Interaction of thrombospondin-1 and heparan sulfate from endothelial cells. Structural requirements of heparan sulfate.

Cell surface-associated heparan sulfate proteoglycans, predominantly perlecan, are involved in the process of binding and endocytosis of thrombospondin-1 (TSP-1) by vascular endothelial cells. To investigate the structural properties of heparan sulfate (HS) side chains that mediate this interaction, the proteoglycans were isolated from porcine endothelial cells and HS chains obtained thereof by beta-elimination. To characterize the structural composition of the HS chains and to identify the TSP-1-binding sequences, HS was disintegrated by specific chemical and enzymatic treatments. Cell layer-derived HS chains revealed the typical structural heterogeneity with domains of non-contiguously arranged highly sulfated disaccharides separated by extended sequences containing predominantly N-acetylated sequences of low sulfation. Affinity chromatography on immobilized TSP-1 demonstrated that nearly all intact HS chains possessed binding affinity, whereas after heparinase III treatment only a small proportion of oligosaccharides were bound with similar affinity to the column. Size fractioning of the bound and unbound oligosaccharides revealed that only a specific portion of deca- to tetradecasaccharides possessed TSP-1-binding affinity. The binding fraction contained over 40% di- and trisulfated disaccharide units and was enriched in the content of the trisulfated 2-O-sulfated L-iduronic acid-N-sulfated-6-O-sulfated glucosamine disaccharide unit. Comparison with the disaccharide composition of the intact HS chains and competition experiments with modified heparin species indicated the specific importance of N- and 6-O-sulfated glucosamine residues for binding. Further depolymerization of the binding oligosaccharides revealed that the glucosamine residues within the TSP-1-binding sequences are not continuously N-sulfated. The present findings implicate specific structural properties for the HS domain involved in TSP-1 binding and indicate that they are distinct from the binding sequence described for basic fibroblast growth factor, another HS ligand and a potential antagonist of TSP-1.

least five proteins with modular and multidomain structure (reviewed in Ref. 1). TSP-1 was initially found in ␣-granules of platelets where it is involved in aggregation and clot formation (2). Later on, the expression of the 450-kDa trimeric glycoprotein by a variety of cells including endothelial cells, fibroblasts, smooth muscle cells, and macrophages has been observed (3)(4)(5)(6). The protein is secreted and incorporated into the extracellular matrix.
The modular organization of TSP-1 enables binding to cells, platelets, and macromolecules, such as different types of collagen, fibronectin, fibrinogen, sulfated glycolipids, and heparan sulfate proteoglycans (reviewed in Refs. 7 and 8). In addition to the role of TSP-1 during platelet aggregation, the molecule is involved in the regulation of cell adhesion, migration, and proliferation of a variety of cells (9 -11). Calcium-binding repeats influence the function of the glycoprotein in cell adhesion and the binding to proteases (12,13).
In cultured endothelial cells, TSP-1 is a major secretory product. The functional properties of TSP-1 in the context of the vascular endothelium are complex. On some substrates TSP-1 promotes endothelial cell adhesion (14). On the other hand, TSP-1 has been shown to reduce focal adhesion and to serve as an antiadhesive matrix component (15,16). TSP-1 promotes migration of endothelial cells but inhibits migration induced by basic fibroblast growth factor. TSP-1 and a 140-kDa TSP fragment act as an inhibitor of angiogenesis in vivo and in vitro (17,18). Furthermore, native TSP-1 and its proteolytic fragments with antiangiogenic activity inhibit proliferation of endothelial cells (19). Because angiogenesis is directly linked to tumor growth and metastasis, it is not surprising that TSP-1 has been implicated in malignant processes (reviewed in Ref. 20).
Essential for the diverse biological effects of TSP-1 is the binding of the molecule to specific cell surface receptors. In the last years the binding to a number of different cell surface molecules has been reported, including CD36, a 105/80-kDa heterodimeric membrane protein, integrins of the ␤3 and ␤1 families, an integrin-associated protein, the low density lipoprotein receptor-related protein, and HS-PG (21)(22)(23)(24)(25)(26)(27).
The interactions with cell surface structures are mediated through different domains of TSP-1. The N-terminal globular domain possesses binding sequences for heparan sulfate proteoglycans (28), a sequence containing SVTCG within type 1 repeats, which interacts with CD36 and a 50-kDa tumor cell receptor (29,30), and a single RGD sequence in the type 3 domain that binds to ␣v␤3 integrins (12,16). Two C-terminal sequences of TSP have substantial affinity to CD47, an integrin-associated protein at the cell surface (24).
In several studies the receptor-like function of heparan sulfate proteoglycans has been demonstrated during the process of binding, uptake, and degradation of TSP-1 by cultured cells (26,28,31,32). Our laboratory recently reported that in porcine endothelial cells, perlecan amounts for at least 90% of membrane-associated HS-PGs and is responsible for the TSP-1-binding activity (33).
Cell surface HS-PGs take part in the membrane binding of various enzymes and growth factors and interact with components of the extracellular matrix (34). The HS-binding sequences for some of the various ligands of HS-PGs are well characterized, and this raises the question, also with respect to biological functions of TSP-1, about the nature of the TSPbinding site in heparan sulfate chains. In the present study, we report the isolation and characterization of distinct heparan sulfate oligosaccharides from endothelial cell-associated HS-PGs that exhibit specific affinity for TSP-1. , heparin, N-acetylated heparin, and unsaturated heparan sulfate disaccharide standards were purchased from Sigma. Heparin-Sepharose, gelatin-Sepharose, Q-Sepharose FF, and DEAE-Sephacel were from Amersham Pharmacia Biotech. Bio-Gel P-6 was obtained from Bio-Rad . All other chemicals used were of analytical grade.

Materials
Preparation of TSP-1 Affinity Columns-Thrombospondin was isolated from human platelets according to previously published procedures (3,35), using chromatography on heparin-Sepharose and gel filtration chromatography on Bio-Gel A 0.5 (Bio-Rad). TSP-1 purity was assessed by SDS-polyacrylamide gel electrophoresis. TSP-1 was coupled to Eurocell ONB-carbonate cellulose beads (Eurochrom, Berlin, Germany) in the presence of N-acetylated heparin (36). Coupling was performed at 4°C overnight according to the manufacturer's instructions at a concentration of 1.5 mg TSP/ml Eurocell ONB beads. The remaining unreacted groups on the beads were blocked with 1 M ethanolamine.
Endothelial Cell Culture-Porcine aortic endothelial cells were isolated and cultured as described previously (37). The cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum. Cells from the second and third passage were used.
In Vivo Labeling of Porcine Endothelial Cells and Isolation of Intact Heparan Sulfate Chains-Porcine endothelial cells were metabolically radiolabeled either with [ 35 S]sulfate (50 Ci/ml) or with [ 3 H]glucosamine (10 Ci/ml) or were double-labeled with both isotopes for up to 48 h. The medium was collected and centrifuged at 2000 ϫ g to remove cellular debris. Urea (6 M) and protease inhibitors were added, and the medium was stored at Ϫ20°C. The cell layer was washed twice with phosphate-buffered saline and extracted with 8 M urea, containing 2% Triton X-100, 1% CHAPS, 100 mM NaCl, 50 mM sodium acetate, pH 6, and a mixture of protease inhibitors (1 mM phenylmethylsulfonyl fluoride, 10 mM benzamidine, 50 mM 6-aminohexanoic acid, 5 mM EDTA) at 4°C for up to 12 h. Unincorporated radioactivity was either removed by chromatography on prepacked PD 10 columns (Amersham Pharmacia Biotech),or the cell layer extract containing HS and CS/DS proteoglycans was applied in the first step to ion-exchange chromatography on a column of Q-Sepharose FF (1-ml bed volume, preequilibrated with 6 M urea, 0.5% Triton X-100, 0.1 M NaCl, and 0.05 M sodium acetate buffer, pH 6.0). The column was washed with 30 bed volumes of this buffer containing 0.4 M NaCl, and the proteoglycans were eluted with 4 M guanidine HCl, 0.5% Triton X-100, 50 mM sodium acetate, pH 6. To obtain free glycosaminoglycan chains by ␤-elimination, the proteoglycan fraction of the cell layer was desalted on a PD-10 column after Q-Sepharose chromatography and thereafter subjected to ␤-elimination for 20 h in 0.1 M NaOH and 1 M NaBH 4 (final concentration) at 37°C.
As shown by control experiments with heparin (Sigma), this procedure did not result in a measurable liberation of inorganic sulfate. After ␤-elimination the mixture was neutralized with 50% acetic acid, diluted with 50 mM NaCl, 20 mM Tris/HCl, pH 7.4, and applied to a 0.5-ml column of DEAE-Sephacel. After a washing step with 300 mM NaCl, 20 mM Tris/HCl, pH 7.4, the free glycosaminoglycan chains were eluted with the same buffer containing 800 mM NaCl. Small amounts of CS/DS were removed from the glycosaminoglycan-containing samples after a desalting step on a PD 10 column by incubation with 0.5 units/ml chondroitin ABC lyase in 20 mM Tris/HCl, 60 mM sodium acetate (pH 8) for 4 h at 37°C. CS/DS oligosaccharides were removed by reapplication to a small DEAE-Sephacel column by a washing step with 0.3 M NaCl, and recovery of the free HS chains was obtained by elution with 0.8 M NaCl.
Preparation of HS Oligosaccharides by Enzymatic Depolymerization-HS chains were depolymerized with heparinase III or heparinase I at a concentration of 20 milliunits/ml in 100 mM sodium acetate, pH 7.0, 2 mM calcium acetate. Heparinase III and heparinase I digestions were performed either in the presence of bovine kidney heparan sulfate (0.25 mg/ml) or of heparin (0.25 mg/ml) as a carrier. Samples were incubated at 37°C for 16 h, and the reaction was stopped by heating at 100°C for 2 min.
Deaminative Scission with Nitrous Acid at Low pH Values-Cleavage of intact HS chains and oligosaccharides with affinity to TSP-1 by HNO 2 treatment at low pH (1.5) was carried out by the method of Shiveley and Conrad (38). Equal volumes of cold 0.5 M H 2 SO 4 and 0.5 M Ba(NO 2 ) 2 solution were mixed on ice and centrifuged at 12,000 ϫ g for 2 min to remove the precipitated barium sulfate. The nitrous acid reagent (50 l) was added to the lyophilized samples, and after 15 min the reaction was stopped by the addition of 1 M Na 2 CO 3 (10 l). To evaluate the occurrence of "anomalous" deamination ring contraction during HNO 2 treatment, control experiments were performed where the oligosaccharides with TSP-1 affinity were subsequently subjected to mild acid treatment (25 mM H 2 SO 4 , 80°C, 30 min) and analyzed further by gel filtration chromatography (39,40).
Preparation of Chemically Modified Heparin Species-Heparin (1 mg/ml) from porcine intestinal mucosa (Sigma) was selectively N-desulfated by acid hydrolysis with 0.04 M HCl for 1 h at 100°C under nitrogen followed by neutralization and dialysis against water. Complete reacetylation was achieved by treating N-desulfated heparin, dissolved in 4.5 M sodium acetate containing 20% (v/v) methanol, with five additions of acetic anhydride in 10-min intervals at ambient temperature, each 1/5 of the original volume. Selective 2-O-desulfation was performed by lyophilization of an aqueous solution of heparin, sodium salt (4 mg/ml H 2 O), brought to pH 12.5 by NaOH addition (41). This treatment reduced the percentage of 2-O-sulfate groups-containing disaccharides from 75% in the mucosal heparin to 3%. For 6-O-desulfation the pyridinium salt of heparin (10 mg/ml pyridine) was treated with N-methyltrimethylsilyl-trifluoroacetamide (200 l/ml heparin solution) for 2 h at 100°C (42). After the addition of an equal volume of 20% methanol, the reaction mixture was dialyzed against water, adjusted to pH 9 -10 with NaOH, and dialyzed again. A reduction of the proportion of 6-O-sulfated disaccharides from 85% in heparin to 17% was thereby achieved.
Affinity Chromatography-For affinity chromatography the intact HS chains or heparinase III-resistant oligosaccharides were dissolved in 20 mM Tris/HCl, 50 mM NaCl, 1 mM CaCl 2 at pH 7.4 (binding buffer) and loaded onto a column of TSP-cellulose beads (1 ml) preequilibrated with this buffer. After washing the column with 10 ml of this buffer, elution of the bound fraction was obtained with a linear gradient of 0.05-1 M NaCl (30 ml) in 20 mM Tris/HCl, pH 7.4. Fractions of 1 ml were collected, and the radioactivity of an aliquot of all fractions was determined. For isolation of TSP-1-binding heparinase III-resistant oligosaccharides, binding assays were performed in Eppendorf tubes by incubating [ 3 H]glucosamine-labeled HS oligosaccharides dissolved in binding buffer (1 ml) with 500 l of immobilized TSP-1. After 2 h of end over end rotation at 4°C, the suspension was transferred to a Pasteur pipette and washed stepwise with binding buffer and buffer containing 150 mM NaCl. Radioactivity found in these steps represent the TSP-1unbound oligosaccharides. TSP-1-binding oligosaccharides were subsequently obtained by elution with 1-ml fractions of 20 mM Tris/HCl, pH 7.4, containing 0.5 M NaCl. The pooled fractions of TSP-1-bound and TSP-1-unbound material were used for gel chromatography and disaccharide analysis. For direct comparison of the disaccharide composition of the TSP-1-bound and -unbound deca-and higher oligosaccharide fraction, the TSP-1-unbound oligosaccharide material was size-fractionated on a Bio-Gel P-6 column (120 ϫ 1 cm), and the octasaccharide and the deca-and higher oligosaccharides were isolated thereof. For inhi-bition of HS/TSP-1-binding by chemically modified heparin species, heparinase III-sensitive oligosaccharides were prepared from [ 35 S]sulfate-labeled endothelial HS and size-fractionated as described above. Deca-and higher saccharides were lyophilized and dissolved in binding buffer at a concentration of about 12 ϫ 10 3 cpm/ml. For binding assays 200 l of immobilized TSP-1 were suspended in 800 l of binding buffer. After centrifugation, the gel was mixed with radioactive ligand (15 l) and 385 l of binding buffer Ϯ 0.2 g inhibitor. After 2 h of end over end rotation at ambient temperature, the suspension was transferred into Pasteur pipettes and washed with 4 ϫ 500 l of binding buffer to determine the amount of unbound material. Bound material was eluted by applying 2 ϫ 500 l of 20 mM Tris/HCl, pH 7.4, containing 2 M NaCl. All assays were done in triplicate.
Gel Filtration Chromatography-Gel filtration chromatography of oligosaccharides after enzymatic or deaminative scission of endothelial HS or HS oligosaccharides with affinity to TSP-1 was performed on a Bio-Gel P-6 column (120 ϫ 1 cm) equilibrated in 0.5 M NH 4 HCO 3 . The column was eluted at a flow rate of 4 ml/h, and 1-ml fractions were collected. An aliquot of each fraction was used for determination of radioactivity by liquid scintillation counting. From the 3 H elution profile of the oligosaccharides the extent of glycosidic bond cleavage could be calculated as described by Malmström et al. (43).
Disaccharide Analysis of HS Oligosaccharides by Strong Anion Exchange-HPLC-Disaccharides were analyzed by strong anion exchange chromatography after complete (more than 90%, as judged by gel filtration) depolymerization by exhaustive digestion with a mixture of heparinases I, II, and III as described (44). The lyases were used at concentrations of 2.5 units/ml (heparinases I and II) and 50 munits/ml (heparinase III). The disaccharides were recovered by chromatography on a Superdex peptide HR 10/30 column (Amersham Pharmacia Biotech) equilibrated and eluted with 0.5 M NH 4 HCO 3 at a flow rate of 0.5 ml/min. The lyophilized digestion products were subjected to strong anion exchange-HPLC on a Phenosphere 5 strong anion exchange 80 column (4.6 ϫ 250 mm; Phenomenex, Hösbach, Germany). After the column had been equilibrated at a flow rate of 0.8 ml with double deionized water and adjusted to pH 3.5 with HCl, samples were injected, and the column was developed with a linear gradient of NaCl (0 -0.75 M in the same mobile phase) over 60 min. Disaccharides were monitored either by UV detection at 232 nm or by liquid scintillation counting of 300-l fractions. Calibration was performed with a set of eight disaccharides kindly provided by Dr. J. T. Gallagher, University of Manchester, UK.

Structure of Cell
Layer-derived HS-In previous studies it was shown that porcine endothelial cells synthesize proteoglycans containing HS and CS/DS glycosaminoglycan chains (33). Whereas CS/DS-containing compounds were found predominantly in the culture medium, one major high molecular weight HS-PG was detected, which upon labeling with radiosulfate, amounted for about 90% of all HS-PG species. It was mainly associated with the cell layer but also partly released into the culture medium. Immunoprecipitation identified the proteoglycan as perlecan, and it was localized in the subendothelial matrix but also closely associated with the cell surface (33). In view of the involvement of HS chains of the cell layer during binding and endocytosis of TSP-1 (31,32), structural features of cell layer-derived HS required for the TSP/HS interactions were investigated.
For characterization of the structural features of the cell layer-derived HS chains, endothelial cells were metabolically labeled with [ 35 S]sulfate or [ 3 H]glucosamine, and the proteoglycan fraction was isolated from the cell extract by ion-exchange chromatography. After ␤-elimination and enzymatic degradation of the small amount (8 -10%) of CS/DS present in the cellular extract, a purified HS fraction was obtained. SDS gel electrophoresis revealed that the intact HS-PG was predominantly found in the stacking gel, whereas the purified HS fraction entered the separation gel and migrated like globular proteins of an apparent molecular mass of about 100 -180 kDa (data not shown).
The cell layer-derived HS chains were characterized by specific cleavage with chemical or enzymatic methods followed by gel filtration. The oligosaccharide fragments obtained either with heparinase I or III or nitrous acid treatment at low pH were separated on Bio-Gel P-6 columns to separate the oligosaccharide fragments.
The major structural requirement of heparinase III for acting on ␣-glycosidic linkages is the presence of glucuronic acid. But the enzyme tolerates both N-sulfated and N-acetylated glucosamine residues at the glycosidic linkage to be split (GlcNR(Ϯ6S)␣1-4GlcA). Resistant oligosaccharides represent highly sulfated regions of the parent HS chains. Based on the [ 3 H]glucosamine-labeling profile heparinase III cleaved 76% of the linkages in the HS chains ( Fig. 1a and Table I). The major cleavage products (65% of total radioactivity) were mono-or unsulfated disaccharides as identified by strong anion exchange-HPLC of the disaccharide fraction (data not shown). Only a small proportion consisted of tetrasaccharides, whereas hexa-and octasaccharides represented the main fraction of oligosaccharides (about 20% of the total 3 H radioactivity). Heparinase I acts on highly sulfated and epimerized sequences of HS and cleaves ␣-glycosidic linkages of N-sulfonylglucosamine if the adjacent uronic acid is iduronic acid 2-sulfate HS fragments were applied to a Bio-Gel P-6 column (120 ϫ 1 cm) and eluted with 0.5 M NH 4 HCO 3 at a flow rate of 4 ml/h. The inset shows an expanded scale of the elution profile to demonstrate low molecular weight degradation products. Fractions of 1 ml were collected and analyzed for radioactivity. Void (V 0 ) and total volume (V t ) were determined by using blue dextran and DNP-alanine, respectively.
(GlcNS(Ϯ6S)␣1-4IdoA(2S)). Only a small proportion of the linkages in the HS chains was susceptible to the enzyme (12, 5%). About 3% of di-and tetrasaccharides were produced. The majority (94%) of the oligosaccharide fraction was not separated by gel chromatography on Bio-Gel P-6 and is therefore dp 16 and higher (Fig. 1b). Treatment of HS chains with nitrous acid at low pH leads to deaminative cleavage of N-sulfonylglucosaminyl residues (GlcNS(Ϯ6S)␣1-4 hexuronic acid (Ϯ2S)). This deaminative scission caused a limited fragmentation of HS chains and yielded different oligosaccharides up to higher than dp 16, with tetrasaccharides as the main fraction (Fig. 1c). From the breakdown products caused by nitrous acid an Nsulfate content of 35% can be calculated. Only marginal differences in the depolymerization patterns were observed between HS chains obtained from cell-derived or medium-released HS-PG (data not shown). The cleavage profiles produced by the different depolymerization conditions provide evidence that endothelial cell-derived HS possesses a domain structure with sulfate-rich and sulfate-poor clusters as reported for other HS species. Moreover, this result allows us to predict some information on the distribution of the susceptible linkages as summarized in Table I. In the case of heparinase I treatment the low yield of di-and tetrasaccharides reflects the fact that GlcNS(Ϯ6S)-IdoA(2S) disaccharide structures appear spaced from each other, whereas the results of the heparinase III treatment demonstrate that the susceptible linkages (GlcNR(Ϯ6S)-GlcA) are contiguously arranged.
TSP-1 Affinity Chromatography of Native and Partially Depolymerized HS-Purified native HS chains, metabolically labeled with [ 35 S]sulfate, were applied to an affinity column containing human TSP-1 isolated from platelets. Over 90% of the HS chains bound to the column. Bound material was eluted with a continuous salt gradient and required 0.32 M NaCl for elution of peak material (Fig. 2a). To identify the major structural determinants of HS necessary for binding to TSP-1, the influence of partially depolymerization with heparinase III was investigated. Approximately 85-90% of the resulting fragments did not bind to the column. Bound material was recovered from the column by elution with a linear salt gradient in a narrow peak at only a slightly lower salt concentration than the intact HS chains (0.28 M NaCl, Fig. 2b). No binding was observed to a bovine serum albumin affinity column (data not shown).
To isolate and to characterize in greater detail the HS oligosaccharides with TSP-1-binding activity, unbound and bound saccharides were size fractionated on a Bio-Gel P-6 column. The results shown in Fig. 3 reveal that the TSP-1-binding fraction consisted only of deca-and higher oligosaccharides. The elution profile of the unbound fraction demonstrated that in addition to oligosaccharides of dp 2-8 a distinct proportion of deca-and higher oligosaccharides was also present in this fraction. Rechromatography of these fractions indicated that they were indeed unable to interact with TSP-1 and excluded the possibility of an insufficient capacity of the column to retain TSP-1-binding fragments. Thus, only 45-50% of the deca-and higher oligomers possessed TSP-1-binding affinity. FIG. 2. TSP-1 affinity chromatography of endothelial cell-derived intact or partially degraded HS. Intact 35 S-labeled HS chains obtained from cell layer-derived HS-PG by ␤-elimination and HS chains after heparinase III treatment were applied to a TSP-1 affinity column (1 ml) in 0.05 M NaCl. After a washing step with binding buffer, the bound material was eluted with a linear gradient of 0.05-1 M NaCl. Elution profile of bound HS chains is shown in a; elution of heparinase III treated HS chains is shown in b.

FIG. 3. Bio-Gel P-6 elution profile of HS oligosaccharides with or without TSP-1-binding affinity. [ 3 H]
Glucosamine-labeled cellderived HS chains were digested with heparinase III, and affinity chromatography with TSP1 was performed as described under "Experimental Procedures." Appropriate amounts of the pooled unbound (E) and TSP-1-bound (f) fraction were analyzed by chromatography on a Bio-Gel P-6 column.

Disaccharide Composition of Intact HS Chains and Oligosaccharides with or without Affinity to TSP-1-Disaccharide analysis of intact HS chains and of oligosaccharide fractions with or
without TSP-1 affinity was performed to define structural requirements necessary for the interaction with TSP-1. Exhaustive depolymerization was performed by digestion with a combination of heparinases I, II, and III. The final disaccharide fraction was recovered by gel filtration on a Superdex Peptide HR column, followed by strong anion exchange-HPLC chromatography. The intact HS chains contain approximately 52% unsulfated disaccharide units, 35% monosulfated disaccharides, and about 9% di-and trisulfated residues, representing Nand O-sulfated structures (Table II). In the TSP-1-binding oligosaccharide fraction the proportion of unsulfated disaccharides is reduced to less than 20%, whereas the content of monosulfated disaccharides remained almost constant. A large increase is observed in the level of di-and trisulfated disaccharides, with 41% in the TSP-binding oligosaccharides compared with 9% in intact HS chains. Comparing the disaccharide composition of nonbinding octasaccharide and the unbound decaand higher oligosaccharide fraction indicates that these two nonbinding compounds possess a very similar disaccharide composition. Compared with intact HS they contain a reduced proportion of unsulfated disaccharide units but an increase in the trisulfated disaccharide content, which was to be expected from the specificity of heparinase III. In relation to the TSP-1binding oligosaccharides the most obvious difference exists in the content of 2-O-and 6-O-sulfates. In both nonbinding compounds a 35-45% reduced level of disaccharide units with 2-Oor 6-O-sulfate groups was observed (Table II).
Structural Arrangement of Disaccharides in TSP-1-binding Oligosaccharides-To determine the arrangement of the constituent disaccharide units in TSP-1-binding sequences further depolymerization of the oligosaccharides with heparinase I and deaminative scission with nitrous acid at low pH were performed. Gel filtration data are shown in Fig. 4. Over 90% of the TSP-binding oligosaccharide fraction (dp10-dp14) were susceptible to heparinase I degradation (Fig. 4a). Heparinase I treatment cleaved the oligosaccharides mainly to hexasaccharides (30% of total radioactivity), tetrasaccharides (28%), and disaccharides (21%). Because heparinase I attacks specifically disaccharide units with 2-O-sulfated iduronic acid, the obtained scission pattern confirms the result of the disaccharide analysis (33% 2-O-sulfated hexuronic acid containing residues; Table  II) and indicates that the 2-O-sulfated hexuronic acid residues represent iduronic acid. The resistant fragments, mainly hexaand tetrasaccharides, represent sequences containing nonsulfated GlcA-or L-iduronic acid residues. Treatment of the oligosaccharides with nitrous acid at low pH produced a high proportion of disaccharides (71% of total radioactivity) but also tetrasaccharides (27%, Fig. 4b). None of the terasaccharides were formed by "anomalous ring contraction" as judged by subjecting the deamination products to mild acid hydrolysis (data not shown). The occurrence of tetrasaccharides indicates the existence of internal linkages within the TSP-1-binding oligosaccharides containing N-acetylated glucosamine residues. Thus, the constituent 2-O-and 6-O-sulfated disaccharide units seem to be arranged in mixed sequences within the TSPbinding oligosaccharides.  in a) with affinity to TSP-1 were obtained as described in Fig. 3 and were depolymerized by either (a) heparinase I (dotted line) or (b) treatment with HNO 2 at low pH. The occurring oligosaccharides were separated on a Bio-Gel P-6 column in 0.5 M NH 4 HCO 3 at a flow rate of 4 ml/h. The disaccharides (dp 2) were partially resolved according to their degree of sulfation.

TABLE II
Disaccharide composition of endothelial cell-derived HS and of oligosaccharide fractions with or without binding affinity to TSP-1 Analysis of the constituent disaccharides of endothelial cell-derived HS and its TSP-1-binding oligosaccharides. The oligosaccharides were prepared by heparinase III depolymerization of intact HS chains and after TSP-1 affinity chromatography dp 8 oligosaccharides and unbound dp 10 and higher oligosaccharides were recovered from the unbound oligosaccharide fraction by size fractioning on a Bio-Gel P-6 column. Complete degradation to disaccharides was obtained with a combination of heparinase I, II, and III, followed by strong anion exchange HPLC as described under "Experimental Procedures."

Disaccharide
Intact HS dp 8 oligosaccharides Unbound dp 10 and higher Bound dp 10 and higher Inhibition of HS/TSP-1 Binding with Chemically Modified Heparin Species-The analysis of the disaccharide composition of TSP-1-binding oligosaccharides does not allow to discriminate between essential structural features and features that can be tolerated within a binding oligosaccharide. To address this problem competition experiments were performed in which low concentrations of chemically modified heparin species (0.5 g/ml) were used for competing with TSP-1-binding oligosaccharides (Fig. 5). At such low concentration of inhibitor heparin was able to reduce the binding to about 50%, whereas Ndesulfation and N-reacetylation of heparin and 6-O-desulfated heparin were completely inactive. Thus, 6-O-sulfation and Nsulfate groups are both necessary for binding activity of HS oligosaccharides to TSP-1. Some inhibition was archived by 2-O-desulfated heparin (17%), suggesting that the presence of 2-O-sulfated iduronic acid is a supportive but not necessarily an essential element.

DISCUSSION
The influence of TSP-1 in diverse and complex biological processes is based on the interaction of the extracellular matrix protein with structural elements at the cell surface. Because of the multiple domain structure, binding sites for integrins, nonintegrin receptors, and heparan sulfate proteoglycans exist in the TSP-1 molecule. In previous studies the involvement of HS during binding and endocytosis of TSP-1 in porcine endothelial cells was reported (31). Loss of HS at the cell surface or inhibition of sulfation of glycosaminoglycan chains by chlorate treatment strongly inhibited the binding and uptake process of TSP-1 (32).
There are several cell-associated heparan sulfate proteoglycan members of the syndecan and glypican family (34). Perlecan is considered as a typical basement membrane component, but we have recently shown that in porcine endothelial cells over 90% of the synthesized HS structures are associated with perlecan (33). Because nearly all HS chains prepared from the cell layer extract are able to interact with TSP-1 it is safe to conclude that the main HS component of endothelial cells with TSP-1-binding properties is represented by perlecan.
TSP-1 interacts with heparin/HS mainly through the Nterminal binding domain, which contains three regions of high affinity for heparin (45). In addition there exists a second binding site in the stalk-like part of the protein and possibly additional sites (46). Up to now no information is available concerning the structural features of the HS chains for the interaction with TSP-1.
The results of the enzymatic and chemical scission demonstrate that endothelial cell-derived HS chains possess the typical structural heterogeneity with domains of highly sulfated disaccharides separated by extended sequences containing predominantly N-acetylated sequences of low sulfation. Detailed analysis revealed a N-sulfation level of 33% and an O-sulfation level of 24%, which leads to a total sulfate content of 0.6 sulfate groups/disaccharide. A very similar composition was reported for human endothelial cells derived from umbilical veins (47) and HS isolated from bovine arterial tissue (48), whereas in other HS preparations of different tissues (49) the composition clearly differed and reflects the structural variability of HS biosynthesis.
Our binding experiments with intact HS chains and heparinase III-treated HS chains allowed us to obtain some information of the HS domains necessary for TSP-1 binding. Over 90% of intact HS chains bound on immobilized TSP-1 with a moderate affinity, as measured by the NaCl concentration required for dissociation. In contrast, only a small portion of the heparinase III-resistant oligosaccharides possessed binding affinity to a TSP-1 column. Size fractioning on Bio-Gel P-6 revealed that the minimal HS fragment capable of binding to TSP-1 are deca-and higher oligosaccharides, whereas oligosaccharides of dp2 to dp8 in size and a distinct proportion of the higher oligosaccharide fraction did not express binding properties.
Compositional analysis of TSP-1-bound and -unbound oligosaccharides showed that the binding oligosaccharides did not contain unique components, but an overall increase in the degree of sulfation. Binding properties seem to be connected with an increase in the content of di-and trisulfated disaccharide units and especially with the occurrence of Ido(2S)-GlcNS(6S). This disaccharide unit was enriched 10-fold in the binding oligosaccharide fraction compared with the native HS chains, and the binding oligosaccharides possessed a 2-fold higher level of this disaccharide compared with the unbound deca-and higher oligosaccharide fraction. Considering the disaccharide analysis of the nonbinding octasaccharide fraction allows us to conclude that the decasaccharide is the minimal length necessary to obtain an optimal structure with binding affinity.
The importance of the various different sulfate groups for the binding of HS oligosaccharides to TSP-1 was evaluated in competition experiments where chemically modified heparin species were used to compete with the active HS fraction. Removal of N-or 6-O-sulfate groups but to a minor degree removal of 2-O-sulfate from heparin was found to impede the interaction with TSP-1, suggesting that 6-O-sulfation and N-sulfation are necessary for binding activity, whereas 2-O-sulfate is supportive but not essential for binding.
As the TSP-binding fraction contained several differently sized oligosaccharides, a definitive structural sequence from the disaccharide units cannot be provided. However, from the analytical data of the disaccharide composition and the susceptibility to deaminative scission and heparinase I treatment some structural features of the binding sequences were obtained. The binding oligosaccharides contain an unsulfated disaccharide unit, probably in most of the cases positioned at the reducing end. In addition, some N-acetylglucosamine con- taining disaccharide units must be placed within the oligosaccharide sequences because treatment with nitrous acid generated not only disaccharides but also tetrasaccharides. The heparinase I sensitivity of the TSP-1-binding oligosaccharides supports the notion that 2-O-sulfated hexuronic acid residues represent IdoA(2S). However, the small fraction of disaccharides obtained by heparinase I scission is an indication that the IdoA(2S) containing disaccharide units are not arranged in clusters within the binding sequences.
Different structural protein-binding domains within the HS chains seem to be designed to interact specifically with a variety of unrelated proteins. For most HS-binding proteins, the binding of HS chains occurs through a single highly sulfated domain of variable size (reviewed in Ref. 50). In a few cases, such as platelet factor 4, two binding sequences separated by a N-acetylglucosamine-rich region are involved in the binding process (51). Binding specificity depends either on the occurrence of unusual structural features or, as found in most cases, binding selectivity is obtained through the specific arrangement of the commonly occurring disaccharide units. A unique structural composition was observed for the HS-binding sequence of antithrombin, where the binding pentasaccharide contains a 3-O-sulfated N-sulfated glucosamine unit (reviewed in Ref. 52). For basic fibroblast growth factor the presence of N-sulfated glucosamine and 2-O-sulfated iduronate is essential for the binding properties of the high affinity tetradecasaccharide (53). By contrast GlcN-6-O-sulfate groups are critical for the HS domain interacting with fibroblast growth factor 1 or fibroblast growth factor 4 (54). This structural requirement was also observed for the interaction with lipoprotein lipase (36) and hepatocyte growth factor (55). In all these cases the binding oligosaccharide contained the trisulfated disaccharide unit IdoA(2S)-GlcNS(6S). From our disaccharide analysis and the competition experiments we can assume that also for TSP-1 the presence of 6-O-sulfate groups and the occurrence of trisulfated disaccharides are structural features required for binding.
An interesting question emerging from our results is whether a modulation of HS structures influences the functional role of TSP-1. There are several reports that describe changes in the composition of HS chains during the process of development, differentiation, and under pathological conditions. For example, during enterocytic differentiation the fine structure and sulfation of perlecan changes, whereas no differences in the protein core were observed (56). The increase in sulfation was primarily affecting 6-O-sulfation, leading to an increase of trisulfated IdoA(2S)-GlcNS(6S) disaccharide units. Therefore, one might propose that similar modifications in the HS structure should have a regulatory and functional influence on TSP-1.