The interaction between basic fibroblast growth factor and heparan sulfate can prevent the in vitro degradation of the glycosaminoglycan by Chinese hamster ovary cell heparanases.

Heparan sulfate proteoglycans on Chinese hamster ovary (CHO) cell surfaces can bind and internalize basic fibroblast growth factor (bFGF). We have investigated whether this interaction affects heparan sulfate catabolism in vitro by measuring the ability of partially purified CHO heparanase activities to degrade 35S-labeled heparan sulfate glycosaminoglycans in the absence or presence of bFGF. Our studies show that the presence of the growth factor prevents partially purified heparanases from degrading the nascent 81-kDa chains to short 6-kDa products, whether the glycosaminoglycan is free in solution or covalently bound to core proteins. A 30-60 molar excess of the growth factor is required to inhibit completely chain degradation by heparanases, implying that multiple bFGF molecules must be bound to the glycosaminoglycan to prevent heparanase-catalyzed catabolism. This hypothesis is supported by protection studies indicating that nascent CHO heparan sulfate glycosaminoglycans have at least four to eight bFGF binding sites/chain. It does not appear, however, that the growth factor inhibits heparanase-catalyzed degradation of the glycosaminoglycan by binding to the sequence cleaved by the enzyme. Both the nascent and short chains bind bFGF with similar affinity (Kd values of 27.0 +/- 3.5 and 38.9 +/- 5.1 nM, respectively), indicating that heparanase activities do not destroy the bFGF binding sites. Rather, our results suggest that the growth factor interferes sterically with heparanase action by binding the heparan sulfate chain at a sequence next to the cleavage site or at a secondary site recognized by the enzyme.

Heparan sulfate proteoglycans on Chinese hamster ovary (CHO) cell surfaces can bind and internalize basic fibroblast growth factor (bFGF). We have investigated whether this interaction affects heparan sulfate catabolism in vitro by measuring the ability of partially purified CHO heparanase activities to degrade 35 S-labeled heparan sulfate glycosaminoglycans in the absence or presence of bFGF. Our studies show that the presence of the growth factor prevents partially purified heparanases from degrading the nascent 81-kDa chains to short 6-kDa products, whether the glycosaminoglycan is free in solution or covalently bound to core proteins. A 30 -60 molar excess of the growth factor is required to inhibit completely chain degradation by heparanases, implying that multiple bFGF molecules must be bound to the glycosaminoglycan to prevent heparanase-catalyzed catabolism. This hypothesis is supported by protection studies indicating that nascent CHO heparan sulfate glycosaminoglycans have at least four to eight bFGF binding sites/chain. It does not appear, however, that the growth factor inhibits heparanase-catalyzed degradation of the glycosaminoglycan by binding to the sequence cleaved by the enzyme. Both the nascent and short chains bind bFGF with similar affinity (K d values of 27.0 ؎ 3.5 and 38.9 ؎ 5.1 nM, respectively), indicating that heparanase activities do not destroy the bFGF binding sites. Rather, our results suggest that the growth factor interferes sterically with heparanase action by binding the heparan sulfate chain at a sequence next to the cleavage site or at a secondary site recognized by the enzyme.
Heparan sulfate proteoglycans (HSPGs), 1 molecules composed of heparan sulfate (HS) glycosaminoglycan chains covalently linked to a protein core, are ubiquitously present on cell surfaces and in extracellular matrix and basement membranes (1)(2)(3). Their expression appears to be developmentally regulated (4,5) and cell-specific (6). Proteoglycans are impli-cated in a number of cellular processes, including cell adhesion, migration, differentiation, and proliferation (for review, see Refs. 7 and 8). Most of the identified proteoglycan functions are attributed to the interaction of the glycosaminoglycan chain with a protein ligand, such as lipoprotein lipase, fibronectin, or various members of heparin-binding growth factor family (9). The last has received a great deal of attention after the presence of heparin or heparan sulfate had been shown to be prerequisite for high affinity binding of basic fibroblast growth factor (bFGF) to its cell surface fibroblast growth factor receptor (FGFR) (10 -12). It was thought originally that binding to HS might change the growth factor conformation so that it can be recognized by the FGFR; however, cocrystallization of bFGF with heparin oligosaccharides demonstrated no structural changes in the growth factor upon the binding (13). Since heparin can form a ternary complex with bFGF and FGFR (14 -16), it has been proposed that the proteoglycans function as a coreceptor for the growth factor and could mediate FGFR dimerization (16), a step that is required for FGFR tyrosine kinase activation and signal transfer. In addition, HS glycosaminoglycans are thought to act as bFGF reservoirs in extracellular matrix (17,18) and facilitate and direct bFGF diffusion to the sites of its action (for review, see Ref. 19).
It has been demonstrated that bFGF can be internalized both while bound to FGFR or HSPG (20) and that these pathways are different (21). Since the interaction with HS protects bFGF from proteolysis and denaturation (22)(23)(24), there was an interesting possibility that the HS in this complex is protected from degradation as well. Normally, HS glycosaminoglycans are released from core proteins and cleaved to shorter chains by heparanases (2,25). This process is thought to occur in endosomes, since short HS chains are produced in cells with defective or inhibited lysosomal degradation (26,27). If bound bFGF inhibits HS catabolism, this might alter the distribution of HSPG on the cell surface or produce HS glycosaminoglycan chains with specific biological functions. In addition, the interaction of HS chains with bFGF could target the complex to specific cellular locations. Any of these effects could have implications in regulation of cellular functions.
We have characterized previously the initial steps in the catabolism of Chinese hamster ovary (CHO) cell HSPGs (25). Because CHO cells express HSPGs that bind and internalize bFGF (20), we decided to investigate the effects of bFGF ligand on HS catabolism in CHO cells. In this work, we show that bFGF inhibits in vitro cleavage of nascent HS chains from CHO cells by heparanase activities partially purified from the same source. The relationship between heparanase cleavage sites and bFGF binding sequences was analyzed by examining heparanase-degraded glycosaminoglycan chains for the growth factor binding, and the number of bFGF binding sites on nascent HS chain was estimated from bFGF protection assays. Our results led to a tentative model for the distribution of heparanase cleavage sites and bFGF binding sites on the nascent HS chains.

EXPERIMENTAL PROCEDURES
Cell Culture-Chinese hamster ovary cells (CHO K1) were obtained from the American Culture Collection (CCL-61). They were cultured in Ham's F-12 medium (Life Technologies, Inc.) supplemented with 5% calf serum, 100 g/ml streptomycin sulfate, and 100 units/ml penicillin G (Sigma) at 37°C in a 5% CO 2 atmosphere with 100% relative humidity. All radioactive labeling experiments were performed with sulfatefree defined medium with 2 or 10 mM glucose, prepared as described previously (28). Cells were passaged by trypsinization every 3-4 days, and after 15-20 passages fresh cells were revived from frozen stocks stored in liquid nitrogen.
HS Glycosaminoglycan Isolation-Confluent CHO cells were incubated with 50 Ci/ml [ 35 S]H 2 SO 4 or [ 3 H]glucosamine (DuPont NEN) in defined F-12 medium to incorporate label into both long, nascent HS chains and short, degraded glycosaminoglycans. 2 Cells were released from the culture dish with 0.125% trypsin, and the long glycosaminoglycans in the trypsinate and the short cellular HS were isolated separately (25). Briefly, labeled glycosaminoglycans were purified on DEAE-Sephacel columns (Pharmacia Biotech Inc.) and precipitated with 80% (v/v) ethanol in the presence of 0.5 mg/ml chondroitin sulfate carrier (25). Residual peptides were removed from the glycosaminoglycans by incubating them with 0.5 M NaOH, 1 M NaBH 4 (25). Labeled HS chains of the desired size were pooled, concentrated by DEAE-Sephacel columns, desalted on a PD-10 gel filtration column (Pharmacia), lyophilized, and resuspended in water.
Unlabeled HS glycosaminoglycans, used as a carrier in the enzyme assays, were prepared from the trypsinate of CHO cells in a similar manner, except higher amounts of starting material were used. On average, one preparation combined trypsinate from about 500 150-mm plates. The elution of the unlabeled HS glycosaminoglycans from the Sepharose CL-6B column was determined by Alcian blue staining of fraction aliquots immobilized on Immobilon-N membrane (Millipore) using a Bio-Dot microfiltration apparatus (Bio-Rad) (29). The concentration of the purified HS was estimated by carbazole reaction (30).
HSPG Isolation-Confluent CHO cells were incubated for 1 h in defined F-12 medium containing 50 Ci/ml [ 35 S]H 2 SO 4 , and then the labeled proteoglycans were extracted from the cell surface with 1% Triton X-100, 10 mM EDTA, 0.15 M NaCl, 50 mM sodium acetate buffer, pH 6.0, containing 10 mM N-ethylmaleimide, 2 mM phenylmethylsulfonide, 1 g/ml leupeptin, and 0.5 g/ml pepstatin to prevent degradation of protein cores (25). Proteoglycans in the extract were isolated by DEAE-Sephacel chromatography and anion exchange HPLC as described previously (25). Free HS chains were separated from intact proteoglycans by gel filtration on TSK 4000 HPLC column (TosoHaas; Montgomeryville, PA) (25).
Heparanase Activity Purification-The heparanase activity used in these experiments was partially purified from CHO cells. Cells were grown to confluence in 150-mm-diameter plates and detached from the dish with 2 ml of 0.125% trypsin at 37°C. Trypsinization was stopped with soybean trypsin inhibitor, and the cells were released from the plate with a pipette. Cells were sedimented in a clinical centrifuge, resuspended in 5 mM HEPES, pH 7.6, containing 1 g/ml leupeptin, 0.5 g/ml pepstatin, 10 mM N-ethylmaleimide, and 0.2 mM phenylmethylsulfonyl fluoride (100 l/plate of cells), and stored at Ϫ20°C until enough cell protein had been collected to begin the purification (around 240 150-mm dishes). Frozen cells were thawed, and fresh protease inhibitors and glutathione (1 mM final) were added to the suspension, then the cells were broken with a Dounce homogenizer. The cell homogenate was centrifuged for 10 min at 15,000 ϫ g (Sorvall SS-34) to pellet nuclei and large organelles. Detergent and salt were added to the 15,000 ϫ g supernatant to a final concentration of 0.1% Triton X-100 and 0.5 M NaCl, then the mixture was centrifuged at 100,000 ϫ g for 60 min to release heparanase activity from microsomal membranes. The 100,000 ϫ g supernatant was diluted to bring the salt concentration below 0.1 M and applied to a DEAE-Sephacel column equilibrated in 0.01 M HEPES, pH 7.6, 0.1 M NaCl, 1 mM glutathione (buffer A). After the column was washed with buffer A, heparanase activity was eluted from the resin by increasing the NaCl concentration. Fractions containing heparanase activity (determined by the glycosaminoglycan degradation assay) were pooled, dialyzed against buffer A, and applied to a heparin Affi-Gel column equilibrated in 0.01 M sodium acetate, pH 5.5, 0.1 M NaCl, 1 mM glutathione (buffer B). The column was washed with buffer B, then heparanase was eluted by a NaCl gradient in buffer B. Fractions containing heparanase activity were pooled, dialyzed against buffer A, and stored at Ϫ20°C until used in experiments.
Heparanase activity was purified over 100-fold with these steps; however, there were still at least 20 silver-stained bands on a sodium dodecyl sulfate-polyacrylamide gel (data not shown). Activity is concentrated against 30,000 molecular weight cut-off membranes, suggesting that the enzyme(s) is greater than 30 kDa. Characterization of the partially purified activity suggests the K m of the enzyme for the long, 81-kDa CHO HS substrate is 10 -20 nM. The enzyme has been shown to be an endoglucuronidase without any exoglycosidic activity (31).
HS Glycosaminoglycan Degradation Assay-Nascent CHO [ 35 S]HS chains (5,000 -15,000 cpm) were mixed with unlabeled CHO HS chains (11 nM final) 3 and incubated for 16 -20 h at 37°C with partially purified CHO heparanase (1-1.5 g of protein) and varying concentrations of human recombinant bFGF (Promega; Madison, WI) or horse heart cytochrome c (Sigma). The total reaction volume was either 30 or 75 l, and 21 mM citrate, 57 mM phosphate buffer was used to maintain pH 5.5. In some assays, 1 mM dithiothreitol (DTT) (Sigma) was included in the reaction mixture. The extent of cleavage was determined by precipitating the HS chains with 1% cetylpyridinium chloride, 0.32 M NaCl, 40 mM sodium acetate, pH 5.5 (25). Under these conditions, HS chains greater than 40 kDa (based on their elution from gel filtration column) precipitate out of solution and can be separated from soluble shorter chains by centrifugation. 2 The amount of degraded chains was determined from the percentage of radioactivity remaining in the supernatant after centrifugation for 10 min.
Gel Filtration HPLC-The size of the heparanase products was examined by gel filtration chromatography on either a TSK 3000 or TSK 4000 HPLC column. Both columns were equilibrated and run in 0.1 M KH 2 PO 4 , pH 6.0, 0.5 M NaCl, 0.2% Zwittergent, at a flow rate 0.5 ml/min. Fractions of 0.5 ml were collected and assayed for radioactivity. The columns were calibrated with heparin, HS, and chondroitin sulfate polysaccharides of known molecular weight. 4 HSPG Degradation Assay-Purified CHO [ 35 S]HSPGs (5,000 cpm) were mixed with either CHO nascent HS (11 nM final) or HSPG (0.9 mg of HS equivalent/ml) isolated from rat liver (32) and incubated for 16 -20 h at 37°C with partially purified CHO heparanase (1-1.5 g of protein) and varying concentrations of bFGF. The reaction conditions were kept the same as for the 35 S-labeled glycosaminoglycan substrate, except 0.17% CHAPS and protease inhibitors were included in the assay. The reaction was stopped with 0.1 M Tris, 0.8 M NaCl, pH 8.0, and the extent of degradation was examined by gel filtration HPLC on a TSK 4000 column. Similar results were observed when either the CHO glycosaminoglycans or the rat liver proteoglycans were used as carrier.
Affinity Coelectrophoresis-The affinity of the interaction between bFGF and HS chains was determined by affinity coelectrophoresis (33). Labeled HS samples were applied into a horizontal slot in a 1% agarose gel (FMC BioProducts; Rockland, ME) and electrophoresed through precast 42-mm lanes containing different concentrations of bFGF. The gels were prepared as indicated by Lee and Lander (33), using 50 mM MOPSO, 125 mM sodium acetate, pH 7.0, 1 mM DTT, and 0.5% CHAPS or 21 mM citrate, 57 mM phosphate, pH 5.5, 1 mM DTT, and 0.5% CHAPS. The electrophoresis running buffer was identical to the buffer used to prepare the gel, except CHAPS was excluded. Electrophoresis was performed for 110 -130 min at 70 V (constant voltage). Gels containing nascent [ 35 S]HS were dried under vacuum and subjected to autoradiography. To visualize the short [ 35 S]HS chains or the 3 Hlabeled glycosaminoglycans, the gels were sequentially transferred into 100% acetic acid (by incubation in 25, 50, and 100% (v/v) acetic acid, 15 min each), then equilibrated for 1 h in 3% 2,5-diphenyloxazol (Sigma) in acetic acid (w/v). The 2,5-diphenyloxazol in the gel was precipitated by a 1-h incubation in water, then the gels were dried under vacuum and incubated with x-ray film (Fuji) at Ϫ70°C. In all cases, radioactive bands were analyzed by a laser densitometer (pdi, Inc., Huntington Station, NY), and apparent dissociation constants were determined from relative retardation of HS by the growth factor (33).
bFGF Protection Assay-Nascent CHO [ 3 H]HS chains (200,000 cpm) were mixed with unlabeled CHO glycosaminoglycans (11 nM final) and bovine serum albumin (1 mg/ml final) and incubated for 6 h at 37°C with 3.7 milliunits heparinase (EC 4.2.2.7, Seikagaku) and heparitinase I (EC 4.2.2.8, Seikagaku), in the absence or presence of 1-1.5 M bFGF. The reaction buffer was 1 mM calcium acetate, 40 mM sodium acetate, pH 7.0. To ensure the reaction went to completion, another 1.2 milliunits of each enzyme was added after 6 h and the mixture incubated an additional 2 h. The degradation products were analyzed by gel filtration on a P2 column (Bio-Rad, 116 ϫ 1 cm), equilibrated in 0.5 M pyridine acetate, pH 5.0. One-ml fractions were collected at a flow rate of 5 ml/h, and an aliquot of each was assayed for radioactivity. The voided 3 H-labeled oligosaccharides protected from cleavage by the growth factor were pooled, and their size was determined on a TSK 3000 gel filtration column. Nascent CHO HS chains are on average 370 residues long, as estimated from their molecular mass (81 kDa 4 ) and average sugar unit mass (220 Da). Then for each oligosaccharide peak (i) from the P2 column, which showed a relative increase in 3 H-labeled cpm when bFGF was included in the reaction mixture, the total number of protected sugar residues/chain (n i ) can be calculated as follows where P and P bFGF are the percentage of radioactivity eluting in the same peak when bFGF was absent or present in the assay, respectively.
The number of bFGF binding sequences/chain (N) is determined as where dp i (degree of polymerization) is the number of sugar residues in the corresponding oligosaccharide.

RESULTS
bFGF Inhibits in Vitro HS Degradation by CHO Heparanases-An in vitro HS degradation assay, developed in our laboratory, was used to examine the effect of bFGF on the cleavage of HS glycosaminoglycans by partially purified CHO heparanase activities. Nascent HS chains from CHO cells (81 kDa, on average), were incubated with heparanase and varying concentrations of bFGF for 20 h. The long incubation time was chosen to allow heparanases to degrade HS substrate completely to short chains. The extent of cleavage was estimated by precipitating uncleaved chains with cetylpyridinium chloride. The results (Fig. 1) show that increasing concentrations of the growth factor inhibit formation of short soluble HS chains, with 50% inhibition achieved at about 500 nM bFGF. Production of short glycosaminoglycans is abolished completely in the presence of 1,000 -2,500 nM bFGF. The inhibition of HS degradation appears to be specific for the growth factor, since cytochrome c, a protein of size and isoelectric point similar to those of bFGF, has no effect on the process (Fig. 1). A direct interaction between HS and the growth factor appears to be required, since the inhibition of cleavage can be reversed by adding more unlabeled HS, but not chondroitin sulfate, to the assay (data not shown). This would also suggest that bFGF inhibits the degradation by binding HS rather than the enzyme, although at present, we cannot exclude that bFGF might interact with CHO heparanase.
To circumvent the possibility that the interaction of bFGF with short HS chains renders them cetylpyridinium chlorideprecipitable, thus causing the apparent inhibition, we used gel filtration to resolve the heparanase degradation products. This method will also establish whether intermediate size chains are formed when bFGF is present in the degradation assay. Once we confirmed that bFGF did not alter the elution profile of either nascent or short HS (data not shown), the size of the heparanase products created in the presence of various bFGF concentrations was analyzed by gel filtration HPLC on a TSK 3000 column. In the absence of the growth factor, most of the nascent 81-kDa chains, which normally elute at the void volume of the column ( Fig. 2A, arrow), are converted by CHO heparanases to short, 6.0 Ϯ 1.1-kDa species (Fig. 2A). Although 76 nM bFGF does not significantly affect the degradation of the long HS by heparanases ( Fig. 2A), a 5-fold increase of the growth factor concentration is sufficient to prevent most of the cleavage, so that the majority of the glycosaminoglycan chains elute at the void volume ( Fig. 2A). In some experiments, intermediate size chains were formed in the presence of bFGF concentrations below and around 350 nM (data not shown). These observations suggest that at lower bFGF concentrations, the growth factor only prevents the catabolism of the glycosaminoglycan in the vicinity of its binding site. To ensure that the voided material did not consist of multiple sized HS species, the experiments were repeated and the products analyzed on a TSK 4000 column, which has better resolution for higher molecular weights. The results obtained correlate well with those from TSK 3000 HPLC. Formation of intermediate sized chains is observed at 200 -300 nM bFGF, but at higher growth factor concentrations the majority of glycosaminoglycans elute from the column at a position comparable to that of the nascent HS chains (data not shown). These experiments suggest that a 30 -60 molar excess of bFGF is required to inhibit HS catabolism completely.
It has been shown that in a nonreducing environment bFGF is able to form dimers through surface cysteine residues (34). Although these dimers do not appear to have a biological function (34), since electrophoretic analysis of the bFGF used in our assays showed that the preparation was a mixture of monomeric and dimeric forms (data not shown), we wanted to examine whether bFGF dimers might carry out the inhibition in a manner different from that of the growth factor monomers. Therefore, we repeated the degradation of nascent HS in the presence of 1 mM DTT and analyzed the degradation products bFGF Inhibits in Vitro Catabolism of Heparan Sulfate on the TSK 3000 column (Fig. 2B). Under reducing conditions when only bFGF monomers are present in the assay, higher concentrations of the growth factor are necessary to achieve the same inhibitory effect as in the absence of DTT (compare Fig. 2,  A and B). These results suggest that the bFGF dimers shield the HS chain from heparanase action more efficiently than the bFGF monomers.
In CHO cells, the initial catabolic steps may occur while the HS chains are still covalently attached to the protein core (25). Therefore, we investigated whether bFGF also inhibits HS chain degradation when the glycosaminoglycans are a part of a proteoglycan. Instead of free glycosaminoglycans, [ 35 S]HSPGs were used as a substrate in the in vitro heparanase assay, and the reaction products were analyzed on a TSK 4000 gel filtration column (Fig. 3). In the absence of the growth factor, heparanases cleave HS chains off the protein core and degrade them to 6-kDa species (Fig. 3). In the presence of 381 nM bFGF, intermediate size chains are formed instead (Fig. 3). Presumably, these products are both partially degraded free glycosaminoglycans and short chains still attached to the protein cores. When the growth factor concentration is increased to 951 nM, the catabolism of the proteoglycans is virtually abolished, and the majority of the HS material elutes from the column with a K av comparable to that of an uncleaved HSPG substrate (Fig.  3). These results suggest that besides inhibiting glycosaminoglycan degradation, bFGF also prevents heparanases from releasing HS chains from protein cores. The inhibition of HS and HSPG catabolism in vitro occurs at a similar concentration of bFGF, indicating that the protein core has no significant effect on the ability of the growth factor to inhibit heparanasecatalyzed degradation of attached HS chains.
bFGF Binds to Both Long and Short HS Chains from CHO Cells-One possibility for the observation that the binding of bFGF to the HS chains prevents heparanases from acting on the glycosaminoglycan would be that the growth factor binds to the same sequence cleaved by the enzyme. If this were the case, one would expect that the short heparanase-degraded glycosaminoglycans would no longer bind bFGF. To test this, nascent and heparanase-degraded HS chains were isolated from CHO cells (see "Experimental Procedures") and examined for their ability to bind bFGF by affinity coelectrophoresis. Most of the experiments were done using glycosaminoglycans labeled with 35 SO 4 , which will incorporate radioactivity only into modified regions of the HS chain. To avoid the possibility that some less modified short glycosaminoglycans would not be detected with the 35 S label, several experiments were performed with glycosaminoglycans labeled with [ 3 H]glucosamine, which uniformly incorporates radioactivity into the chain. Tritiated and 35 Slabeled HS chains behaved identically when analyzed for bFGF binding (data not shown).
Heparanases do not appear to destroy bFGF binding sites, since the migration of both nascent and heparanase-degraded chains is retarded in the presence of bFGF (Fig. 4). The affinity of the long and short HS chains for bFGF was determined from the retardation coefficients obtained from affinity coelectrophoresis (33), and a nonlinear least square analysis was used to estimate the dissociation constants for the growth factor-HS interaction (Fig. 5). At pH 7.0, the apparent dissociation constant for the complex of long HS and bFGF was estimated to be 27.0 Ϯ 3.5 nM (Fig. 5A), which is reasonably close to the value measured by Moscatelli for the interaction of bFGF with CHO cell surface HSPGs (35). The apparent dissociation constant for short chains at the same pH was slightly higher than the value for the nascent glycosaminoglycans, averaging 38.9 Ϯ 5.1 nM (Fig. 5B). To correlate our results with findings from the heparanase assays, several runs were performed at pH 5. bFGF Inhibits in Vitro Catabolism of Heparan Sulfate it can be concluded that the bFGF binding sequences are not altered significantly by heparanase action. Therefore, these results would argue against a model in which bFGF prevents heparanase from acting on the substrate by binding in the middle of the enzyme cleavage site. However, our data do not rule out the possibility that the heparanase cleavage site and the bFGF binding sequence lie next to each other on the glycosaminoglycan chain or that heparanases must bind to a second recognition site that is blocked by the growth factor.
Although bFGF binds to long and short CHO glycosaminoglycans with similar affinities, there are some differences in the affinity coelectrophoresis profiles between nascent and heparanase-degraded HS species (Fig. 4). Long, nascent HS chains behave uniformly with regard to bFGF binding (Fig.  4A), indicating they all have a similar average affinity for the growth factor. In contrast, the short HS species behave as a rather heterogeneous population (Fig. 4B). This apparent heterogeneity may arise from populations of chains with different affinities for the growth factor. To test this possibility, short chains were separated into subpopulations based on their migration in an agarose gel with a single bFGF concentration and were reanalyzed for bFGF binding by affinity coelectrophoresis. Against the same concentration of growth factor, the migration profiles of the short glycosaminoglycan subpopulations were nearly identical to each other and to the parental population (data not shown), indicating that they all have similar affinity for bFGF. This would suggest that charge and length variations between the glycosaminoglycan chains, or a higher diffusion rate caused by their smaller size, may be responsible for the apparent heterogeneity of the short HS.
Nascent CHO HS Chains Carry Multiple bFGF Binding Sites-The results from our enzyme assays and bFGF binding affinity studies are consistent with a model where bFGF binds to the glycosaminoglycan at a site next to the sequence recognized and/or cleaved by CHO heparanase and thereby blocks the enzymes from degrading the HS chain. Heparanases cleave the nascent CHO HS on average 13 times, since the long chains are approximately 81 kDa and the short heparanase products are 6 kDa. 4 If our model is correct, there must be multiple bFGF binding sites on the glycosaminoglycan chain as well. We attempted to measure the number of bFGF binding sites on a CHO HS chain by retaining the bFGF-glycosaminoglycan complexes on positively charged membranes or DEAE microfilters, but we were unsuccessful because of the high nonspecific binding of the growth factor to both supports. Instead, we decided to use the bacterial polysaccharide lyases to estimate how many bFGF molecules could bind a CHO HS chain.
Unlike the mammalian heparanases, the specificities of the bacterial polysaccharide lyases, heparinase (EC 4.2.2.7) and heparitinase I (EC 4.2.2.8), are relatively well characterized. Heparinase cleaves the (1)(2)(3)(4) linkage between N-sulfated glucosamine and 2-O-sulfated iduronic acid (36), destroying the highly sulfated bFGF binding sites on HS chains (37). Heparitinase I, on the other hand, acts primarily at the (1-4) linkages between N-acetylated or N-sulfated glucosamine and glucuronic acid (38,39) and leaves intact the binding sites for the growth factor (37). Accordingly, we have observed that in our Relative mobilities of HS species in the presence and absence of bFGF (Fig. 4) were determined by a laser densitometer (pdi, Inc.) and used for calculating the retardation coefficient R. For both nascent and short HS chains, retardation coefficients from four independent experiments were averaged, and the apparent dissociation constant K d was obtained by fitting the data into the equation r ϭ R max *[bFGF]/(K d ϩ [bFGF], where R max is retardation coefficient for HS saturated with bFGF. bFGF Inhibits in Vitro Catabolism of Heparan Sulfate assays, bFGF prevented heparinase from degrading the CHO glycosaminoglycan, presumably by binding at the same site cleaved by the enzyme, whereas the action of heparitinase I was unaffected by comparable growth factor concentrations (data not shown).
Both heparinase and heparitinase I degrade the long HS to shorter chains (data not shown), indicating that the CHO glycosaminoglycans contain the sequences cleaved by the bacterial enzymes. When nascent [ 3 H]HS chains are exhaustively cleaved with a mixture of heparinase and heparitinase I, the glycosaminoglycan is almost completely degraded to trisaccharides and disaccharides, which can be resolved on a P2 gel filtration column (Fig. 6). If excess bFGF is present in the incubation mixture, longer 3 H-labeled oligosaccharides are observed, which presumably were protected from the enzymes by the growth factor (Fig. 6). These protected sequences elute from the P2 column as octasaccharides and as longer chains at the void volume (Fig. 6). Analysis of the voided oligosaccharides on the TSK 3000 HPLC column shows that the majority are tetradecasaccharides (data not shown). 5 The fraction of oligosaccharides in the bFGF-protected peaks corresponds to three to seven tetradecasaccharide sequences and one octasaccharide sequence/chain, which suggests that each nascent CHO HS chain has at least four to eight bFGF binding sites. DISCUSSION HSPGs have been shown to participate in the catabolism of several ligands, including lipoprotein lipase (40,41), thrombospondin (42), and bFGF (20,21). However, it is not generally known whether the binding of a ligand to the glycosaminoglycan chain affects the catabolism of the HSPG. In this work, we demonstrate that the specific interaction between bFGF and HS glycosaminoglycans can prevent partially purified CHO heparanase activities from degrading free HS chains or HSPGs in vitro (Figs. 1-3). A 30 -60-fold molar excess of the growth factor was required to inhibit completely the formation of the 6-kDa HS degradation products, which suggests that multiple bFGF binding sites must be occupied to prevent heparanases from cleaving the chain. Size analysis of nascent and degraded HS chains indicates that, on average, the long CHO glycosaminoglycans are cleaved by heparanases to 14 short pieces. Since all of the short HS chains bind the growth factor (Figs. 4 and 5), it suggests that the nascent glycosaminoglycan may have as many as 14 bFGF binding sequences. Our bFGF protection assay confirms that long HS chains contain multiple bFGF binding sites (Fig. 6). There appear to be four to eight binding sites/chain; however, this number may be underestimated. Since the bFGF⅐HS complex dissociates rapidly (35), it is possible that the heparin lyases were able to degrade the chain at bFGF binding sites during the incubation. A higher concentration of the growth factor and shorter incubation times with the heparin lyases may be necessary to detect all of the bFGF binding sequences on the CHO HS chains.
The affinities of the nascent and degraded HS chains for the growth factor are similar (Fig. 5), indicating that the bFGF binding sites on the long glycosaminoglycans are not destroyed by heparanase action. This result argues against a model where bFGF prevents heparanases from degrading the chain by binding to the same sequence cleaved by the enzyme. Further support for our conclusion that the heparanase cleavage sites and bFGF binding sites do not coincide comes from experiments examining the CHO heparanase cleavage site (31). These studies show that the disaccharide at the reducing end of the short, heparanase-derived chains is [N-acetylglucosamine (1-4) glucuronic acid], which is not likely to occur in the middle of the highly modified bFGF binding sequence (43)(44)(45). Instead, we propose that bFGF binds the glycosaminoglycan chain at a site where it blocks the access of the enzyme to the sequence that is cleaved (Fig. 7). By binding next to the cleavage site, bFGF may shield the substrate from the enzyme or alter the structure of nearby sugar residues (46) that are required for heparanase recognition. Structural analysis of the heparanase-derived HS indicates that at least one population of short chains has a highly modified sequence that begins 3-5 residues from the cleaved glycosidic bond (31), and it is likely that the growth factor could bind to this sequence and prevent heparanases from acting. Alternatively, bFGF could be preventing heparanases from binding at a site other than that cleaved by the enzyme. Most intracellular heparanases require the substrate to be at least 10 kDa to recognize and cleave the chain (2), which would suggest that the enzyme may bind the glycosaminoglycan at more than one place to align the chain correctly at the catalytic site. If bFGF is bound at this secondary site, heparanases would be unable to bind the substrate, and thus chain degradation would be inhibited. At the present time, we cannot discriminate between bFGF blocking enzyme 5 The bFGF-protected tetradecasaccharides were isolated, and the affinity of the fragments for the growth factor was analyzed by affinity coelectrophoresis. These studies showed that the protected fragments bind bFGF with affinity similar to that of the nascent or short HS chains.
FIG. 6. bFGF protects octa-and tetradecasaccharides from cleavage by a mixture of bacterial heparin lyases. Nascent [ 3 H]HS chains were degraded exhaustively by a mixture of heparinase and heparitinase I in the presence (q) or absence (E) of bFGF, and the cleavage products were resolved on P2 gel filtration column. The number of sugar residues in the oligosaccharides is indicated by the numbers above the peaks. The oligomer eluting in the void volume was determined to be 14 sugars in length by gel filtration HPLC on TSK 3000 column. This graph represents one of four independent experiments, where we observed a 10 -27% and 1-3% increase in material eluting in the void volume and as octasaccharides, respectively, when bFGF was present in the incubation mixture. bFGF Inhibits in Vitro Catabolism of Heparan Sulfate access at the cleavage site or at a secondary site necessary for substrate recognition.
Our observation that the mixed monomer-dimer bFGF population was a more effective inhibitor of heparanase-catalyzed HS degradation than the monomer alone (Fig. 2) would suggest that the dimer creates a larger spatial block to deny enzyme access. This could be accomplished by having the dimer protrude far from the glycosaminoglycan or by having two bFGF molecules bound to adjacent sites on the HS chain. The latter hypothesis is supported by our finding that the bFGF inhibition of HS degradation, measured by precipitation of uncleaved chains, can best be approximated by a sigmoidal fit (Fig. 1), which indicates that the inhibition depends on a binding of two bFGF molecules. Furthermore, the concentration of bFGF necessary to achieve 50% inhibition of HS degradation is estimated to be about 500 nM (Fig. 1), which is close to the square of the K d for the bFGF monomer-HS interaction determined by affinity coelectrophoresis (27 nM) (Fig. 5), again indicating that binding of two bFGF monomers is required to produce the inhibitory effect. Adjacent bFGF binding sites on nascent HS glycosaminoglycans may be relevant to the heparin-mediated FGF oligomerization, which was demonstrated in some studies (12,47).
It is not likely that bFGF would affect bulk HS catabolism inside CHO cells. Since multiple bFGF molecules must be bound to the HS chain to inhibit fully its degradation by heparanases in vitro (Figs. 1-3), the concentration of bFGF encountered by the cell HSPGs in vivo (48,49) will probably not be high enough to prevent their catabolism significantly. Indeed, heparanase activities secreted from platelets, neutrophils, or lymphoma cells were found to be able to liberate bFGF bound to matrix proteoglycans (18) or human endothelial cell perlecan (50). Although the difference in the effect of bFGF may be the result of distinct specificities between intracellular and extracellular heparanases (2,25,51) or may arise from structural differences among different HSPGs (44,52,53), it could also indicate that the in vitro inhibition we have observed may not be physiologically relevant.
On the other hand, the interaction between HS and bFGF could alter the cellular destination of either growth factor or HS species. Since the short HS chains interact with bFGF with affinity similar to that of the nascent HS (Figs. 4 and 5), heparanase-degraded chains would still be able to bind and/or protect bFGF. Interaction of short HS with bFGF may alter the secretion of the glycosaminoglycans from the cells and thus affect their possible biological function. This interaction may be also relevant for translocation of bFGF to the nucleus, which has been proposed to be an alternative signal transduction pathway (54). Both 125 I-bFGF and [ 35 S]HS glycosaminoglycans were detected in CHO and bovine endothelial cell nuclei (55), suggesting that the growth factor may need to be complexed to the HS chain to be protected from degradation and delivered to the nucleus.
The finding that a specific interaction with the glycosaminoglycan chains interferes with their degradation may be valid for other HS ligands. Cell surface proteoglycans serve as receptors for numerous extracellular molecules (9), and since these ligands are often internalized along with the HSPG, they may be expected to affect proteoglycan catabolism. The factors that are likely to determine whether the ligand inhibits HS degradation and whether it occurs in vivo would include the affinity and specificity of the interaction, the glycosaminoglycan sequence bound by the ligand, and the stability of the complex in the acidic endosomal environment. The notion that ligands other than bFGF have the ability to inhibit HS cleavage is supported by our additional studies that show ␤-amyloid, an-other molecule known to be complexed with HS in vivo (56), also prevents CHO heparanases from degrading the glycosaminoglycan in vitro. 6 If inhibition of proteoglycan catabolism by some ligands occurs inside the cells, it could be one of the factors determining the fate of the HSPG-ligand complex (21,40,42).