Characterization of heparan sulfate oligosaccharides that bind to hepatocyte growth factor.

Proteoglycans from rat liver had the ability to bind hepatocyte growth factor (HGF). Digestion of the proteoglycans with heparitinase resulted in the complete loss of the activity, while the digestion with chondroitinase ABC had no effect. Heparan sulfate (HS)-conjugated gel also bound HGF, and the binding was competitively inhibited by heparin and bovine liver HS, but not by Engelbreth-Holm-Swarm sarcoma HS, pig aorta HS, or other glycosaminoglycans, suggesting the specific structural domain in HS for the binding of HGF. Among limited digests with heparitinase I of bovine liver HS, octasaccharide is the minimal size to bind HGF. Comparison of the disaccharide unit compositions revealed a marked difference in IdoA(2SO4)-GlcNSO3(6SO4) unit between the bound and unbound octasaccharides. The contents of this disaccharide unit were calculated to be 2 mol/mol for the bound octasaccharide but 1 mol/mol for the unbound one. Considering both the substrate specificity and properties of heparitinase I, the above results suggest that the bound octasaccharide should contain two units of IdoA(2SO4)-GlcNSO3(6SO4) contiguously or alternately in the vicinity of the reducing end. The bound decasaccharide was more than 20 times as active as the unbound one with regard to the ability to release HGF bound to rat liver HS proteoglycan. The ability was comparable to the one-fourth of that of heparin.

HS 1 has been shown to have activities to bind to various molecules (1). Of those, heparin-binding growth factors are particularly important, considering the physiological significance of potential ligands of HS (1). bFGF is such a typical molecule and was detected as a complex with HSPG in the extracellular matrix such as basement membranes of the kid-ney glomerulus (2). In addition, the low affinity receptor for bFGF on the cell surface was identified to be a cell-surface HSPG (3,4). Recent studies (5)(6)(7)(8) have shown that the binding of bFGF to the cell-surface and/or extracellular matrix HSPG is essential for the interaction of bFGF with its high affinity receptor. Heparin or HS may also be involved in protecting bFGF from protease digestion or heat/acid inactivation (9). It is of note here that the binding of bFGF to HS requires the domain structure composed of a cluster of IdoA(2S)-GlcNS units (10 -13).
HGF was identified initially as a mitogen for hepatocytes (14,15). Subsequently, HGF was found to be identical not only with a scatter factor (16) but also with a tumor cytotoxic factor (17). Thus, HGF promotes the dissociation of epithelial cells and vascular endothelial cells in vitro and stimulates angiogenesis in vivo (18,19). In addition, HGF is considered to be a unique pleiotropic factor that acts as a mitogen, a tumor suppresser, a motogen, and a morphogen. Further, HGF may mediate epithelial and mesenchymal interactions during embryogenesis, organ repair, and neoplasia (20).
HGF is known to have the ability to bind to heparin, and there are two classes of receptors for HGF with the different affinities (16,(21)(22)(23)(24). The high affinity receptor (K d 4.6 pM) (21) on rat hepatocytes was identified as the c-met proto-oncogene product, a transmembrane tyrosine kinase that is expressed predominantly on epithelial cells (16,22,25). The low affinity receptor (K d 275 pM) (21) was found to be a HSPG at the cell surface. Possible functional consequences after binding are as follows; stabilization of HGF (26,27), induction of conformational changes to fit HGF to the high affinity receptor (28,29), or, conversely, blocking of the biological activity due to ligand sequestering (30). HSPGs in rat liver are identified as perlecan, syndecan, and fibroglycan (31)(32)(33)(34). However, it remains to be determined which is likely for a low affinity receptor. A mutant HGF without the affinity for heparin showed neither the affinity for c-met protein nor the biological activity (35)(36)(37)(38)(39). However, exogenous addition of heparin reduced the interaction of HGF with c-met protein (23,28) and, consequently, reduced the mitogenic (40,41) and motogenic (42) responses of cells to HGF. This was explained by the observation that a HGF-exogenous heparin complex could not be bound to c-met protein (28), which suggests, interestingly, that exogenous heparin does not function as the cell-surface HSPG. Certain molecular structures and/or spatial localization of endogenous HSPG may be important in regulating the binding of HGF to c-met protein (28). Therefore, the significance of interaction between cell-surface HSPG and HGF may be the same as that of bFGF, but the mechanism appears to be different and complex. To understand it, the precise analysis for the interaction between HSPG and HGF is needed.
In this study, we fractionated HS oligosaccharides prepared from the HS digested with heparitinase I, in accordance with the different affinities to HGF, and characterized a possible * This work was supported in part by grants-in-aid from the Ministry of Education, Culture and Science, Japan, Special Coordination Founds of the Science and Technology Agency of the Japanese Government, and a special research fund from Seikagaku Corporation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

EXPERIMENTAL PROCEDURES
Materials-Heparin was purchased from Sigma. HSs from pig aorta, pig liver, bovine liver, and EHS sarcoma were gifts of K. Yoshida  Preparation of Proteoglycans from Rat Liver-Liver was quickly excised. Livers from five rats (total wet weight, approximately 65 g) were rinsed with PBS, cut into small pieces, and then homogenized in the 4 M guanidine HCl extraction solution containing 50 mM sodium acetate, 10 mM EDTA, 10 mM N-ethylmaleimide, 1 mM phenylmethanesulfonyl fluoride, 0.1 M 6-aminohexanoic acid, 20 mM benzamidine HCl, 2% (v/v) Triton X-100. The homogenate (approximately 360 ml) was stirred at 4°C for 48 h. Insoluble residues were removed by centrifugation at 12,000 ϫ g for 30 min at 4°C. The supernatant was recovered. Twenty ml of the supernatant solution were diluted with 19 volumes of 7 M urea buffer (7 M urea, 20 mM Tris-HCl, pH 7.2, 10 mM EDTA, 5 mM Nethylmaleimide, 0.5 mM phenylmethanesulfonyl fluoride, 2% (v/v) Triton X-100), and was applied to DEAE-Sephacel (2 ml) equilibrated with 7 M urea buffer at 4°C. The column was washed with 10 ml of 0.2 M NaCl in 7 M urea buffer. Proteoglycans were eluted with 6 ml (3 volumes of the column) of 2 M NaCl in 7 M urea buffer. For the complete separation, the elute was diluted with 9 volumes of 7 M urea buffer, then applied to the second DEAE-Sephacel (1 ml). The column was washed twice with 5 ml of 0.2 M NaCl in 7 M urea buffer. A proteoglycan fraction was eluted with 3 ml of 2 M NaCl in 7 M urea buffer, precipitated with 3 volumes of 95% (v/v) ethanol containing 1.3% (w/v) potassium acetate. The precipitate was dissolved in 300 l of H 2 O.
Preparation of Digoxigenin-conjugated HGF and [ 125 I]HGF-Digoxigenin-conjugated or 125 I-labeled HGF was prepared according to the method recommended by the manufacturer. Briefly, 10 g of HGF in 200 l of 0.2 M phosphate buffer, pH 8.5. were added into N-acetylated heparan sulfate and then mixed with 8.75 nmol of digoxigenin in dimethyl sulfoxide followed by 2 h of incubation at room temperature. The HGF solution was applied to 0.5 ml of heparin-Sepharose gel equilibrated with phosphate-buffered saline (PBS; 0.1 M sodium phosphate, 1.37 M NaCl, 2.7 mM KCl, pH 7.2) containing 0.02% (v/v) Triton X-100 and 1 mg/ml BSA (solution A). Heparin-Sepharose gels were washed with 5 ml of solution A. Digoxigenin-conjugated HGF was then eluted with 2.5 ml of 2 M NaCl in solution A.
IODO-BEADS (Pierce) were kept in 100 l of 0.1 M sodium phosphate containing 0.5 mCi of [ 125 I]NaI at room temperature for 5 min. Then 3 g of HGF were added, and the suspension was kept for 10 min at room temperature. 125 I-Labeled HGF was desalted using a Sephadex G-25 column (0.9 cm ϫ 3.9 cm). Specific radioactivity of 125 I-HGF was 2.5 ϳ 5.7 ϫ 10 4 dpm/ng.
Binding Assay of Digoxigenin-HGF to PG from Rat Liver-15 l of the PG fraction (equivalent to 0.2 g of rat liver) was subjected to 5% SDS-PAGE under nonreducing conditions, electrotransferred to a poly-(vinylidene fluoride) membrane (ProBlott) (Applied Biosystems Japan) at 10 V and 4°C overnight. Each membrane was blocked with a blocking solution (Boehringer Mannheim Biochemica) at room temperature for 30 min and then digested with a mixture of 10, 1, and 10 milliunits/ml heparitinase I, II, and heparinase (the HSase mixture) plus or minus chondroitinase ABC in 50 mM Tris-HCl, pH 7.2, 1 mM CaCl 2 , 0.5 mg/ml BSA in the presence of protease inhibitors excepted no addition of EDTA as described previously (43). Some membranes were digested only with 33 milliunits/ml chondroitinase ABC in 0.5 mM Tris-HCl, pH 8.0, 0.5 mg/ml BSA at 37°C for 1 h. Membranes were washed three times with TBS (50 mM Tris HCl, pH 7.5, 0.15 M NaCl), and then subjected to HGF binding in the solution containing 0.2 g/ml digoxi-genin-HGF, 0.2 mg/ml chondroitin 4-sulfate, 0.9 mM CaCl 2 . After 1 h at room temperature, unbound digoxigenin-HGF was removed by washes with TBS as described above. Membranes were then treated with antidigoxigenin-AP, Fab fragments (1:500 dilution) for 1 h. Unbound antibodies were washed out as described above, and membranes were soaked in 5-bromo-4-chloro-3-indolyl phosphate, 4-toluidine salt (1:200 dilution) and nitro blue tetrazolium (1:260 dilution).
Preparation of HS-conjugated Sepharose Gel-HS-Sepharose gel was prepared by the method reported previously with a minor modification (44). 3-Amino-2-hydroxypropyl-derivatized Sepharose gel was prepared from epoxy-activated Sepharose 6B gel. A portion (1 g) of amino-Sepharose gels thus obtained was suspended in 1 ml of 0.2 M phosphate buffer, pH 7.2, and conjugated with 30 mg of HS (pig liver) by adding 3 mg of NaBH 3 CN. The suspension was kept at room temperature for 48 h with a gentle shaking. The gel was washed several times with PBS. The amount of immobilized HS was 2.4 mg/ml of gel. The gels were, then, suspended in PBS(ϩ) containing 20 mg/ml BSA, and gently stirred for 1 h at room temperature to block nonspecific binding sites. After an extensive wash with PBS(ϩ), the gels were suspended in PBS(ϩ) containing 0.02% NaN 3 to give a 25% (w/v) suspension and stored at 4°C until use.
Competitive Inhibition Assay of [ 125 I]HGF Binding to Immobilized HS with GAGs-The binding reaction was performed in 100 l of solution containing 1.25% (w/v) HS-conjugated gel, 1 ϫ 10 4 dpm of 125 I-HGF, 0.1ϳ100 g/ml GAG, and 1 mg/ml BSA. After 1 h of incubation at 4°C with gentle agitation, the mixture was diluted with 3 volumes of PBS(ϩ) and centrifuged (630 ϫ g, 3 min) in a microcentrifuge tube with a membrane filter (UFC30HV00; Millipore, Bedford, MA). The gel on the membrane was washed thoroughly with PBS(ϩ), and the radioactivity bound to the gel was determined in a ␥-radiation counter. Nonspecific binding was determined as the radioactivity bound to the gel in the presence of 100 g/ml heparin.
Fractionation of HS-Bovine liver HS was fractionated by Dowex 1 column chromatography. The fraction eluted with 0.5 ϳ 1.25 M NaCl was termed bovine liver HS fraction 1. The fraction eluted with 1.25 ϳ 1.75 M NaCl was further fractionated by DEAE-Sephacel column chromatography. The subfractions eluted with 0.42 ϳ 0.48 M and 0.48 ϳ 0.62 M NaCl in 50 mM Tris-HCl, pH 7.2, were termed bovine liver HS fractions 2 and 3, respectively. bFGF-bound HS and unbound HS were prepared from EHS mouse sarcoma HS by the method reported previously (10).
Preparation of HGF-conjugated Sepharose Gel-HGF-conjugated Sepharose gel was prepared by the reported method (10). HGF (1 mg) was coupled to 1.8 ml of CNBr-activated Sepharose 4B gel according to the method recommended by the manufacture. N-Acetylated heparin (10 mg) was added to the coupling reaction mixture to protect the heparan sulfate-binding sites in HGF.
HGF Affinity Chromatography of HS and Heparin Oligosaccharides-About 4 nmol of the HS or heparin oligosaccharide fractions containing 1 ϳ 2 ϫ 10 5 dpm of 3 H-label were dissolved in 1 ml of 10 mM Tris-HCl, pH 7.2, 0.15 M NaCl, 0.9 mM CaCl 2 , 0.2 mg/ml chondroitin 4-sulfate (solution B), and applied to a syringe column of HGF-Sepharose (1 ml) equilibrated with solution B at 4°C. Chondroitin 4-sulfate was included in solution B to prevent the nonspecific binding. The column was shaken gently for 1 h, then washed with 10 ml of solution B, and eluted with 3 ml of 2 M NaCl in10 mM Tris-HCl, pH 7.2. The radioactivity of the eluate was detected in a liquid scintillation counter.
Mono Q Column Chromatography-The eluate from HGF affinity column was subjected to gel chromatography on Sephadex G-50 (1.2 cm ϫ 120 cm) to remove coexisting chondroitin 4-sulfate. The oligosaccharides were recovered from the retarded fractions and then desalted using a fast desalting column (Pharmacia). The fractions were applied to a mono Q column (Pharmacia). The chromatography was performed by a linear gradient elution from 0 to 2.0 M NaCl in 50 mM Tris-HCl, pH 7.2.
Composition Analysis of HS and Its Oligosaccharides-About 1 g of HS or HS oligosaccharides was digested with a mixture of 1 milliunit of heparitinase I, 0.1 milliunit of heparitinase II, and 1 milliunit of heparinase in 50 l of 50 mM Tris-HCl, pH 7.2, 1 mM CaCl 2 , 5 g of BSA at 37°C for 1 h. Unsaturated disaccharide products were analyzed by HPLC using a polyamine-bound silica PAMN column (YMC). The elution was performed with a linear gradient from 40 to 550 mM KH 2 PO 4 and with a subsequent wash with 550 mM KH 2 PO 4 at a flow rate of 1.2 ml/min at 40°C. The elution was monitored by uv absorption at 232 nm. Each peak was identified by its retention time which was standardized with authentic unsaturated disaccharides as described previously (46).
Degradation of about 1 g of HS oligosaccharides with nitrous acid at pH 1.5 and reduction of degradation products with [ 3 H]NaBH 4 were carried out as described by Shively and Conrad (45). The products were desalted using Fast desalting columns. The fractions containing disaccharides were collected and analyzed by HPLC on a Partisil-10 SAX column (Whatman, Clifton, NJ) as described by Bienkowski and Conrad (47). The elution was monitored by measuring the radioactivity in a liquid scintillation counter.
HGF-releasing Activity of HS Oligosaccharides and Heparin-The releasing activity was measured by ELISA by the method recommended by the manufacture with a minor modification. A 96-well Nunc-Immuno Plate MaxiSorp (A/S Nunc, Roskilde, Denmark) was coated with 0.1 nmol (as hexuronic acid) of rat liver proteoglycans overnight at 4°C. Wells were washed three times with 200 l of PBS and then blocked with 200 l of PBS containing 10 mg/ml BSA (solution C) for 1 h at 37°C with a gentle shaking. Wells were washed three times with 200 l of PBS. Then 100 l of the solution C containing 0.2 g/ml digoxigenin-HGF, 0.2 mg/ml chondroitin 4-sulfate, 0.9 mM CaCl 2 was added into each well. After 1 h at room temperature, unbound digoxigenin-HGF was removed by washes as described above. Then, 100 l of PBS containing 1 ng to 10 g of heparin or 1 pmol to 1 nmol as hexuronic acid of HS oligosaccharides were added into wells. After 1 h at room temperature, wells were washed as above, and then alkaline phosphataseconjugated Fab fragments of anti-digoxigenin antibody (1:500 dilution) were added. After 1 h at room temperature, unbound Fab fragments were removed by washing, and the alkaline phosphatase substrate (1 mg/ml of pNPP in 1 mol/liter diethanolamine, pH 9.8, containing 0.5 mmol/liter) was added into each well. The enzyme activity in each well was measured by a MTP-100 microplate reader (Corona Electric Co., Ibaragi, Japan).

RESULTS
Binding of HGF to Rat Liver Proteoglycans-PG preparations from whole rat liver were subjected to SDS-PAGE. PGs separated on the gel were transferred to a membrane for the blot analysis of HGF binding using digoxigenin-conjugated HGF. At least three species of PGs showed the affinity for HGF, of which molecular masses were 220, 180, and 120 kDa (Fig. 1,  lane 1). When these PGs on the membrane were digested with a mixture of heparitinases I and II and heparinase (the HSase mixture) before exposing to HGF, none of them could bind HGF (Fig. 1, lane 2). However, the digestion of the PGs with chondroitinase ABC had no effect on the HGF binding (Fig. 1, lane  3). The results, therefore, suggested that HGF appeared to bind to proteoglycans only with HS chains, but not with chondroitin sulfate or dermatan sulfate chains.
HGF Binding Activities of Various Glycosaminoglycans-The activities of various GAGs were assessed by their capacities to inhibit 125 I-HGF binding to pig liver HS-conjugated Sepharose gel as described under "Experimental Procedures." Table I shows IC 50 values of various GAGs which were their concentrations to inhibit 50% the total radioactivity of 125 I-HGF bound to the HS-conjugated gels. Heparin exhibited the highest inhibition activity (IC 50 ϭ 0.15 g/ml). Bovine liver HS fraction 3 (IC 50 ϭ 0.75 g/ml) was approximate in the inhibition activity to heparin. Bovine liver HS fraction 2 exhibited inhibition activity less than that of heparin (IC 50 ϭ 5.4 g/ml). When bovine liver HS fractions 2 and 3 were digested with the HSase mixture (see "Experimental Procedures") prior to the addition, the inhibition activity completely disappeared (data not shown). This result also supported the fact that HS chains bound HGF. However, bovine liver HS fraction 1 and pig liver HS exhibited weak inhibition activity (IC 50 ϭ 45 and 38 g/ml, respectively), and neither pig aorta HS nor EHS tumor HS showed inhibition activity. The results suggest that HSs vary depending on their differences in species and tissue origins with respect to their affinity for HGF. None of other GAGs tested exhibited inhibition activity.
Heparin and bovine liver HS fraction 3 that showed the high inhibition activity are higher in the sulfation degree (2.59/ disaccharide and 2.12/disaccharide, respectively) than other GAGs, suggesting the involvement of the negative charge in the activity. However, bovine liver HS fraction 2 with the significant inhibition activity is apparently lower in the sulfation degree than chondroitin sulfate E or chemically sulfated dermatan sulfate that showed no inhibition activity (1.21/disaccharide for bovine liver HS fraction 2, compared with 1.43/ disaccharide for chondroitin sulfate E or 1.31/disaccharide for chemically sulfated dermatan sulfate). Taken together, it is likely that binding of HGF to HS/heparin is not simply due to an electrostatic interaction, but may depend on some unique structural units in HS. Indeed, because HS from EHS tumor, which had such units for bFGF-binding (IdoA(2SO 4 )-GlcNSO 3rich domain) (10), had no inhibition activity, binding of HGF to HS may require structural units of HS distinct from the ones for bFGF binding.
Fractionation of HGF-bound HS Oligosaccharides-To determine HGF binding structures in HS, we first prepared HS oligosaccharide with various HGF binding activities from bovine liver HS fraction 2. Limited digestion of the fraction with heparitinase I was performed, which attacks preferentially glucosaminic linkages to nonsulfated hexuronic acid residues in HS. Oligosaccharide products were reduced with [ 3 H]NaBH 4 , and 3 H-labeled HS oligosaccharides thus obtained were subjected to a molecular size fractionation by Sephadex G-50 column chromatography (Fig. 2). Fractions of HS oligo- saccharides with different sizes were rechromatographed on the same column for further purification and designated as shown in Fig. 2. The apparent molecular weights of HS oligosaccharide fractions calculated from their relative elution positions to those of standard oligosaccharides were as follows; HS-I, 800; HS-II, 1300; HS-III, 1700; HS-IV, 2100; HS-V, 2600; and HS-VI, 3000.
Each 3 H-labeled HS oligosaccharide fraction (4 nmol) was applied to a column of HGF-conjugated Sepharose equilibrated with solution B (10 mM Tris-HCl, pH 7.2, 0.15 M NaCl, 0.9 mM CaCl 2 , 0.2 mg/ml chondroitin 4-sulfate). After a wash with solution B, the bound 3 H-labeled oligosaccharides were eluted with 2 M NaCl in 10 mM Tris-HCl, pH 7.2. The percent proportion of the bound radioactivity to the applied radioactivity for each fraction is shown in Fig. 3A. The proportion increased as the molecular size increased. However, a sharp increase in the proportion was observed between HS-III and HS-IV (4 and 17%, respectively). The results suggest that HS-IV is the smallest size of the structures required for HGF binding, which was estimated to be HS octasaccharide judging from its molecular weight and disaccharide composition as described below (see Table II). The chain size dependence of the heparin-binding to HGF was also determined using 3 H-labeled heparin oligosaccharides (Fig. 3B). The octasaccharide (Hep-8) was also the smallest fraction to show a sharp increase in the binding proportion, although the proportions tended to increase as the size of oligosaccharides increased.
The results suggest that the sizes of HS/heparin saccharides are one of the structural factors required for the binding of HS/heparin to HGF and the octasaccharides are the minimal.
Characterization of HGF-bound and -unbound Oligosaccharides-Bound and unbound oligosaccharides of HS-IV were prepared as described under "Experimental Procedures." Rechromatography of the HS-IV-unbound fraction showed that more than 95% of the radioactivity passed through the HGF column reproducibly (data not shown), indicating no significant contamination of HGF-bound species. Both bound and unbound fractions of HS-IV were further fractionated in accordance with their negative charges by ion-exchange chromatography on a Mono-Q column (Fig. 4). Most of HS-IV-bound fraction was eluted at the NaCl concentration of above 0.88 M (fractions 43-50; designated IV-B in Fig. 4A). On the other hand, the HS-IV-unbound fraction was eluted with a broad distribution pattern. But 16% of the HS-IV-unbound fraction was recovered in the subfraction similar in the elution positions to HS-IVbound fraction (fractions 44 -50; designated IV-UB in Fig. 4A). Therefore, the difference in HGF affinity between IV-B and IV-UB may be due to structural factors other than their net negative charges.
Both nonlabeled IV-B and IV-UB, after the extensive digestion with the HSase mixture, were subjected to the compositional analysis by HPLC on a polyamine silica column as described under "Experimental Procedures" (Table II). Comparison of the unsaturated disaccharide compositions between them showed a marked difference: 47% of the disaccharides obtained from IV-B were ⌬Di-(N,6,U)triS, whereas only 26% were in those obtained from IV-UB. Considering the molecular weights of IV-B and IV-UB, these composition data suggested that IV-B and IV-UB corresponded to the octasaccharide (4 disaccharide units) containing at least 2 HexA(2SO 4 )-GlcNSO 3 (6SO 4 ) units and a mixture of the octa-and decasaccharides containing only 1 above unit, respectively. Moreover, considering both the substrate specificities and catalytic properties of enzymes used for the preparation of these HS oligosaccharides, nonreducing ends of the HS oligosaccharides are supposed to have nonsulfated unsaturated HexA. Hence, 2 HexA(2SO 4 )-GlcNSO 3 (6SO 4 ) units in HGF-bound octasaccharides should be localized contiguously or alternately at or near the reducing ends.
HS-V fraction was also fractionated into HGF-bound and -unbound fractions by HGF affinity chromatography. Both V-B and V-UB were fractionated on a Mono-Q column (Fig. 4B), and the resulting fractions (V-B and V-UB) were subjected to the compositional analysis. V-B that was estimated to be a decasaccharide contained more than 50% HexA(2SO 4 )-GlcNSO 3 (6SO 4 ), but V-UB contained only 12% (Table II). Thus, the composition analysis gave similar results to those obtained with IV-B and IV-UB fractions.
To identify the hexuronic acid residues participating in HGF binding, IV-B was treated with nitrous acid at pH 1.5 and then I]HGF to HS-conjugated gel 100 l of PBS containing 0, 0.01, 0.1, 1, 10 g of GAG and 1 ϫ 10 4 dpm (0.2 ng) of [ 125 I]HGF were added to a HS-conjugated Sepharose gel. The reaction mixture was incubated at 4°C with agitation. After 1 h, the mixture was centrifuged (640 ϫ g, 3 min) in microcentrifuge tubes. The gel on the membrane was washed with PBS(ϩ), and radioactivity bound to the gel was determined in a ␥-radiation counter. Nonspecific binding was determined using 100 g/ml heparin. Bovine liver HS fractions 1, 2, and 3 were fractionated by Dowex 1 column chromatography and DEAE-Sephacel column chromatography.  (45). 85% of the total labeled saccharides were recovered in the disaccharide fraction (data not shown). The disaccharides were identified by HPLC on a SAX column. Of these disaccharides, 52% were IdoA(2SO 4 )AMan R (6SO 4 ), and only 2% were GlcA(2SO 4 )AMan R (6SO 4 ). Therefore, HexA(2SO 4 )-GlcNSO 3 (6SO 4 ), which was a major disaccharide component of IV-B, was an IdoA-type. The identification of hexuronic acid residues was also performed with the other HGF-bound fraction, V-B. Molar ratios of disaccharides per mol of IV-B or V-B estimated from both the results of Table II and the above identification of hexuronic acid residues are shown in Table III. In both IV-B and V-B, IdoA(2SO 4 )-GlcNSO 3 (6SO 4 ) was the only component with the content close to or exceeding 2 mol/mol, suggesting an essential involvement of this disaccharide unit in the HGF binding. Other disaccharide components were present in less than 1 mol/mol. However, contents of N-sulfated disaccharides such as IdoA-GlcNSO 3 and GlcA-GlcNSO 3 (6SO 4 ) were relatively high, compared to those of N-acetylated disaccharides, and the sum of these N-sulfated disaccharide contents was more than 1 mol/mol. The results suggest that clustering of 2 IdoA(2SO 4 )-GlcNSO 3 (6SO 4 ) units and one N-sulfated component (HexA-GlcNSO 3 or HexA-GlcNSO 3 (6SO 4 )) may form the binding site for HGF.
HGF Releasing Activities of HS Oligosaccharides and Heparin from the Complex of HGF and HSPGs-Affinities to HGF of HS-bound and -unbound oligosaccharides and heparin were assessed by their releasing activities of HGF from the complex of HGF and HSPGs. The HSPG preparation from rat liver were used to coat ELISA plates. Digoxigenin-HGF was bound on the plate via coated HSPGs. After 1 h of incubation with oligosaccharides at various concentrations on the plate, digoxigenin-HGF yet bound on the plate was determined using anti-digoxigenin Fab fragment as described under "Experimental Procedures." The HGF-releasing activity was compared among HGF-bound HS oligosaccharide (V-B), HGF-unbound HS oligosaccharide (V-UB), and heparin (Fig. 5). The concentrations to give a 50% release of bound HGF were 1.3, 5, and 110 ng/ml for heparin, V-B, and V-UB, respectively. The releasing activity of V-B was 20 times more active than V-UB and only one fourth less than heparin. DISCUSSION Our present study has shown that HGF bound only to heparin and some species of HS, suggesting possible involvements of some unique structures on the chains in the binding (Table  I). HGF affinity gel chromatography of HS oligosaccharides prepared by a limited digestion of bovine liver heparan sulfate with heparitinase I has shown that minimal sizes of the chains for HGF binding are octasaccharide (Fig. 3). Bound and unbound octasaccharides thus obtained were subjected to structural analyses. HS-bound octasaccharides (IV-B) characteristically comprised 2 mol of IdoA(2SO 4 )-GlcNSO 3 (6SO 4 ) per molecule (Table III). These results, considering the fact that their nonreducing ends were nonsulfated, unsaturated hexuronic acid, suggest that at least two IdoA(2SO 4 )-GlcNSO 3 (6SO 4 ) units are present contiguously or alternately each other at or near the reducing ends (see Fig. 6). The presence of this structural unit was also detected in the HSbound decasaccharide fraction (V-B) (Table III).     3. Percent proportions of oligosaccharides with the binding activity to HGF affinity column. Oligosaccharide fractions (4 nmol) containing 1 ϫ 10 5 dpm of 3 H-label which were prepared from bovine liver HS fraction 2 as shown in Fig. 2 A and from heparin by degradation with nitrous acid at pH 1.5 (B) were subjected to a HGF affinity chromatography as described under "Experimental Procedures." After incubated at 4°C for 1 h, the column was washed with solution B and then eluted with 2 M NaCl, 10 mM Tris-HCl, pH 7.2. The elution was analyzed for radioactivity.
Lyon et al. (48) have also suggested that heparan sulfate with a high affinity to HGF apparently has a sequence rich in IdoA and GlcNSO 3 (6SO 4 ) residues. However, according to their results, no contiguous sequence of two or more IdoA(2SO 4 )containing disaccharides appeared to be absolutely necessary for the interaction with HGF, because most of fragments prepared from fetal skin fibroblast HS by digestion with heparinase I which specifically attacks N-sulfated disaccharides containing IdoA(2SO 4 ) residue still retained a HGF affinity.
It is in question in our present study whether HexA(2SO 4 )-GlcNSO 3 units are involved in the binding of HGF to HS directly, since these HexA(2SO 4 )-GlcNSO 3 units comprised only 3.2% of the starting material, bovine liver HS fraction 2.  5. HGF releasing activity of V-B, V-UB, and heparin from the complex with HSPG. Releasing activity was detected by ELISA as described under "Experimental Procedures." Digoxigenin-HGF was added into wells coated with rat liver proteoglycans (0.1 nmol as hexuronate). After 1 h, unbound digoxigenin-HGF was removed, and then V-B (E), V-UB (q), and heparin ( ) at various concentrations were added. After 1 h, the wells were washed, then anti-digoxigenin-AP, Fab fragments were added to yield color. Nonspecific binding was determined using 100 ng/ml heparin.

TABLE III Disaccharide composition per mol of IV-B and V-B
Bound fractions from HS-IV and HS-V were prepared as shown in Fig. 3, and treated with nitrous acid at pH 1.5. The products were labeled by reduction with [ 3 H]NaBH 4 and then subjected to gel chromatography connected to two fast desalting columns. An aliquot of the disaccharide fraction was applied to a Partisil-10 SAX column as described under "Experimental Procedures." The approximate molar ratios of the disaccharide per mol of IV-B and V-B are from the data in Table II