Characterization of Growth Factor-binding Structures in Heparin/Heparan Sulfate Using an Octasaccharide Library*

Heparan sulfate (HS) chains interact with various growth and differentiation factors and morphogens, and the most interactions occur on the specific regions of the chains with certain monosaccharide sequences and sulfation patterns. Here we generated a library of octasaccharides by semienzymatic methods by using recombinant HS 2-O-sulfotransferase and HS 6-O-sulfotransferase, and we have made a systematic investigation of the specific binding structures for various heparin-binding growth factors. An octasaccharide (Octa-I, ΔHexA-GlcNSO3-(HexA-GlcNSO3)3) was prepared by partial heparitinase digestion from completely desulfated N-resulfated heparin. 2-O- and 6-O-sulfated Octa-I were prepared by enzymatically transferring one to three 2-O-sulfate groups and one to three 6-O-sulfate groups per molecule, respectively, to Octa-I. Another octasaccharide containing 3 units of HexA(2SO4)-GlcNSO3(6SO4) was prepared also from heparin. This octasaccharide library was subjected to affinity chromatography for interactions with fibroblast growth factor (FGF)-2, -4, -7, -8, -10, and -18, hepatocyte growth factor, bone morphogenetic protein 6, and vascular endothelial growth factor, respectively. Based upon differences in the affinity to those octasaccharides, the growth factors could be classified roughly into five groups: group 1 needed 2-O-sulfate but not 6-O-sulfate (FGF-2); group 2 needed 6-O-sulfate but not 2-O-sulfate (FGF-10); group 3 had the affinity to both 2-O-sulfate and 6-O-sulfate but preferred 2-O-sulfate (FGF-18, hepatocyte growth factor); group 4 required both 2-O-sulfate and 6-O-sulfate (FGF-4, FGF-7); and group 5 hardly bound to any octasaccharides (FGF-8, bone morphogenetic protein 6, and vascular endothelial growth factor). The approach using the oligosaccharide library may be useful to define specific structures required for binding to various heparin-binding proteins. Octasaccharides with the high affinity to FGF-2 and FGF-10 had the activity to release them, respectively, from their complexes with HS. Thus, the library may provide new reagents to specifically regulate bindings of the growth factors to HS.

Heparan sulfate (HS) 1 exists ubiquitously as a component of proteoglycans on cell surfaces and in extracellular matrix and basement membranes and has divergent structures and functions (1)(2)(3). HS chains are known to interact with a variety of proteins such as heparin-binding growth and differentiation factors (HBGFs), morphogens, extracellular matrix components, protease inhibitors, protease, lipoprotein lipase, and various pathogens (4 -8). These interactions have been shown to play a pivotal role in various patho-physiological phenomena as well as in tissue morphogenesis, as uncovered by recent genetic studies (9 -12). In some of these phenomena, the interactions of HS with certain proteins have been shown in regions of the HS with specific monosaccharide sequences and sulfation patterns. Such functional domains are thought to be generated after the sequential modification steps during the biosynthesis of HS. In these modification steps, HS N-deacetylase/N-sulfotransferases (13)(14)(15)(16), C5 epimerase (17), HS 2-O-sulfotransferase (HS2ST) (18), HS 6-O-sulfotransferases (HS6ST) (19,20), and HS 3-O-sulfotransferase (21,22) are involved. The expression patterns of the individual modification enzymes have been shown to differ from tissue to tissue, and as a result different functional structures of HS may be generated in the different tissues. Such structural diversity introduced to HS may result in the different response of each tissue to various heparin-binding proteins.
So far, the sequences in HS that interact with FGF-1 or FGF-2 have been studied by biochemical and x-ray crystallographic analysis. It became apparent that the FGF-1-binding region was distinct from the minimal FGF-2-binding region (23)(24)(25)(26)(27). In addition to the studies on FGF-1 and FGF-2, the HS sequences that mediate binding and/or activation of some HB-GFs have been reported in the systems including FGF-4 (23,24), FGF-8b (28), hepatocyte growth factor (HGF) (29 -31), and platelet-derived growth factor (32). These studies on the binding structures in HS appear to support the idea that each heparin-binding growth factor may recognize the respective unique structure. However, our knowledge about the heparin/HS structures involved in the interaction with a variety of HBGFs is still limited, and the greater part of the interactions between HS and the heparin-binding proteins remains to be studied. Furthermore, structural analyses of HS with binding activity for HBGFs often give different results when HS is isolated from different sources (29,30).
In the present study, we prepared an octasaccharide library consisting of well defined sulfated octasaccharides. The library comprised 2-O-sulfated or 6-O-sulfated octasaccharides generated from CDSNS-heparin-derived octasaccharide (Octa-I) by in vitro reactions with HS2ST or HS6ST. By using this library, we examined the structures that were specifically bound to the various HBGFs including FGF-2, FGF-4, FGF-7, FGF-8, FGF-10, FGF-18, HGF, BMP-6, and VEGF. Our results show that these HBGFs could be classified roughly into five groups on the basis of the difference in affinity with the oligosaccharides, and offer further evidence for specific interactions between heparinbinding growth factors and the corresponding domain structures in HS. Furthermore, for a physiological relevance of this study, we demonstrated specific release of FGF-10 and FGF-2 from HS by the addition of 6-O-sulfated Octa-I and 2-O-sulfated Octa-I, respectively. Preparation of Octasaccharide Fractions from CDSNS-heparin and Heparin-One octasaccharide composed of HexA-GlcNSO 3 (Octa-I) and another composed of HexA(2SO 4 )-GlcNSO 3 (6SO 4 ) (Octa-II) were prepared from CDSNS-heparin and heparin, respectively, as follows. One hundred milligrams of CDSNS-heparin was digested with 0.1 unit of heparitinase I, which is known to preferentially cleave glucosaminidic linkages to nonsulfated HexA residues in heparin/HS, at 37°C for 2 h. The unsaturated oligosaccharide products were separated by Superdex 30 chromatography based on size. Octasaccharide fractions were pooled and lyophilized. The lyophilized materials were applied to a Mono Q column. The Mono Q column was developed by using a linear gradient from 0.2 to 1.2 M NaCl in 50 mM glycine-HCl, pH 3.0. The fractions eluted around 0.34 to 0.38 M NaCl were pooled and desalted by PD-10 column chromatography. The purified octasaccharide thus obtained was designated Octa-I. About 250 nmol (as HexA) of Octa-I was obtained. One hundred milligrams of heparin was digested with 0.1 unit of heparinase, which is known to cleave preferentially glucosaminidic linkages to 2-O-sulfated IdoUA in heparin/HS, at 37°C for 2 h. Octa-II was purified from the heparin digests by the methods described above except that the fractions eluted over 1.0 M NaCl were pooled in the Mono Q chromatography. About 2.4 mol (as HexA) of Octa-II was obtained. Aliquots of Octa-I and Octa-II were reduced with [ 3 H]NaBH 4 as described by Shively and Conrad (33). The specific activities of 3 H-labeled Octa-I and Octa-II were 1.5 ϫ 10 4 and 1.2 ϫ 10 4 dpm/nmol, respectively.
HBGF Affinity Chromatography of Various Octasaccharides-0.1 nmol of 35 S-labeled octasaccharide was dissolved in 0.5 ml of PBST, 0.9 mM CaCl 2 , and 0.2 mg/ml chondroitin 4-sulfate (Binding buffer) and applied to a syringe column of GF-Sepharose (0.3 ml) equilibrated with Binding buffer at 4°C. The column was shaken gently for 1 h and then washed with 2.5 ml of PBS containing 0.9 mM CaCl 2 and 0.05% Tween 20 (PBST(ϩ)). The column was eluted stepwise with 1 ml of 0.20, 0.25, 0.30, 0.35, 0.40, 0.45, and 0.50 M NaCl in PBST(ϩ) and 2 ml of 2 M NaCl in PBS containing 0.05% Tween 20 (PBST) (Elution buffer). In some cases, the octasaccharides were eluted with 2 ml of Elution buffer after the wash with 2.5 ml of PBST(ϩ). The elution profiles were monitored by measuring the radioactivity in a liquid scintillation counter (30).
Surface Plasmon Resonance Analysis-Real time analysis of the interaction of growth factors and heparin/modified heparin was performed with a BIAcore 2000 SPR biosensor. Streptavidin-conjugated Sensor Chip SA was used to immobilize various glycosaminoglycans. Glycosaminoglycans were biotinylated according to the method recommended by the manufacturer. Two hundred micrograms of heparin, 2ODS-heparin, 6ODS-heparin, or chondroitin 4-sulfate was incubated with 74 g of NHS-LS-Biotin (Pierce) in 100 l of 50 mM sodium bicarbonate buffer, pH 8.5, for 30 min at room temperature. Biotinylated glycosaminoglycans were precipitated with 2.5 volumes of 95% (v/v) ethanol containing 1.3% (w/v) potassium acetate, and the process was repeated three times. In order to immobilize GAGs on the sensor chip SA, 0.5-2 g/ml biotinylated GAGs in PBST were injected at a flow rate of 5 l/min. The injection of biotinylated GAGs produced 10 -200 response units of immobilized GAG on the biosensor surface. The amount of bound material on the biosensor chips was measured in arbitrary response units. All measurements were carried out at room temperature, and refractive index errors due to bulk solvent effects were corrected by subtracting away responses on the non-coated sensor chip for GF concentrations used. Each GF stock solutions were diluted with PBST.
Various concentrations of growth factors were injected across the GAG-coated surface at a flow rate of 5 l/min. The steady state binding level was monitored for 300 s. Sensorgrams were evaluated using BIAevaluation software. Dissociation constants (K D ) for binding could be extracted from the dependence of steady state binding levels on GF concentration. In a steady state, this model calculates K D from a plot of R eq against C according to the equation, K D ϭ R eq (R max Ϫ R eq )/C, where R eq is the steady state response level for the growth factor; R max is the maximal capacity of the sensor chip to bind growth factors expressed in response units; and C is the molar concentration of growth factor.
HBGF-releasing Activity of Octasaccharides from HS-The releasing activity was measured by ELISA as described previously with a minor modification (30). A 96-well streptavidin-coated plate (Thermo Labsystems, Finland) was coated with 0.1 nmol (as hexuronic acid) of biotinylated pig aorta HS for 1 h at room temperature. Wells were washed three times with 200 l of PBS and then blocked with 200 l of PBS containing 10 mg/ml BSA for 1 h with gentle shaking. Wells were washed three times with PBS. Then 100 l of Binding buffer (see above) containing 40 ng/ml digoxigenin-conjugated HBGF and 10 mg/ml BSA was added into each well. After 1 h at room temperature, unbound digoxigenin-conjugated HBGF was removed by three washes with PBST(ϩ). Then 100 l of Binding buffer containing 10 mg/ml BSA and 1 pmol to 1 nmol of each octasaccharide were added into the wells. After 1 h at room temperature, wells were washed, and then alkaline phosphatase-conjugated Fab fragments of anti-digoxigenin antibody (1:1000 dilution) were added. After 1 h at room temperature, unbound Fab fragments were removed by three washes with PBST, and the alkaline phosphatase substrate (1 mg/ml p-nitrophenyl phosphate in 1 M diethanolamine, pH 9.8) was added into each well. The enzyme activity in each well was measured by using a microplate reader. The experiments were independently repeated three times, and statistical analyses were performed using Student's t test. The criterion of significance was shown by p values. Compositional Analysis of Octasaccharides-About 0.3 nmol of octasaccharides 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 , and 5 g of BSA at 37°C for 1 h. Unsaturated disaccharide products were analyzed by fluorometric postcolumn high performance liquid chromatography (HPLC) as reported previously (35).
2-O-Sulfated and 6-O-sulfated Octa-I were prepared by incubating Octa-I with the recombinant HS2ST and HS6ST-1, respectively, together with PAPS as described under "Experimental Procedures." The in vitro sulfated products were separated with Mono Q chromatography (Fig. 1). Both 2-O-sulfated Octa-I and 6-O-sulfated Octa-I were separated into three peaks: 2S-1, 2S-2, and 2S-3 for 2-O-sulfated Octa-I; and 6S-1, 6S-2, and 6S-3 for 6-O-sulfated Octa-I. All of these peaks were eluted at higher NaCl concentration than Octa-I and at lower NaCl concentration than Octa-II. Each peak was pooled sepa-rately. To examine the structure of these products, aliquots of these sulfated octasaccharides were digested extensively with the mixture of heparitinase and heparinase and subjected to HPLC as described under "Experimental Procedures." From the disaccharide compositions shown in Table I, it is evident that 2S-1, 2S-2, and 2S-3 contained 1, 2, and 3 units, respectively, of the HexA(2SO 4 )-GlcNSO 3 component; and 6S-1, 6S-2, and 6S-3 have 1, 2, and 3 units, respectively, of HexA-GlcNSO 3 (6SO 4 ) component. Under the maximum reaction conditions used, more than 60% of Octa-I was sulfated by each sulfotransferase. The yields of these sulfated Octa-I were decreased as the content of 2-O-sulfated or 6-O-sulfate units was increased (Table I).
Binding Abilities of Octa-I and Octa-II to Various HBGFs-Before determining the binding activity of oligosaccharides, we examined the binding ability and capacity of the growth and differentiation factor-conjugated Sepharose 4B affinity columns using 3 H-labeled heparin (2 nmol as HexA). Every affinity column conjugated with FGF-2, FGF-4, FGF-7, FGF-8, FGF-10, FGF-18, HGF, VEGF, or BMP-6 bound more than 80% of the applied 3 H-labeled heparin (data not shown). To examine the binding activity of the oligosaccharide libraries synthesized enzymatically, we first applied [ 3 H]NaBH 4 -reduced Octa-II (0.5 nmol as octasaccharide) to the growth factor-conjugated columns. The amounts of Octa-II bound to various growth factor-conjugated columns are shown in Fig. 2. FGF-2, FGF-4, FGF-18, and HGF bound Octa-II strongly. FGF-10 and FGF-7 also bound Octa-II slightly weaker than FGF-2. In contrast, FGF-8, BMP-6, and VEGF hardly bound Octa-II (less than 15% of the applied). These results indicate that FGF-2, FGF-4, FGF-7, FGF-10, FGF-18, and HGF have high affinity to Octa-II as observed for heparin, but FGF-8, BMP-6, and VEGF have very weak or no affinity to Octa-II, although these proteins showed high affinity to heparin. The results suggest that oligosaccharides longer than octasaccharides may be required for the binding of FGF-8, BMP-6, or VEGF.
Octa-II contains both O-sulfate and N-sulfate. To address which sulfate groups are necessary for the binding to these growth factors, we examined the binding of Octa-I which is devoid of O-sulfate but has N-sulfate. As shown in Fig. 2, Octa-I hardly bound at all to the growth factor-conjugated affinity columns (open bars in Fig. 2). These results clearly indicate that the O-sulfate groups in Octa-II are essential for the binding to these growth factors.
Affinity of the Octasaccharide Library to Various HBGFs-Octa-II is composed of 3 units of HexA(2SO 4 )-GlcNSO 3 (6SO 4 ). To determine whether either or both of the 2-O-sulfate and 6-O-sulfate groups interact with the growth factors, we examined the binding of 2-O-sulfated or 6-O-sulfated Octa-I to the growth factor-conjugated columns that could retain Octa-II. In Fig. 3, the binding of 2S-3 and 6S-3 to FGF-2, FGF-4, FGF-7,   3. Binding of 2S-3 and 6S-3 to various growth and differentiation factors. 100 pmol of octasaccharides "2S-3" and "6S-3" was subjected to the GF affinity column chromatography as described under "Experimental Procedures." After incubation at 4°C for 1 h, the column was washed with Binding buffer and then eluted with Elution buffer (bound fraction). Closed bars and gray bars indicate amount of bound 2S-3 and 6S-3, respectively. one, two, or three 6-O-sulfate groups (6S-1, 6S-2, and 6S-3). To the FGF-2-conjugated column, 57% of 2S-1 and more than 90% of 2S-2 and 2S-3 applied were bound. The majority of the bound 2S-1 was released by 0.35 M NaCl, but a complete release of 2S-2 and 2S-3 was attained with 2.0 M NaCl (Fig. 4A). These results indicate that the presence of only 1 unit of HexA(2SO 4 )-GlcNSO 3 is sufficient to retain the octasaccharides on FGF-2, and that the affinity of octasaccharides containing 2 or 3 units of HexA(2SO 4 )-GlcNSO 3 to FGF-2 is much higher than that of the other growth factors examined. To the FGF-10-conjugated column, 32% of 6S-3 was bound but neither 6S-2 nor 6S-1 was bound (Fig. 4B). These results suggest that 3 units of HexA-GlcNSO 3 (6SO 4 ) are required for the binding of 6-O-sulfated Octa-I to FGF-10. Because 48% of the bound 6S-3 was released by 0.25 M NaCl, the interaction of FGF-10 with 6-O-sulfated octasaccharides should be relatively weak. As observed in Fig.  3, FGF-18 and HGF showed affinity to both 2-O-sulfate and 6-O-sulfate, and the affinity to 2-O-sulfate seemed to be higher than to 6-O-sulfate. These results were confirmed by the binding experiments using different sulfated Octa-I molecules (Fig.  4). To the FGF-18-conjugated column, 90% of 2S-3, 42% of 6S-3, and 23% of 2S-2 were bound, but 6S-2, 6S-1, and 2S-1 were not bound. Quantitative elution of the bound 2S-3 and 6S-3 was achieved with 0.4 and 0.3 M, respectively, of NaCl (Fig. 4, C and  D). Nearly the same results were obtained for the HGF-conjugated column (Fig. 4, E and F); 75% of 2S-3, 35% of 2S-2, and 19% of 6S-3 were bound, but 2S-1, 6S-2, and 6S-1 were not bound. Quantitative elution of the bound 2S-3 and 6S-3 was achieved with 0.35 and 0.25 M, respectively, of NaCl. However, a clear difference between FGF-18 and HGF was observed in the affinity to 6S-3; the affinity of HGF to 6S-3 was much lower than that of not only FGF-18 to 6S-3 but also HGF to 2S-2.
Analysis of the Interactions of VEGF 165 and BMP-6 with Modified Heparins Using Surface Plasmon Resonance Biosensor-As observed above, VEGF 165 , BMP-6, and FGF-8 were able to bind heparin but did not bind Octa-II. To examine how 2-O-sulfate and 6-O-sulfate in heparin contributed to the binding to VEGF 165 , BMP-6, and FGF-8, we determined the disso-ciation constant (K D ) between one of these HBGFs and intact 2-O-desulfated (2ODS) or 6-O-desulfated (6ODS) heparin. K D values were determined by surface plasmon resonance (SPR) biosensor. As shown in Table II, 2-O-sulfate groups were completely removed from 2ODS-heparin, whereas a few 6-O-sulfate groups (up to 1.8%) remained in 6ODS-heparin. The major disaccharide components of heparin, 2ODS-heparin, and 6ODS-heparin were thus thought to be HexA(2SO 4 )-GlcNSO 3 (6SO 4 ), HexA-GlcNSO 3 (6SO 4 ), and HexA(2SO 4 )-Glc-NSO 3 , respectively. As positive controls, K D values were also determined for FGF-2 and HGF. The K D value observed in the system of FGF-2/heparin, FGF-2/2ODS-heparin, and FGF-2/ 6ODS-heparin was 23, 340, and 23 nM, respectively (Fig. 5, A-C, and Table III). Because the K D for FGF-2/6ODS-heparin was the same as the K D for FGF-2/heparin, the presence of 6-O-sulfated groups on GlcNSO 3 residues in heparin appeared to have no effect on the interaction between FGF-2 and heparin. In contrast, the K D for FGF-2/2ODS-heparin was 15-fold that for FGF-2/heparin. In the case of HGF, the measured K D for HGF/heparin, HGF/2ODS-heparin, and HGF/6ODS-heparin was 12, 86, and 58 nM, respectively (Fig. 5, D-F, and Table  III). Both the K D for HGF/2ODS-heparin and the K D for HGF/ 6ODS-heparin were higher than the K D values for HGF/heparin. These observed K D values for the interactions of FGF-2 or HGF with the modified heparins are consistent with the results obtained from the octasaccharide library, and confirm the importance of 2-O-sulfate in the binding of FGF-2 and the importance of both 2-O-and 6-O-sulfate in the binding of HGF. Thus, the K D values obtained by SPR biosensor using biotinylated glycosaminoglycans seem accurate enough to be compared quantitatively. Under the same conditions used for the interaction of FGF-2 and HGF, we determined K D values for the interaction between VEGF 165 or BMP-6 and heparin or the modified ones. The measured K D for VEGF 165 /heparin, VEGF 165 /2ODS-heparin, and VEGF 165 /6ODS-heparin was 165, 524, and 592 nM, respectively (Fig. 5, G-I, and Table III). Both the K D for VEGF 165 /2ODS-heparin and the K D for VEGF 165 / 6ODS-heparin were 3-fold the K D values for VEGF 165 /heparin. The measured K D values for BMP-6/heparin, BMP-6/2ODSheparin, and BMP-6/6ODS-heparin was 6.3, 11, and 15 nM, respectively (Fig. 5, J-L, and Table III), which indicate the extremely high affinity of BMP-6 to heparin or the modified heparins. Even so, both the K D for BMP-6/2ODS-heparin and the K D for BMP-6/6ODS-heparin were 2-or 3-fold the K D values for BMP-6/heparin. These results suggest that both 2-O-sulfate groups on HexA residues and 6-O-sulfate groups on GlcNSO 3 residues contained in heparin/HS are equally required for the interactions with VEGF 165 and BMP-6.
Because FGF-8 was bound nonspecifically to the sensor chips, the K D values for FGF-8/modified heparins could not be determined. It should be of note here that no interactions between these growth factors and chondroitin 4-sulfate have been recognized (data not shown).
FGF-releasing Activity of Octasaccharide Library-As shown above, FGF-2 and FGF-10 bound the octasaccharides typically different from each other in that FGF-2 required 2-O-sulfate but FGF-10 needed 6-O-sulfate. Therefore, we examined whether the addition of 2-O-sulfated Octa-I and 6-O-sulfated Octa-I could release FGF-2 and FGF-10, respectively, from their complexes with pig aorta HS, which has the affinity for both FGF-2 and FGF-10 and is thought to represent natural HS in normal tissues. Streptavidin-coated ELISA plates were coated with the biotinylated HS. Digoxigenin-labeled FGF-2 or FGF-10 was then added to form complexes with the HS on plates. The incubation with various octasaccharides at different concentrations added to the plates released the digoxigenin-labeled FGF from the complexes, which was assessed by quantitating the labeled FGF left on the plates after several washes as described under "Experimental Procedures." The addition of 1 nmol/ml 6S-3 and 2S-3 released 29 and 8% of bound FGF-10, respectively (Fig. 6A), which corresponded to the releasing activity observed by the addition of 0.3 and 0.03 nmol/ml Octa-II, respectively (Fig. 6B), suggesting that 6S-3 is 10-fold more active in releasing FGF-10 than 2S-3. The addition of 1 nmol/ml Octa-II released 45% of bound FGF-10. Other O-sulfated Octa-I have none or only a slight effect on the releasing. In the case of the FGF-2 binding to the HS, the addition of 0.2 nmol/ml 2S-2, 2S-3, and Octa-II had almost the same releasing activity (57-61%), and that of 0.2 nmol/ml 2S-1 and 6S-3 released 26 and 11% of bound FGF-2, respectively (Fig. 6C), which corresponded to the releasing when 0.05 and 0.02 nmol/ml Octa-II were added, respectively (Fig. 6D). 2S-1 still showed 25% of the Octa-II activity although it had only one O-sulfate, but 6S-3 showed less than 10% of the Octa-II activity even though it had three O-sulfate residues. The observed differences of octasaccharide library in the releasing activity are good reflections of the differences in the binding specificity and activity of octasaccharide library to FGF-2 and FGF-10, which were described above. DISCUSSION We demonstrated that octasaccharide libraries composed of well defined sulfated octasaccharides generated by recombinant HS-O-sulfotransferases were useful for characterizing the binding structure for HBGFs. For FGF-2, FGF-4, FGF-7, FGF-8, FGF-10, FGF-18, HGF, BMP-6, and VEGF 165 , we analyzed systematically the binding structures present in heparin/ heparan sulfate, including the octasaccharide library. Based upon differences in affinity, these growth factors could be classified roughly into five groups.  (Fig. 7). The results indicate that the structural domain in heparin/heparan sulfate exhibiting affinity to each HBGF could be differentiated in terms of chain size, position to which the sulfate is attached,   E, H, and K), and 6ODS-heparin (C, F, I, and L). The response in the steady state was plotted against the concentration of each GF. The dissociation constant for the interaction was evaluated with these curves using BIA evaluation software. We also demonstrated relevance of the use of an octasaccharide library consisting of defined sulfated octasaccharides for characterizing the binding structures for HBGFs by their releasing activities of HBGFs from the complexes of HS and HBGFs. The affinity analyses using octasaccharide library revealed unique and almost opposite structural requirements for the bindings of FGF-10 and FGF-2. Therefore, we examined whether the addition of octasaccharides having the affinity to FGF-10 and FGF-2 could release FGF-10 and FGF-2, respectively, from their complexes with aorta HS. As expected from the affinity analyses, octasaccharides with the high affinity to FGF-10 or FGF-2 released specifically the respective FGF from their complexes with HS, and in addition, octasaccharides with the higher affinity gave the higher releasing activity (Fig. 6). These results suggest that the interactions of HS with certain HBGFs involved regions of the HS with specific sulfation patterns, and the exogenous addition or the endogenous generation of such regions as oligosaccharides may affect the bindings of HBGFs to HS and also to HBGF receptors. In the case of FGF, because HS is also known to bind to the growth factor receptors, the supply of HS oligosaccharides, whether it is exogenous or endogenous, may have some effect on their signaling. Further investigation remains on functions as the signaling regulators of these oligosaccharides.
Heparin-binding structures of the FGF family have been investigated by x-ray crystallographic analysis and biochemical analysis (36 -38). From these studies, several basic amino acid residues were found to exist opposite the 2-O-sulfate group in heparin. Our results indicated that some members of the FGF family examined here had affinity to 2-O-sulfate, but FGF-10 exhibited little affinity to 2-O-sulfate. Instead, FGF-10 had affinity to the 6-O-sulfate group. To investigate whether the amino acid residues interacting with 2-O-sulfate groups are conserved in FGF-10, the putative heparin-binding region of FGF-10 was aligned with the same regions of other members of the FGF family having affinity to the 2-O-sulfate group (Fig. 7). It has been shown that the 2-O-sulfate-binding region of FGF-2 consists of Lys-125, Gln-134, Lys-135, and Ala-136 (39), whereas that of FGF-1 is composed of Asn-18, Lys-113, Lys-118, and Gln-127 (40) (see boldface amino acid residues in the partial amino acid sequences of FGF-2 and FGF-1 in Fig. 7). When the glycine boxes, which are commonly found in the FGF family and thought to correspond to motifs for heparin-binding sites, are aligned among FGF-2, FGF-4, FGF-7, FGF-8, FGF-10, and FGF-18, either Gln or Lys residues found in these putative 2-O-sulfate-binding sites are conserved in FGF-4, -7, -8, and -18 (see amino acid residues in the gray boxes in Fig. 7). However, these residues are not conserved in FGF-10 at all but are substituted for other amino acids. Crystallographic analysis of the FGF-10-heparin oligosaccharide complex will reveal the structure of the heparin-binding domain of FGF-10.
We demonstrated previously that the HGF-bound octasaccharide prepared from bovine liver HS contained 2 units of IdoUA(2SO 4 )-GlcNSO 3 (6SO 4 ), whereas the HGF-unbound octasaccharide from the same HS contained only 1 unit of IdoUA(2SO 4 )-GlcNSO 3 (6SO 4 ). On the other hand, Lyon et al. (29) reported that 2-O-sulfate groups contributed marginally to the interaction between HGF and fibroblast HS. Such an apparent discrepancy may be due partly to the difference in the heparan sulfate used. The present study clearly shows that HGF interacts with both 2-O-sulfate and 6-O-sulfate, because both the 2-O-sulfated octasaccharides containing 2 or 3 units of We demonstrated that VEGF 165 did not bind octasaccharides with both 2-O-and 6-O-sulfate groups at all but bound the native heparin, indicating that VEGF 165 requires a longer binding domain than octasaccharide for binding heparin. Both 2-O-sulfate groups on HexA residues and 6-O-sulfate groups on GlcNSO 3 residues appear to contribute equally to the interaction between VEGF 165 and heparin, because the K D value for VEGF 165 /2ODS-heparin was nearly equal to the K D value for VEGF 165 /6ODS-heparin. Although the K D value for VEGF 165 / heparin was much larger than that for FGF-2/heparin or HGF/ heparin, the interaction between VEGF 165 and heparin may still have relevance to some physiological processes, because the activity of VEGF 165 was stimulated in vitro by heparin, and mutant mice expressing only VEGF 121 lacking the heparinbinding domain showed abnormalities of microvascularization. Four spliced variants of VEGF, VEGF 145 , VEGF 165 , VEGF 189 , and VEGF 206 , may thus have important physiological functions through the interaction with HS. It remains to be studied what structures in HS could interact with the spliced variants other than VEGF 165 and how strong binding activity is.
To our knowledge, the direct interaction between BMPs and heparan sulfate has never been studied. However, there are a number of reports to suggest important roles of the interaction in differentiation and morphogenesis. For example, glypican-3, one of a family of six cell surface heparan sulfate proteoglycans in vertebrates, plays an important role in BMP signaling during limb patterning and skeletal development (42) and renal branching morphogenesis (43). The interaction of Dally, a glypican homolog, and Decapentaplegic, a TGF-␤/BMP homolog, in Drosophila functions in the gradient formation of Decapentaplegic as a morphogen necessary for the wing and sensory organ formation (44 -46). In the present study, we have shown that BMP-6 did not bind any octasaccharide, but by SPR analysis it bound heparin and 2ODS-or 6ODS-heparin with conspicuously high affinity. Chondroitin sulfate and CDSNS-heparin were a very poor ligand (data not shown). Therefore, the interaction with heparan sulfate should be very strong if the O-sulfated region is long enough. It is known that inducing activities of some members of the BMP family are tightly restricted to the region around the cells that produce them (47). Therefore, the observed high affinity of BMP-6 in the binding to HS likely reflects a role of HS in trapping the BMP. Because there are a number of the BMP family molecules with different activities (48), the present study raises an interesting question how each of them interacts with HS.
In our study Octa-II, octasaccharide containing three HexA(2SO 4 )-GlcNSO 3 (6SO 4 ) units derived from heparinase-digested heparin did not have the affinity enough to bind to FGF-8-conjugated column under the used conditions (Fig. 2). Because heparin was almost retained to this column, we judged FGF-8 needed the longer size for the binding. However, the minimal saccharide domain for the binding of FGF-8b, a spliced variant of FGF-8, has been reported to encompass 5-7 monosaccharide units of heparin (28). The observed discrepancy might have been caused by our usage of FGF-8, which is not a spliced variant, by some alteration in binding properties such as partial denaturation of our FGF-8 or by the difference in the reducing and nonreducing ends of oligosaccharide structures due to the preparation methods between ours (heparinase digestion) and theirs (nitrous acid degradation).
We prepared sulfated octasaccharides using the recombinant HS2ST and HS6ST-1.  (34,36). All of the FGFs used in this alignment are from human. The sequences of the putative heparin-binding regions of FGFs that are structurally superimposed are shown, where the 2-O-sulfate-binding residues are indicated in boldface. Square shows glycine box, which is thought to correspond to motifs for heparin-binding sites (37). was obtained at a ratio of 1.0:0.40:0.15. The one having 1 unit of 2-O-sulfate was a major product at the beginning of the reaction (Fig. 1, insets). It is thus unlikely that the sulfated octasaccharides behave better as an acceptor for HS2ST than the nonsulfated octasaccharide. As observed in the products of HS6ST-1, 6-O-sulfated Octa-I also separated into three major peaks; the octasaccharides with 1 unit of 6-O-sulfate were obtained in the highest yield even at the end of the reaction.
HS6ST-1 appears to prefer nonsulfated octasaccharide as well. A small peak was observed after 6S-3 in the Mono Q chromatography. This peak may be an octasaccharide with 4 units of 6-O-sulfate groups, although the disaccharide composition of this peak has yet to be determined. It is of interest to examine the order and extent of 6-O-sulfation with HS6ST-1.
There are now a number of studies to demonstrate that HS is important to activate the growth factor binding to its receptor via the complex formation among them (5,9), and longer HS oligosaccharides such as decasaccharides are involved in those interactions (25). Therefore, it is becoming important to generate oligosaccharide library of various sizes. In this regard, our present study may be the initial step for such a future study.
In conclusion, our approach using an octasaccharide library prepared in vitro appears to be useful for defining the specific structures required for binding various heparin-binding proteins. Furthermore, more divergent oligosaccharide libraries will be generated by the combined use of various HS modification enzymes such as other HS6ST isoforms and acceptor oligosaccharides of various sizes, and these oligosaccharides may allow more selective regulation of growth factor activities (see "Addendum"). In fact, the potential of such oligosaccharide libraries, containing mixed 2-O-and 6-O-sulfate substituents, has been demonstrated recently in analysis of sequence requirements for saccharides interacting with FGF-1 and FGF-2 (50).