. 4G). Most of these, except for a minor portion, were of the same size or smaller than the PG (cf. pool II in Fig. 4A). However, the chains differed markedly in degree of sulfation. The smallest chains, eluting around fraction 50, were mostly nonsulfated, whereas some of the larger, minor ones, eluting around fraction 35, had a [ 35 S]/ [ 3 H] ratio of approximately 4:1. Hence, the small GLP-1 glycoform from BFA-treated cells contained chains and chain stubs varying in size from 60 kDa down to less than 5 kDa (23). Apparently, the core protein simultaneously carried both short and long chains as the intact PG did not separate into discrete components. Moreover, each PG molecule should only contain one long chain. This chain would dominate the excluded volume of the PG, because the core protein is expected to have a rather compact structure (1). The presence of short HS chains may have facilitated its essentially quantitative recovery by immunoisolation. The small PG may be an immature form trapped in an early compartment of the secretory pathway.
Identification of GlcNH 2 in Different GLP-1 Glycoforms-The presence of GlcNH 2 was indicated by selective deaminative cleavage at the reducing side of these residues in alkali-released HS chains using HNO 2 at pH 3.9 (24) followed by gelpermeation chromatography on Superose 6. The small GLP-1 glycoform obtained from unperturbed cells (Fig. 3, A-C) yielded a major pool of HS-chains (see bar in Fig. 3D) with an estimated size of 30 -50 kDa. These chains were rechromatographed before (Fig. 3E) and after (Fig. 3F) deaminative cleavage. The results showed that most of the chains were resistant, but a small proportion had been cleaved into medium-sized fragments (fractions 45-50).
The long HS side chains of the immunoisolated large GLP-1 glycoform from BFA-treated cells (Fig. 4C) were extensively depolymerized upon deaminative cleavage at pH 3.9 (Fig. 4E), indicating that GlcNH 2 residues were present in all chains to a variable degree. The degradation products included fragments of a relatively uniform distribution (fractions [25][26][27][28][29][30]. However, the major products were a series of intermediate-size (fractions 30 -45, approximately 20 -60 kDa) to small-size fragments (fractions 45-50, approximately 10 kDa) with gradually diminishing [ 35 S]/[ 3 H] ratios. The latter fragments should be derived from the unmodified regions (see Fig. 1). The HS chains of the material that was not immunoreactive (Fig. 4D) yielded essentially the same type of products after deaminative cleavage (Fig. 4F). Hence, any syndecan-derived HS that might be present in the PG pool (18) may also contain a similar amount of GlcNH 2 residues. It should also be pointed out that nitrous acid causes N-desulfation and deaminative cleavage at GlcNSO 3 residues under more acidic conditions, such as pH 1.5 (24). Release of free sulfate would be indicative of such a reaction. In all cases (Figs. 3F, 4E, and 4F) very little, if any, free sulfate could be detected (see below).
The [ 3 H]glucosamine-and [ 35 S]sulfate-labeled GLP-1 glycoforms from suramin-treated cells arise from the larger PG precursor by partial endoglycosidic degradation of HS (Fig. 2,  top right). These truncated HS chains were also examined for the presence of GlcNH 2 residues. PG and PG-derived material not reacting with the antiserum and immunoreactive material were separately chromatographed on Superose 6 (Fig. 5, A and  B, respectively). The immunoisolated material (Fig. 5B) afforded a profile similar to that of GLP-1 from unperturbed cells (Fig. 3, A-C). The supernatant obtained after immunoisolation kDa with an average at 40 kDa). A comparison with the profile obtained previously (18) with the total radiolabeled, polyanionic material from suramin-treated cells (see dashed line in Fig. 5A) indicated that approximately half of the PG material from these cells was immunoreactive.
The alkali-released HS chains of the immunoisolated GLP-1 glycoform from suramin-treated cells had approximately the same average size (Fig. 5C) compared with those obtained from the major GLP-1 glycoform of untreated cells (Fig. 3D). Similarly, there was only a slight shift in the profile when HS chains of GLP-1 from suramin-treated cells were subjected to deaminative cleavage (Fig. 6D). Essentially the same result was obtained with chains derived from the material that was not immunoreactive (data not shown).
The marginal effect of nitrous acid on the partially degraded HS side chains of GLP-1 from suramin-treated cells may be due to prior cleavage near the GlcNH 2 residues by an endoheparanase not inhibited by suramin. Alternatively, endogenously formed nitrite may have consumed all available GlcNH 2 sites. Therefore, we also examined the GLP-1 glycoforms obtained from nitrite-deprived cells as well as those from both suramintreated and nitrite-deprived cells. The GLP-1 glycoforms from cells depleted of nitrite by treatment with NOS inhibitor, neocuproine, and sulfamate ( Fig. 6A) were more size-heterogeneous than corresponding material from untreated (Fig. 3, A-C) or suramin-treated cells (Fig. 5B). Still, a substantial portion was recovered by immunoisolation (cf. Fig. 6A and bar  in B). The HS chains derived from these glycoforms (Fig. 6C) were also more heterogeneous and somewhat larger (average size of approximately 50 kDa) than corresponding material from untreated (Fig. 3D) or suramin-treated cells (Fig. 5C). However, despite nitrite deprivation, very little cleavage was observed after treatment with HNO 2 at pH 3.9 (Fig. 6D).
GlcNH 2 residues and sites for endoheparanase cleavage could be adjacent or very closely located on HS chains. Therefore, the HS chains of GLP-1 glycoforms produced by nitritedeprived cells may indeed contain GlcNH 2 residues, but cleavage nearby by endoheparanase would make them difficult to detect with the present approach. We therefore examined HS from glycoforms made by cells that were simultaneously nitrite-deprived and treated with suramin to inhibit endoheparanase. The GLP-1 glycoforms isolated from these cells were heterogeneous and included some large PG material (Fig. 7A). In the supernatant obtained after immunoisolation, there were both unreactive material of the same size as the reactive material (pool I in Fig. 7B) and HS chain fragments (pool II). When material from the latter pool was recovered and treated with HNO 2 at pH 3.9, no effect was seen (data not shown).
The immunoisolated glycoforms from nitrite-deprived and suramin-treated cells (Fig. 7A) were quantitatively bound to octyl-Sepharose. To remove any free HS chains and chain fragments from the unreactive PG material in pool I (Fig. 7B), this material was also passed over octyl-Sepharose. Pooled octylbound materials from both immunoreactive and unreactive material were treated with alkali, and released HS stubs were rechromatographed (Fig. 7C). The average size and size-distribution of these chains were essentially the same as for corresponding material from only nitrite-deprived cells (Fig. 6C). However, the HS-material derived from simultaneously suramin-treated and nitrite-deprived cells was more sensitive to HNO 2 at pH 3.9 (Fig. 7, C and D) than were chains derived from cells treated with only suramin (Fig. 5, C and D) or only nitrite-depleted (Fig. 6, C and D). The HS chains derived from PG obtained after the combined treatment were therefore used to locate the GlcNH 2 residues (see below).
Location of GlcNH 2 Residues in HS Chains-If HS chains have the general domain structure depicted in Fig. 1, the fully modified regions (filled boxes) should contain IdoUA  residues joined at their nonreducing side to GlcNSO 3 constituting sites for cleavage by heparin lyase (marked with boldface S). Fragments generated from distal or middle portions of an HS chain by cleavage with heparin lyase can be separated from proximal HS stubs linked to the core protein by passage over octyl-Sepharose, which adsorbs the latter material (12). Thus, octyl-Sepharose-bound glycoforms, obtained from cells subjected to combined suramin treatment and nitrite deprivation, were digested with heparin lyase and again passed over octyl-Sepharose. Approximately 20% of the material (based on [ 3 H]glucosamine) passed through the column, indicating that heparin lyase-sensitive sites were relatively rare. The released HS fragments were subjected to gel-permeation chromatography on Superose 6 (Fig. 8A). Most of the material consisted of fragments smaller in size (approximately 30 kDa on an average) than those generated endogenously in suramin-treated cells (Fig. 5A). The heparin lyase-generated HS fragments were insensitive to nitrous acid at pH 3.9 (Fig. 8B), indicating that GlcNH 2 residues are rare distal to the heparin lyase cleavage sites (Fig. 1).
If a HS chain has the structure shown in Fig. 1, the octyl-Sepharose-bound heparin lyase degradation products should consist of the core protein substituted with more or less truncated HS chains. These chain stubs should consist of a major stretch of unmodified GlcUA-GlcNAc repeats (open box) immediately following the linkage region and sometimes extended with modified regions. The unmodified region should contain multiple sites for cleavage by HS lyase. Hence, the heparin lyase-degraded PG material was further degraded by treatment with HS lyase, and the products were again passed over octyl-Sepharose. More than 80% of the material passed through the column, indicating that nearly all of the radiolabeled HS chains had been released from the core protein. Because the released fragments should consist of small oligosaccharides, they were chromatographed on Superdex peptide (Fig. 8C). The material separated into one excluded fraction (pool I) and a series of included smaller-size saccharides (pool II). The former material, probably decasaccharides (approximately 2.5 kDa or larger), was treated with HNO 2 at pH 3.9 and rechromatographed (Fig. 8D). The results showed extensive degradation into di-, tetra-, and hexasaccharides, indicating the presence of clustered GlcNH 2 residues that had survived HS lyase treatment. Although the fragments generated by HS lyase had low [ 35 S]/[ 3 H] ratios, some sulfate could reside in surviving GlcNSO 3 residues. To test this, materials from pools I and II were treated with HNO 2 at pH 1.5. As shown in Fig. 8, E and F, respectively, they were completely resistant. Therefore, the saccharides containing GlcNH 2 may also contain some O-sulfate. Taken together, the results obtained so far indicate that clustered GlcNH 2 residues are concentrated to a proximal region between the core protein and the first heparin lyase-sensitive site (Fig. 1). To determine the placement of the . This material was quantitatively bound to octyl-Sepharose (data not shown). Radiolabeled polyanionic material was also recovered from the supernatant of the immunoisolate by passage over DEAE-cellulose and chromatographed on Superose 6 (B). Material from the latter preparation and corresponding in size to the immunoisolated material (pool I) was purified by passage over octyl-Sepharose. Pooled octyl-bound materials were treated with alkali and rechromatographed (C). Released glycan chains were pooled (see bar), treated with HNO 2 at pH 3.9, and rechromatographed (D). Material from pool II in B was also treated with HNO 2 at pH 3.9 and again rechromatographed. No change in elution position was observed (data not shown). 3 H (E----E); 35 S (f--f)V o , void volume; V t , total volume. endoheparanase cleavage sites relative to the clustered Gl-cNH 2 residues, the following experiments were performed.
Large-size radiolabeled PG was isolated and purified from BFA-treated ECV cells. Because the long radiolabeled HS chains produced in BFA-treated cells appear to be made by extension of pre-existing, unlabeled HS stubs (18), cells were first incubated with [ 3 H]glucosamine in the absence of BFA to maximize labeling of the HS chain stubs. Then the cells were incubated with [ 35 S]sulfate in the presence of BFA to radiolabel HS chain segments made during accumulation of the large PG form. The doubly labeled purified PG was then incubated with ECV cells to allow cell-surface-located (if present) endoheparanase to degrade its HS chains. As a control, incubations were also made in the presence of suramin. The results showed (Fig.  9) that exogenously supplied PG was degraded to medium-size PG-like (pool II) and oligosaccharide-like (pool III) material (Fig. 9A). No degradation took place in the presence of suramin (Fig. 9B), indicating that the degradation was caused by endoheparanase. The degradation products in pool II (Fig. 9A) were sensitive to alkali (Fig. 9C) confirming their PG nature. The material in pool III (Fig. 9A) was alkali-resistant when analyzed on Superose 6, confirming their oligosaccharide nature (data not shown). They were probably of decasaccharide size or larger as judged from the Superdex peptide chromatogram (Fig. 9D). The HS chains derived from the PG degradation product (Fig. 9C) were treated with nitrous acid at pH 3.9 and rechromatographed on Superose 6 (Fig. 9E). The results showed extensive degradation into both sulfate-rich and sulfate-poor fragments. When compared with the profiles obtained after deaminative cleavage of the HS chains of the substrate (Fig. 4, E and F) it is evident that the longer chain fragments (eluting in fractions 20 -30) were missing, indicating that these were released by endoheparanase and should be recovered in the oligosaccharide fraction (pool III in Fig. 9A). When the oligosaccharide material was subjected to deaminative cleavage, no degradation could be observed upon chromatography on Superose 6 (data not shown). To look for release of very small saccharides, the oligosaccharides were also chromatographed on Superdex peptide after treatment with nitrous acid at pH 3.9 (Fig. 9F). Again, no sign of degradation could be observed (cf. Fig. 9D). These results thus indicate that the clustered GlcNH 2 residues are located between the core protein and the first endoheparanase cleavage site (Fig. 1).
A PG substrate that was prepared from cells radiolabeled with [ 3 H]glucosamine and [ 35 S]sulfate simultaneously and in the presence of BFA was also used for incubations with ECV cells. The HS chains of this PG should mainly be radiolabeled in the peripheral regions (18). Using this PG substrate, only radiolabeled HS oligosaccharides but no radiolabeled PG-type degradation products were obtained (data not shown). This is consistent with the idea that the large PG in BFA-treated cells is made by extension of existing chains on a preformed PG and that the extension starts somewhere near an endoheparanase cleavage site.

DISCUSSION
Most GLP-1 glycoforms in unperturbed ECV cells contain HS chains that rarely exceed 50 kDa in size. However, some of the FIG. 8. Localization of N- Fig. 7B (pool I). After further purification by passage over octyl-Sepharose, bound material was digested with heparin lyase and again passed over octyl-Sepharose. Unbound degradation products were chromatographed on Superose 6 before (A) and after treatment with HNO 2 at pH 3.9 (B). Bound degradation products obtained after digestion with heparin lyase were further digested with heparan sulfate lyase and again passed over octyl-Sepharose. Unbound degradation products were chromatographed on Superdex peptide (C). Material in pool I was treated with HNO 2 at pH 3.9 and rechromatographed (D). Materials in pools I and II were also treated with HNO 2 at pH 1.5 and rechromatographed in E and F, respectively. 3  three available sites may contain shorter stubs indicating that elongation of HS chains on the different sites may not be synchronized. Moreover, because ECV cells should endogenously produce nitrite, it is perhaps not surprising that only a few GlcNH 2 residues were detected. Hence, the short stubs may also be the result of deaminative cleavage.
The HS chains of the large PG from BFA-treated cells were approximately twice the size of those of the smaller PG, and GlcNH 2 residues occurred in multiple places and were, surprisingly, more abundant than in chains obtained from nitritedeprived cells. The reason for this could be that nitric oxide and subsequently nitrite were not produced in BFA-treated cells or that the PG was inaccessible to nitrite. Moreover, large Gl-cNH 2 -rich PG exogenously supplied to ECV cells suffered endoheparanase-catalyzed cleavage without substantial loss of GlcNH 2 residues, probably because the substrate and/or nitrite concentration were not sufficiently high in this situation. In contrast, when endogenously produced PG was degraded by endoheparanase during pulse-chase experiments, most of the clustered GlcNH 2 residues were lost, probably by deaminative cleavage.
Structural analysis of the HS chains from the PG produced in BFA-treated cells as well as of the HS stubs on the endoheparanase-generated PG product from exogenously supplied PG indicated that clustered GlcNH 2 residues were concentrated to the region between the core protein and the first sites cleaved by endoheparanase and heparin lyase (Fig. 1). The analysis of the products also indicated that mainly the external portions of the HS chains are removed and replaced during recycling (Fig. 2). Similar observations were made with GLP from human fibroblasts (25,26).
Recently, a mammalian endoheparanase with endo-␤-glucuronidase specificity (17) was cloned (27)(28)(29)(30)(31). Because only one gene was found and the protein appeared unique, it was assumed that the same enzyme is expressed in both normal and transformed cells. The enzyme, which localizes to both endosomes and the cell-surface recognizes the sequence -GlcUA-GlcNR-HexUA(2-SO 4 )-GlcNR (17). There are two principal varieties of this sequence where HexUA is either GlcUA or IdoUA. Therefore, the observations that there are multiple endoheparanase activities and that some sites are more preferred (15,16) could be ascribed to the presence of factors that bind to the enzyme and modulate its specificity. Alternatively, basic compounds such as polyamines and growth factors that bind to specific sequences in HS may protect certain sites from cleavage (1,3,32).
Although endoheparanase is supposed to be inhibited by suramin (13), the GLP-1 glycoforms obtained from suramintreated ECV cells contained truncated HS chains and were accompanied by HS degradation products indicative of some endoheparanase cleavage. It has been reported that human microvascular endothelial cells take up suramin via caveolae

FIG. 9. Localization of N-unsubstituted glucosamines in degradation products obtained after incubating large [ 3 H]-, [ 35 S]GLP-1 with ECV cells.
Large-size GLP-1 was obtained from ECV cells (175-cm 2 dish) that were preincubated for 24 h with [ 3 H]glucosamine and then with [ 35 S]sulfate in the presence of BFA for another 24 h. PG was isolated from a detergent extract (Triton X-100) of the cells by passage over DEAE-cellulose followed by size chromatography on Superose 6 and ion-exchange chromatography on MonoQ as described (18). Purified PG was dissolved in serumfree DMEM and added to 75-cm 2 dishes of ECV cells and incubated in the absence or presence of suramin for 24 h. Medium was then decanted, and the cell layer was extracted with Triton X-100, concentrated, and chromatographed on Superose 6. The results obtained with untreated and suramin-treated cells are shown in A and B, respectively. Products obtained in A were pooled into fractions I, II, and III as indicated by bars. Material in pools II and III were recovered, treated with alkali, and rechromatographed on Superose 6. Results obtained with material from pool II are shown in C. Alkalitreated material from pool III did not change its position on Superose 6 (data not shown) and was eluted in the void volume of Superdex peptide (D). The alkali-treated materials from pools II and III were finally subjected to deaminative cleavage with HNO 2 at pH 3.9 and rechromatographed on Superose 6 (E) and Superdex peptide (F), respectively. 3 H (E----E); 35 S (f--f); V o , void volume; V t , total volume. and that it is transported to the nucleus (33). It is thus possible that suramin does not have access to endoheparanase acting on HS when GLP-1 is recycling through, e.g. early endosomes. However, cell-surface-located enzyme was inhibited by suramin as shown in this study. Taken together, the results suggest that the potential endoheparanase site closest to the linkage region and near the clustered GlcNH 2 residues (filled arrow in Fig. 1) is not cleaved.
The HS fragments released from peripheral regions of the chains by endoheparanase or heparin lyase did not seem to contain much GlcNH 2 , at least not in internal positions. However, the precursor of the suramin-arrested material, i.e. the large PG in BFA-treated cells, contained a heterogeneous population of HS chains, some of which must contain GlcNH 2 residues also in more distal transition regions (Fig. 1). We had expected that the number of GlcNH 2 residues should be significantly increased in HS produced in nitrite-deprived cells. However, this was not the case unless the cells were simultaneously treated with suramin. Taken together, these results suggest that endoheparanase cleavage sites and single GlcNH 2 residues could be placed closely together (see Fig. 1, filled arrows). The finding that HS-chain fragments obtained after either endoglycosidic or deaminative cleavage had similar size distribution also supports this notion. Results of heparin lyase degradation indicated that also some of these sites were present in the near vicinity. This is in agreement with previous results (23), which indicated that heparin lyase and endoheparanase cleavage sites are juxtaposed. Hence, HS chains may contain "hotspots" where sites for cleavage by endoheparanase, heparin lyase, and HNO 2 are close together as in -GlcNH 2 -GlcUA-GlcNSO 3 -IdoUA(2-SO 4 )-GlcNH 2 . The GlcNH 2 residue at the reducing end is a potential target for 3-O-sulfation (9). GlcNH 2 residues could also be present in segments not susceptible to heparin lyase, such as -GlcNH 2 /GlcNR-GlcUA-GlcNH 2 /GlcNR-GlcUA(2-SO 4 )-GlcNH 2 /GlcNR, where the glucosamine residue on the nonreducing side of GlcUA(2-SO 4 ) is probably N-sulfated. There could also be consecutive GlcUA-GlcNH 2 repeats or they could be alternating with GlcUA-GlcNAc repeats. Indications of the latter two arrangements were found in this study. The formation of GlcNH 2 residues may be the result of insufficient N-sulfation after N-deacetylation, caused by a specific N-deacetylase/sulfotransferase isoform or by an endo-sulfamidase removing certain N-sulfate groups after completion of synthesis.
As mentioned above, the GlcNH 2 -containing repeats that were clustered near the linkage region may inhibit further erosion by endoheparanase (14,17). This may give rise to HS stubs that are poor primers for re-elongation (18). Formation of NO at relatively high concentrations within the endothelial cell should result in rapid conversion to nitrite before NO diffuses out of the cell (18,19). It is thus possible that NO indirectly, via nitrite, regulates recycling by removing inhibitory "telosaccharides," thus facilitating re-elongation. Recycling GLP-1 may be a vehicle for endo-/exocytosis of growth factors and morphogens constituting an additional function of HSPG in signaling and developmental patterning (1,3,32,34). Deaminative cleavage at GlcNH 2 gives rise to reducing terminal anhydromannose (24) in the released oligosaccharides and/or, in combination with endoheparanase, free anhydromannose depending on the structural context (see above). This unit and its reduced ver-sion anhydromannitol have not yet been found in cells, neither in free form nor at the reducing end of an HS oligosaccharide. However, exogenously supplied anhydromannitol is phosphorylated by phosphofructokinase to anhydromannitol-1-phosphate and -1,6-bisphosphate (Ref. 35 and references therein). This causes inhibition of gluconeogenesis and glycogenolysis and stimulation of glycolysis (35)(36)(37). The results described in the present study thus invite inquiries into the possibility of a connection between HS turnover and energy metabolism.