Monomer Complexes of Basic Fibroblast Growth Factor and Heparan Sulfate Oligosaccharides Are the Minimal Functional Unit for Cell Activation*

The interaction of basic fibroblast growth factor (bFGF) with heparan sulfate (HS)/heparin has been shown to strongly enhance the activity of the growth factor although the mechanism of activation is unclear. We have addressed the issue of the minimal stoichiometry of an active HS oligosaccharide·bFGF complex by chemically cross-linking the two components to form novel covalent conjugates. The cross-linking procedure produced both monomeric and dimeric bFGF·oligosaccharide complexes, which were purified to homogeneity. Dimer conjugates were shown to have been formed as a result of disulfide bridging of monomer conjugates. These monomer conjugates were subsequently found to be biologically active in a mitogenesis assay. We therefore conclude that a monomeric bFGF·oligosaccharide complex is the minimal functional unit required for mitogenic stimulation.

Basic fibroblast growth factor (bFGF) 1 is one of a family of at least 18 polypeptides (reviewed in Ref. 1; also see Refs. [2][3][4][5], and has been shown to influence a variety of cellular processes such as proliferation, migration, and differentiation (6 -8). It has been implicated in a number of disease states, including tumor growth, rheumatoid arthritis, and diabetic retinopathy, due to its ability to stimulate angiogenesis (9,10). This family of growth factors act primarily through high affinity tyrosine kinase receptors (FGFRs) (11), although in addition their activity is modulated by lower affinity heparan sulfate (HS) proteoglycan receptors (12)(13)(14). However, the mechanism by which this occurs is far from understood. The FGFR family comprises four related molecules, which each contain a highly conserved tyrosine kinase domain (15), further diversity being provided by the existence of alternatively spliced isoforms (15)(16)(17). Intracellular signaling is believed to be initiated by receptor dimerization and receptor transphosphorylation (18,19); however, signaling without receptor dimerization cannot be ruled out, as it has not yet been shown whether FGFRs are activated in their monomeric forms. Indeed it has been suggested that bFGF activates multiple signaling pathways by utilizing either monomeric FGFRs or FGFR dimers/multimers (20).
Several HS binding sites on bFGF have been identified with the major site comprising residues Asn-27, Lys-26, Lys-125, Lys-135, and Arg-120 (21)(22)(23). A second possible site was also identified using synthetic di-and trisaccharides and x-ray crystallography (24). A HS binding site has also been identified on the FGFR and was found to be located between the receptors Ig loops I and II (25). The primary high affinity FGFR binding domain has also been identified on bFGF and was found to consist of three hydrophobic amino acids (Tyr-103, Leu-140, and Met-142) and two polar amino acids (Arg-44 and Asn-101) (26,27). A secondary low affinity FGFR binding site was also identified (on bFGF) and comprises amino acids Lys-110, Tyr-111, and Trp-144 (26 -28). Various groups have identified sequences within HS/heparin that interact strongly with bFGF (29 -34). These oligosaccharides were found to be enriched in IdceA(2S)␣1,4GlcNSO 3 (Ϯ 6S) disaccharides, with their affinity for bFGF increasing with 2-O-sulfate content. In addition, no role was established for 6-O-sulfation in the interaction with the primary bFGF binding site (23). The minimum length of saccharide required to activate bFGF in mitogenic assays has been variously reported as ranging from di-to dodecasaccharide (14,24,31,32,(35)(36)(37).
The investigations described above have led to a number of models being proposed by which HS⅐bFGF⅐FGFR interact to bring about a biological response. These include the simultaneous binding of two bFGF molecules to a HS oligosaccharide sequence within the HS chain so as to present bFGF dimers to the FGFRs, thereby enabling two high affinity FGFRs to be brought into close proximity to one another for receptor dimerization and transphosphorylation (14,22,29,38,39). An alternative HS⅐bFGF monomer model has been suggested, in which HS acts as a bridge by simultaneously binding single molecules of bFGF and FGFR to form a signaling complex (25,31). Another HS⅐bFGF monomeric model has also been proposed, in which a monomeric complex of HS and bFGF facilitates receptor dimerization through two FGFR-binding interfaces on the growth factor (26). Other models have also been proposed, with complexes as large as bFGF tetramers being implicated as an active complex (40), as in cytokines such as interferon ␥ and platelet factor 4 (41,42). However, conditions used to observe these complexes differ considerably from the component ratios found to be active in cell assay systems.
One of the major issues still to be resolved in the bFGF mitogenic signaling mechanism is the exact nature of the sig-naling complex and, in particular, the stoichiometry of the HS⅐bFGF complex. In this study, we have produced novel covalently cross-linked conjugates of bFGF and HS oligosaccharides, which no longer bind any further HS but are fully active biologically. These conjugates have enabled us to determine the minimal bFGF⅐HS oligosaccharide stoichiometry required for initiating mitogenic activity. Preparation of HS Dodecasaccharides (dp12)-100 mg of porcine mucosal HS in 0.5 ml of heparinase buffer (100 mM sodium acetate, 0.1 mM calcium acetate, pH 7.0), was incubated initially with 0.25 IU of heparinase III followed by two further additions after 24 and 48 h. The digest was monitored by absorbance at 232 nm until no further increase occurred. Oligosaccharides produced by heparinase III digestion were then resolved using a Bio-Gel P6 column (1.5 ϫ 170 cm) in 0.5 M NH 4 HCO 3 at a flow rate of 6 ml/h. Fractions of 1 ml were collected and oligosaccharides detected by monitoring the absorbance at 232 nm. Size-defined oligosaccharides dp2-16 were pooled, freeze-dried, and stored at Ϫ80°C until required.

Materials
Zero Length Cross-linking-Cross-linking was carried out essentially as described previously by Grabarek and Gergely (43); however, conditions were chosen so as to optimize yields of cross-linked products. Briefly, for preparative reactions, dp12 oligosaccharides in coupling buffer (0.1 M NaCl, 0.1 M MES, pH 6.0) were incubated for 15 min at 25°C with 6 mM EDC and 15 mM S-NHS. The reaction was terminated by addition of ␤-mercaptoethanol to a final concentration of 20 mM. Activated oligosaccharides were then added to a 50 M solution of bFGF in coupling buffer, to give a bFGF:dp12 molar ratio of 1:4 and incubated for 2 h at 25°C. Products were analyzed by standard SDS-PAGE (12% gel) and Western blotting, followed by enhanced chemiluminescence (ECL) immunodetection prior to purification.
Immunodetection of bFGF-Western blots of SDS-PAGE gels on nitrocellulose membranes were blocked overnight with 1% (w/v) bovine serum albumin in PBS and then washed three times with PBS, 0.05% (v/v) Tween 20. The membrane was then incubated with a 1:200 dilution of rabbit anti human bFGF monoclonal antibody for 2 h at 4°C. The antibody solution was then removed and the membrane washed three times with PBS, 0.05% (v/v) Tween 20. The membrane was then further incubated with a 1:1000 dilution of horseradish peroxidaseconjugated swine anti-rabbit IgG for 1 h. The membrane was then finally washed eight times in succession with PBS, 0.05% (v/v) Tween 20 and the presence of bFGF visualized by ECL (Amersham Pharmacia Biotech) following the manufacturer's protocol.
Purification of Cross-linked Products-Cross-linked samples were applied to two coupled Superose 12 columns linked to a Gilson HPLC system, equilibrated in 50 mM phosphate buffer containing 2.0 M NaCl, pH 7.4. Samples were eluted at a flow rate of 0.5 ml/min, fractions (250 l) were collected, and elution profile monitored by absorbance at 280 nm. Fractions containing protein were desalted by microdialysis (M r cut-off 12,000) against 5 mM Tris-HCl, pH 7.4, at a flow rate of 1.5 ml/min overnight at 4°C. Desalted samples were analyzed for purity by SDS-PAGE, followed by Western blotting and ECL immunodetection. Fractions containing cross-linked products were pooled and reapplied to the columns and the procedure repeated as above. Finally, purified components were lyophilized and stored at Ϫ80°C until required.
F32 Cell Mitogenesis Assay-BaF3 cells transfected with FGFR1 (designated F32 cells) (14) were routinely maintained in RPMI 1640 medium, 10% horse serum supplemented with interleukin-3 conditioned medium (prepared from WEHI 3b cells) at 37°C, 5% CO 2 . For the assay system, F32 cells were plated into 96-well plates at a density of 50,000 cells/well in 100 l of RPMI 1640 medium, 10% horse serum supplemented with cross-linked conjugates. Cells were incubated for 46 h before addition of [ 3 H]thymidine (0.3 Ci/well) for another 2 h. Incorporation of [ 3 H]thymidine was stopped by harvesting cells on a Filtermate-196 cell harvester. Cells were allowed to air-dry before addition of 25 l of Microscint O to each well, and incorporated radioactivity counted on a top count system.
Filter Binding Assay-The assay was performed using a membrane filtration apparatus (Millipore, Watford, Herts, UK). Briefly, 4 g of bFGF or bFGF⅐HS complex was applied to the filter in binding buffer (10 mM Tris-HCl, pH 7.3). Filters were then washed with 10 ml of 2.0 M NaCl to remove any non-cross-linked oligosaccharide that may have been present, and then re-equilibrated by washing with binding buffer. Metabolically labeled endothelial cell [ 3 H]HS, prepared as described previously (44), was applied in 5 ml of binding buffer and cycled through the filter three times. Filters were then washed twice with 5 ml of binding buffer to remove unbound [ 3 H]HS. Bound [ 3 H]HS was then removed from the growth factor by washing, first with three 5-ml aliquots of 0.3 M NaCl, followed by further washing with three 5-ml aliquots of 2.0 M NaCl in binding buffer. Fractions (5 ml) were collected and eluted material quantified by scintillation counting.

RESULTS
In order to produce potentially active HS oligosaccharide⅐bFGF conjugates, we have employed a specific method of zero length cross-linking. As no spacer is introduced using this method, crosslinking should only occur between the oligosaccharide and amino acid side chains within the actual HS binding site of bFGF. The cross-linking process is a two-step procedure, initially involving a brief incubation of HS oligosaccharide with EDC in the presence of NHS. This results in the conversion of some of the oligosaccharide carboxyls into succinimide esters. This oligosaccharide activation reaction is then terminated by neutralizing the EDC with ␤-mercaptoethanol prior to addition of growth factor. Cross-linking arises from the reaction of the succinimide esters with the lysine ⑀-amino groups of the growth factor. A two-step process has the advantage that only the oligosaccharide comes into contact with the active cross-linking reagents; hence, only oligosaccharide protein crosslinks are formed. This eliminates possible complications arising from protein-protein cross-linking, which could perturb further interaction of the conjugates with FGFRs.
Analysis of Products Obtained from the Cross-linking of bFGF to dp12 Oligosaccharides-Preliminary experiments were carried out using the zero length cross-linking procedure, in order to assess the type of conjugates formed and to optimize the yield of products. Oligosaccharides of six disaccharides in length (dp12) were chosen, as these have previously been shown to be the minimum size saccharide fragments that have maximal bFGF stimulating activity, when compared with intact chains (31,37). Fig. 1A shows an SDS-PAGE and ECL immunodetection analysis of native bFGF (approximate mass of 18 kDa) and the products obtained by cross-linking of dp12 oligosaccharides to bFGF. It can be seen that the products of cross-linking include both dimer (CL-dp12⅐bFGF dimer) and monomer (CL-dp12⅐bFGF monomer) conjugates of oligosaccharide and growth factor. Increasing the ratio of oligosaccharide to growth factor resulted in higher yields of both CL-dp12⅐bFGF monomers and CL-dp12⅐bFGF dimers, with the most striking increase seen with the CL-dp12⅐bFGF monomers. However, the majority of the growth factor remained in the native form. This is most likely due to a combination of the half-life of the active oligosaccharide esters being only 1-2 h, and the slow off-rate of the oligosaccharide when bound to the growth factor. Control experiments in the absence of crosslinking reagents showed no conjugate bands and no increase in disulfide-bonded bFGF dimers (Fig. 1A, lanes 1 and 2). In order to optimize yields of CL-dp12⅐bFGF dimers and CL-dp12⅐bFGF monomers within a single reaction, the ratio of bFGF:dp12 oligosaccharide was varied from 4:1 to 1:32 (results not shown). Additionally, the ratio of EDC to S-NHS has been shown to effect coupling efficiency (43); in light of this, the ratio of EDC:S-NHS was also varied (results not shown). As a result of these experiments, optimal cross-linking conditions for production of both monomer and dimer conjugates were chosen to be a 1:4 ratio of bFGF:dp12 and EDC/S-NHS concentrations of 6/15 mM, respectively. A typical distribution of reaction products is shown in Fig. 1B. The lack of cross-linked product when denatured bFGF was used (results not shown) ruled out the possibility of the process being the result of nonspecific random collision of active oligosaccharide-succinimide esters and primary amines on the growth factor.
Purification of bFGF⅐Oligosaccharide Conjugates-Products of the cross-linking reaction were separated by gel-filtration chromatography using two Superose 12 columns connected in series, as described under "Experimental Procedures," and a typical elution profile is shown in Fig. 2A. This clearly shows the conjugates (56 -61 min) eluting prior to the free oligosaccharide (62-65 min as identified previously by absorbance at 232 nm), and much earlier than the native growth factor (results not shown) (73-76 min for native bFGF disulfide dimer and 77-80 min for the native bFGF monomer). This early elution of bFGF conjugates (just prior to free dp12 oligosaccharide) may be due to the rigid extended nature of the oligosaccharide dominating the conjugate's hydrodynamic shape. In previous work (results not shown), we have noticed that, due to the strength of the binding of oligosaccharide to bFGF, noncross-linked complexes of bFGF and oligosaccharide also elute at a position just prior to the free oligosaccharide peak. This indicates that the cross-linked conjugates resemble at least in size/shape non-cross-linked dp12⅐bFGF oligosaccharide complexes. Indeed examination of the fractions by SDS-PAGE and ECL immunodetection, showed some contamination of the conjugates by both native bFGF disulfide dimer and monomer (Fig. 2B) as a result of the strong native bFGF⅐dp12 interaction described above. In order to finally remove all traces of free growth factor from the samples, fractions containing CL-dp12⅐bFGF dimers and CL-dp12⅐bFGF monomers were pooled separately and reapplied to the Superose 12 column in the presence of 2 M NaCl. This resulted in the recovery of conjugates with no free growth factor (by SDS-PAGE) or oligosaccharide contamination (as detected by absorbance at 232 nm results not shown). Fig. 3A shows a typical Superose 12 elution profile for the CL-dp12⅐bFGF monomer conjugate, with SDS-PAGE and ECL immunodetection analysis (Fig. 3B) confirming the absence of native growth factor. Purification of the CL-dp12⅐bFGF dimer conjugates was found to require several additional cycles of gel filtration in order to obtain an homogeneous preparation.
Structural Characterization of the CL-dp12⅐bFGF Dimer Conjugate-A number of possible structures were envisaged for the CL-dp12⅐bFGF dimer conjugates. These were: 1) individual cross-linking of two separate bFGF molecules to a single oligosaccharide, as may expected in the case of the dimer activation model which has been previously proposed (14,22,29,38,39); 2) disulfide-bridged dimers comprising one CL-dp12⅐bFGF monomer conjugate, and a molecule of native bFGF; and 3) disulfide-bridged dimers of two CL-dp12⅐bFGF monomers. To clarify this, we ran non-reduced and reduced samples of the CL-dp12⅐bFGF dimer conjugate on SDS-PAGE (Fig. 4). This clearly shows that the dimer product consists of disulfide-  4 -8). B, typical SDS-PAGE and ECL immunodetection analysis of the products from a preparative dp12⅐bFGF cross-linking reaction. Identity of bands is as indicated in the figure.

FIG. 2. Purification of cross-linking reaction products. A, typical gel-permeation elution profile of cross-linking reaction products.
Products of the two-step cross-linking reaction were applied to two Superose 12 columns attached in series and eluted in 50 mM phosphate buffer, 2.0 M NaCl, pH 7.4, at a flow rate of 0.5 ml/min. Elution was monitored by absorbance at 280 nM and 250-l fractions collected. Fractions containing protein were desalted by microdialysis and analyzed by SDS-PAGE. B, SDS-PAGE and ECL immunodetection analysis as described under "Experimental Procedures" of alternate fractions collected between 56 and 64 min. Fractions containing cross-linked CL-dp12⅐bFGF dimer and CL-dp12⅐bFGF monomer conjugates were pooled separately, freeze-dried, and stored at Ϫ80°C until required. bridged CL-dp12⅐bFGF monomers. Initial experiments also showed that small amounts of hybrid disulfide dimers were formed, which degraded to CL-dp12⅐bFGF monomers and unmodified bFGF monomers on reduction. However, more extensively purified dimer preparations such as that used in Fig. 4 were devoid of these hybrids. The fact that no true cross-linked dimer (two bFGFs to one dp12 oligosaccharide) is formed provides evidence that the binding of two bFGF molecules to a single oligosaccharide via identical growth factor binding sites is not an accurate model for the activation of signal transduction by HS⅐bFGF. As a result of this experiment, only the CL-dp12⅐bFGF monomer conjugate remained as a potential minimal mitogenic activation complex.
Monomer Conjugate Activity in F32 Cell Assay-The BaF3 cells are a lymphoblastoid line that do not express FGFRs and fail to respond to bFGF. These cells were transfected with FGFR1 to produce the F32 cell clone. The F32 cells, which are devoid of functional cell surface heparan sulfate proteoglycans, will respond to bFGF only in the presence of exogenous heparin or HS (14). Addition of purified CL-dp12⅐bFGF monomer conjugates to the assay system showed the complexes to be active ( Fig. 5), with activity ([ 3 H]thymidine incorporation) reaching a plateau at levels equivalent to maximum stimulation with intact HS and native bFGF. It was noted, however, that a concentration in excess of 100 ng/ml CL-dp12⅐bFGF monomer conjugate was needed to bring about maximum stimulation, with 50% maximum stimulation achieved at 20 ng/ml. In this assay the native bFGF and soluble dp12 oligosaccharide give maximum activity at a concentration of only 10 ng/ml bFGF. The lower activity of the conjugate is probably due to the presence of some oligosaccharides in the structurally heterogeneous dp12 mixture that bind bFGF strongly but fail to activate it (45). This would lead to a reduction in the effective concentration of active conjugate, as a proportion of the growth factor is irreversibly linked to non-activating oligosaccharides. This is in contrast to the native form of the growth factor, which can freely dissociate from non-activating oligosaccharides and further contribute to cell stimulation by binding active oligosaccharides. The higher amounts of conjugate needed to reach maximum stimulation may also indicate a level of specificity in terms of the exact positioning of the bFGF molecule along an activating oligosaccharide. However, the results do suggest that a bFGF⅐HS oligosaccharide monomer is a sufficient minimal structural unit for the initiation of mitogenic activity.
Analysis of the Affinity of Conjugates for HS/Heparin-Assessment of the binding of HS to the monomer conjugate is essential in terms of validating the monomer unit as being the minimum active structural unit of a HS⅐bFGF⅐FGFR signaling complex. Filter binding experiments (Fig. 6)  These data suggest that the oligosaccharide is covalently linked into the growth factor's HS binding site, resulting in the site being completely obscured from further HS interactions. DISCUSSION The ability of bFGF to bind to its high affinity FGFR is largely dependent on the binding of HS to the growth factor. This has led to the proposal of a "dual receptor system" for FIG. 3. Purification of CL-dp12⅐bFGF monomer conjugates. A, typical gel-permeation elution profile of the CL-dp12⅐bFGF monomer conjugates when reapplied to two Superose 12 columns attached in series and run as described in Fig. 2. B, SDS-PAGE and ECL immunodetection analysis as described under "Experimental Procedures" of CL-dp12⅐bFGF monomer containing fractions.

FIG. 5. [ 3 H]Thymidine uptake of F32 cells after stimulation
with CL-dp12⅐bFGF monomer conjugates. Cells were plated out at a density of 50,000 cells/well in RPMI medium, 10% horse serum. CL-dp12⅐bFGF monomer conjugates were added over a range of concentrations and incubated for 46 h before addition of [ 3 H]thymidine for another 2 h. Cells were harvested and incorporated radioactivity determined by scintillation counting. Controls for the experiment shown were 14,874 Ϯ 961 cpm for 10 ng/ml bFGF and 1 g/ml intact HS (determined optimal conditions for stimulation by bFGF) (positive control) and 5046 Ϯ 564 cpm for the cell suspension in the absence of bFGF and heparin (negative control).
signaling function (46). To date much is known about the interaction between HS and bFGF, i.e. the structural properties of HS such as the oligosaccharide length and sulfation pattern required to modulate FGF signaling. However, a major question still to be resolved is the exact molecular nature of the signaling complex comprising HS⅐bFGF⅐FGFR, and in particular the minimum stoichiometry of the active HS⅐bFGF complex. In this study we have addressed this problem by evaluating the biological activity of novel covalent conjugates of HS oligosaccharides and bFGF, which were prepared using a zero length cross-linking procedure. Our data showed that CL-dp12⅐bFGF monomer conjugates were biologically active and that any CL-dp12⅐bFGF dimers formed consisted only of disulfide-bonded CL-dp12⅐bFGF monomers.
The evidence that these monomer conjugates are a realistic representation of the native signaling configuration of bFGF and HS oligosaccharide is fourfold. First, it was shown that the CL-dp12⅐bFGF monomer conjugates were biologically active in a cell assay system that is devoid of functional heparan sulfate proteoglycans. Second, the formation of these types of zero length cross-links is dependent on the presence of an ion-pair interaction between a COO Ϫ group and a NH 2 group, in this case via the carboxylic acids of the oligosaccharides and amine groups of lysines present in the growth factor. Many studies in the past have shown the primary HS binding site of bFGF to consist of a number of lysine residues (22,(47)(48)(49)(50)(51), but a more recent report of the x-ray crystal structure of a heparin hexasaccharide bound to bFGF has shown that a COO Ϫ group of a IdceA unit is in close proximity (approximately 3 Å) to the nitrogen atom of the NH 2 group of lysine 136 of the growth factor's primary HS binding site (23). This distance is within the range for an ion-pair interaction to occur and is therefore a potential cross-linking site within the growth factor's HS binding site. Third, occupancy of this site is inferred from filter binding experiments which showed that the growth factor (after cross-linking with dp12 oligosaccharides), had no appreciable affinity for [ 3 H]HS chains. Finally, no conjugates were formed when cross-linking was performed using denatured protein, indicating that the reaction was specific for the native HS binding site(s) in bFGF and was not the result of random molecular collisions.
Several models have been proposed by which HS⅐bFGF and FGFRs interact to mediate signal transduction, with the most prevalent being the dimer presentation model. Initially the model simply suggested that biologically active HS oligosaccharides of at least 5-6 disaccharides in length can simultaneously bind two FGF molecules, hence presenting a pseudo-FGF dimer to the FGFR's (14,22,24,29,38,39,52), yet the evidence for its existence was initially circumstantial. However, in recent years, this dimer model has been more substantially investigated using molecular modeling, NMR, and x-ray crystallography techniques. Venkataraman and co-workers (53) suggested that bFGF preferentially self-associates to form true dimers and larger oligomers that were stabilized by a trisaccharide or heparin oligosaccharide, via two independent mechanisms. Oligomerization of bFGF by HS/heparin oligosaccharides was also implicated in a NMR and light scattering study, in which bFGF tetramers comprising two cis-oriented FGF dimers were formed in the presence of a decasaccharide (40).
Other growth factors such as platelet-derived growth factor, transforming growth factor-␤, and nerve growth factor have previously been shown to form homo-and heterodimers in order to facilitate receptor activation. In these examples, receptor dimerization occurs if each monomer unit binds to a separate growth factor receptor, so enabling two adjacent receptor molecules to associate (54). In the light of the wealth of physical and chemical evidence so far described, there now seems to be little doubt that HS/heparin chains or oligosaccharides can bring about oligomerization of bFGF under certain experimental conditions. Nevertheless, questions still remain as to the validity of such models in activity assays, as concentrations of growth factor and oligosaccharides used in biophysical studies, far exceed those present under cell culture or physiological conditions.
During the course of this study, the validity of this dimeric presentation model was put into question, as successful ␤-mercaptoethanol reduction of the CL-dp12⅐bFGF dimers formed during the cross-linking procedure showed them to comprise two disulfide-bonded CL-dp12⅐bFGF monomers. Not even small amounts of CL-dp12⅐bFGF dimers were formed through direct protein/oligosaccharide cross-linking between two bFGF molecules and a single oligosaccharide. The latter product would be expected if HS oligosaccharides were acting as a bridge between two bFGF molecules via the same protein/ oligosaccharide binding sites utilized in the formation of the CL-dp12⅐bFGF monomer conjugates. We believe that this result casts doubt on the likelihood of a dimeric bFGF presentation model mediating the growth factor's mitogenic activity. Furthermore, since we have shown that the CL-dp12⅐bFGF monomer conjugates have no measurable affinity for [ 3 H]HS chains, it is unlikely that any association of CL-dp12⅐bFGF monomers to form dimers or larger oligomers will involve further interactions with the dp12 oligosaccharide of the conjugate. It may be that dimerization of these monomer conjugates by mechanisms other than HS oligosaccharide bridging is occurring. However, a strong non-covalent association seems unlikely, as gel filtration analyses of these conjugates at physiological salt concentrations and under reducing conditions, showed they had no tendency for self-association (results not shown). In fact, as previously shown, the only oligomerization of these CL-dp12⅐bFGF monomer conjugates observed during this study, was the result of disulfide bonding. It seems plausible to discount this as a method for presenting dimeric conjugates to the receptor, as previous experimental data have shown that disulfide bonding of bFGF is not necessary for the growth factor's biological activity (55)(56)(57). In addition, we found that the formation of bFGF disulfide dimers was not influenced by the presence of HS oligosaccharides.
Additional models have also been proposed, both of which are based on data that indicated a direct interaction between HS and FGFR1 (25). This led to the suggestion that a short oligosaccharide (e.g. 5-6 disaccharides in length) contains two bind- FIG. 6. Filter binding analysis of the affinity of the CL-dp12⅐bFGF monomer conjugates for HS chains. Growth factor samples (4 g) were applied to a nitrocellulose filter in binding buffer 10 mM Tris-HCl, pH 7.3. The immobilized growth factor was then washed with 2 M NaCl to remove any non-cross-linked oligosaccharide and re-equilibrated in binding buffer. [ 3 H]HS was applied to the filters and washed with binding buffer; bound material was released by washing with 0.3 M and 2.0 M NaCl in binding buffer. Fractions (5 ml) were collected and radiolabeled material quantified by scintillation counting. The data shown are the total [ 3 H]HS (cpm) eluted at each NaCl concentration. Ⅺ, native bFGF; p, CL-dp12⅐bFGF monomers.
ing sites, one for bFGF and a second for the FGFR, with the oligosaccharide acting as a template for bringing the two proteins into close proximity to form a signaling unit. This has very strong similarities to the catalytic action of HS/heparin on the inactivation of thrombin by antithrombin III (58). Guimond and co-workers (31) provided further evidence to support this theory (that HS/heparin had to bind to a second protein) when, by the use of selectively desulfated heparins, they showed that C-6 O-sulfation was not required for bFGF binding, but was essential for mitogenic activity. These findings suggest a role for 6-O-sulfation in binding to a second protein, the most likely candidate being FGFR, with signaling occurring possibly via a single HS⅐bFGF⅐FGFR complex, or alternatively through oligomerization of these units. More recently, Pantoliano and co-workers (26,27) have suggested a mechanism akin to that involved in human growth hormone action (59,60), in which a monomeric complex of HS oligosaccharide and bFGF facilitates receptor dimerization through two FGFR binding interfaces on the growth factor. In essence, the principal theme of these latter two mechanisms is the monomer stoichiometry of the minimal active bFGF⅐HS complex, which is in agreement with our experimental evidence. It should perhaps be noted that it has previously been implied by Krufka and co-workers that HS/heparin may act along with bFGF to initiate different signaling pathways by employing either monomeric FGFR's or FGFR dimers/multimers (20). Indeed, their study suggested that FGFR dimerization may not be required for initiating a mitogenic response. Hence, both bFGF⅐HS oligosaccharide monomer models described above could be utilized in transducing different downstream signals.
In conclusion, our data suggest that the minimal bFGF⅐HS oligosaccharide activating unit for mitogenic stimulation is monomeric and not a dimeric presentation of bFGF mediated by a HS bridge. The accumulated data in this study tend to favor the two models proposed by Guimond et al. (31) and Pantoliano et al. (26), both of which have a bFGF⅐HS oligosaccharide monomer as their essential structural element, although ultimately their existence and possible roles in determining the final bFGF signaling events is yet to be determined. The present study provides the first direct demonstration that bioactive covalent complexes can be formed between HS and its dependent growth factors; such complexes will in the future facilitate further mechanistic studies and may also have considerable therapeutic potential.