Hepatocyte Growth Factor/Scatter Factor Binds to Small Heparin-derived Oligosaccharides and Stimulates the Proliferation of Human HaCaT Keratinocytes*

Hepatocyte growth factor/scatter factor (HGF/SF) acts via a dual receptor system consisting of the MET tyrosine kinase receptor and heparan sulfate or dermatan sulfate proteoglycans. In optical biosensor binding assays, competition by oligosaccharides for binding of HGF/SF to immobilized heparin showed that disaccharides failed to compete, whereas tetrasaccharides inhibited HGF/SF binding (IC50 8 μg/ml). The inhibitory potency of the oligosaccharides increased as their length increased by successive disaccharide units, to reach a maximum (IC50 1 μg/ml) at degree of polymerization (dp) 10. In binding assays, HGF/SF was found to bind directly to oligosaccharides as small as dp 4, and the binding parameters were similar for oligosaccharides of dp 4–14 (k a 2.2–45.3 × 106 m −1 s−1, k d 0.033–0.039 s−1, and K d 9–16 nm). In human keratinocytes, HGF/SF stimulated DNA synthesis, and this was dependent on a sustained phosphorylation of p42/44MAPK. In chlorate-treated and hence sulfated glycosaminoglycan-deficient HaCaT cells, the stimulation of DNA synthesis by HGF/SF was almost abolished. Heparin-derived oligosaccharides from dp 2 to dp 24 were added together with HGF/SF to chlorate-treated cells to determine the minimum size of oligosaccharides able to restore HGF/SF activity. At restricted concentrations of oligosaccharides (4 ng/ml), HGF/SF required decasaccharides, whereas at higher concentrations (100 ng/ml) even tetrasaccharides were able to partly restore DNA synthesis. The results suggest that HGF/SF binds to a tetrasaccharide and that although this is sufficient to enable the stimulation of DNA synthesis, longer oligosaccharides are more efficient, perhaps by virtue of their ability to bind more easily other molecules.

HGF/SF 1 is a well described heparan sulfate (HS) and dermatan sulfate (DS) binding growth factor with mitogenic, mor-phogenic, and motogenic activities toward many normal and neoplastic epithelial cells (1,2), as well as at least some stromal cells (3). In vivo, HGF/SF mediates epithelial-mesenchymal interactions, which are crucial for embryonic development, as well as tissue regeneration processes (4). By virtue of its motogenic and angiogenic activities, HGF/SF is involved in tumorigenesis and metastasis (5)(6)(7)(8). The diverse biological effects of HGF/SF are transduced by activation of its transmembrane receptor MET, encoded by the c-met protooncogene (1,9). Binding of HGF/SF, which is thought to induce MET dimerization and autophosphorylation, activates multiple signaling cascades (1). Two naturally occurring HGF/SF isoforms, which consist of the N-terminal domain and either the first kringle repeat (NK1) or the first two kringle repeats (NK2), bind MET and HS/heparin and function as agonists or antagonists of the full-length HGF/SF (10 -14).
The interactions of HGF/SF with HS and DS are of high affinity, with K d values ranging from 0.2 to 20 nM (15,16). By combining scission of HS and DS with sequence-specific enzymes and affinity chromatography on HGF/SF, a minimal binding sequence in HS was approximated to [IdoA-GlcNS(6-OSO 3 )] 3 (17) and in DS to [IdoA-GalNac(4-OSO 3 )] 3 (15). Although the absolute number and positioning of critical residues is unknown, these analyses demonstrate the critical importance of the iduronate residues themselves, because chondroitin sulfate, similarly sulfated to DS but lacking iduronate, fails to bind HGF/SF. Moreover, it is clear that both N-sulfation of hexosamine and 2-O-sulfation of iduronate plays no role in HGF/SF binding (15). Heparin was shown to be required for NK1 to increase the tyrosine phosphorylation of MET and downstream signaling in mutant Chinese hamster ovary cells devoid of HS and DS (11). However, in hematopoietic cells lacking HS, heparin potentiated the activity of HGF/SF and NK1 but was not absolutely required for HGF/SF (12,14). In contrast, the activity of HGF/SF is dependent on HS or on DS in mutant Chinese hamster ovary cells deficient in these GAGs (18). Moreover, in cell systems rendered deficient in sulfated GAG by treatment with chlorate, the mitogenic and motogenic responses to HGF/SF are strongly dependent on the presence of GAG (19,20). Heparin-binding sites have been identified in the N-terminal domain of HGF/SF, as well as in kringle domains 1 and 2 (11,(21)(22)(23)(24)(25).
Therefore, the balance of evidence favors a model in which the diverse biological activities of HGF/SF are dependent upon the presence of GAG to which it can bind. Oligomerization of NK1 in solution has been found to be promoted by heparin (11,12,26), and this has been suggested as a possible mechanism for the dependence of the cellular activities of HGF/SF on GAGs. In marked contrast, a recent study using 1:1 covalent complexes of HGF/SF and heparin/DS suggests that dimerization of HGF/SF by GAG is unlikely to be relevant (18). However, the structures in HS/DS that are required to support the biological activities of HGF/SF remain to be fully defined. To address this issue, we have measured quantitatively the interactions of HGF/SF with oligosaccharides of different lengths. These studies have been extended by examining the dependence of HGF/SF-stimulated cell proliferation on the length of heparin-derived oligosaccharides. Our results show that a tetrasaccharide is the minimal structural unit, which is recognized by HGF/SF, and that this interaction is not affected by increasing chain length. However, there is a relationship between oligosaccharide length and biological activity for HGF/ SF, whereas oligosaccharides of degree of polymerization (dp) 4 to dp 8 are equipotent, increasing the length of oligosaccharides by successive disaccharide units from dp 8 to 14 increases their biological potency. These results argue against a model in which the GAG acts to dimerize the HGF/SF and instead suggest the GAG may act by bridging HGF/SF and another molecule, such as MET.

EXPERIMENTAL PROCEDURES
Materials-Human recombinant HGF/SF was obtained from R & D Systems (Abingdon, UK). Heparin-derived oligosaccharides (dp 4 -26), prepared from partial heparinase I digests of pig mucosal heparin, were obtained from Iduron (Manchester, UK); the main disaccharide unit in these saccharides (Ͼ75%) being IdoA,2S-GlcNS,6S. Porcine intestinal mucosal heparin, trisulfated heparin disaccharide, and streptavidin were from Sigma. All reagents for electrophoresis were purchased from Bio-Rad. PD098059 and biotin-XX-hydrazide were from Calbiochem, and N-hydroxysuccinimide(LC)-biotin was from Pierce and Warriner (Chester, UK). The pan extracellular signal-regulated kinase antibodies, which recognize p42/44 MAPK regardless of their state of phosphorylation, the antibodies against the phosphorylated forms of p90 RSK , and the antibodies against the dually phosphorylated Thr(P) 183/202 / Tyr(P) 185/204 forms of p42/44 MAPK were purchased from New England Biolabs (Hitchin, UK). Secondary peroxidase-labeled anti-IgG antibodies were from Amersham Biosciences.
Measurement of DNA Synthesis-The immortalized human HaCaT keratinocytes (generously provided by Dr. N. Fusenig, Germany) were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum and the antibiotics penicillin (1000 units/ml) and streptomycin (1 mg/ml) (Invitrogen) (27). Cells were maintained in a humidified atmosphere of 95% air and 5% carbon dioxide at 37°C. HaCaT keratinocytes were plated into 24-well plates for 24 h. They were then washed twice in PBS, and the culture medium was replaced with Dulbecco's modified Eagle's medium supplemented with 250 g/ml bovine serum albumin. After 24 h, HGF/SF was added at 30 ng/ml. [methyl-3 H]Thymidine (40 Ci/ml, 0.8 M) was added directly to the culture medium 18 h later for 1 h. The cells were washed twice in PBS, and then macromolecules were precipitated at 4°C in 5% (w/v) trichloroacetic acid for at least 30 min, washed in 95% ethanol, air dried for 20 min, and finally lysed in 0.1 M NaOH for 30 min at 37°C. The radioactivity incorporated into DNA was measured by liquid scintillation counting.
Sulfated glycosaminoglycan-deficient HaCaT cells were prepared as described previously (20,28,29) for other cell lines. Cells were incubated for 4 h in sulfate-free Dulbecco's modified Eagle's medium and supplemented with 10% (v/v) dialyzed fetal calf serum and 15 mM NaClO 3 . Following trypsinization, the cells were seeded in 24-well plates in sulfate-free Dulbecco's modified Eagle's medium supplemented with 15 mM NaClO 3 for 24 h, as described for DNA synthesis assays, except that sulfate-free DMEM supplemented with 15 mM NaClO 3 was used throughout. As others have described (30 -33), cells can be kept viable in sulfate-free medium containing chlorate for up to 1 month. Checks were always made on cell death and protein synthesis.
Identification of Phosphorylated Forms of p42/44 MAPK and p90 RSK by Western Blotting-Cells were seeded at equal densities in 10-cm diameter cultured dishes and then treated identically as for the DNA synthesis assay up to the addition of growth factors. In some experiments an inhibitor of mitogen-activated protein kinase/extracellular signalregulated protein kinase kinase, PD098059 (2-(2Ј-amino-3Ј-methoxyphenyl)oxanaphtalalen-4-one) (34 -36) diluted in Me 2 SO was added directly to the medium at a concentration of 50 M and 15 min before the addition of the growth factors. At the times indicated in the figure legends, cells were washed twice with ice-cold PBS and were lysed in 300 l of lysis buffer (50 mM Tris-HCl, pH 6.8, 1% (w/v) SDS, 10% (v/v) glycerol, 0.006% bromphenol blue (w/v), 2% (v/v) ␤-mercaptoethanol, and protease inhibitor mixture), scraped with a rubber policeman, and collected in 1.5-ml Eppendorf tubes. All steps were performed at 4°C. Identical amounts of protein were separated by SDS-PAGE. After transfer to nitrocellulose, the membranes were blocked for 30 min in blotting solution (Tris-buffered saline containing 5% (w/v) nonfat dry milk; 0.05% (v/v) Tween 20). Incubation with the primary antibody diluted at 1:500 was carried out overnight at 4°C in the blotting solution. After 5 washes in Tris-buffered saline containing 0.05% (v/v) Tween 20, the nitrocellulose membrane was incubated with secondary peroxidase-conjugated antibodies to IgG antibodies, diluted 1:1000 in the blotting solution. Following several washes with Tris-buffered saline containing 0.05% (v/v) Tween 20, immunoreactive proteins were revealed with the SuperSignal chemiluminescent detection system (Pierce and Warriner, Chester, UK) on Hyperfilm (Amersham Biosciences).
Biotinylation of Oligosaccharides-Porcine intestinal mucosal heparin was biotinylated on amino groups with N-hydroxysuccinimide(LC)biotin, and free biotin was removed by gel filtration chromatography (37), whereas oligosaccharides were biotinylated at their reducing ends (38). Briefly, oligosaccharides (1 mg/ml in PBS) and biotin-XX-hydrazide (5 mM) were allowed to react for 2 h at room temperature. Unreacted biotin reagent was removed by ion exchange chromatography. For dp 4, 6, and 8, a Hypersil (250 ϫ 4.6 mm, 5 m pore; ThermoQuest, Runcorn, Cheshire, UK) strong anion exchange high pressure liquid chromatography column was used, and the oligosaccharides were eluted with a 1 M NaCl step. For oligosaccharides of dp Ն 10, a 0.5-ml column of DEAE-Sephacel was used, and the oligosaccharides were eluted with 1 ml of 1.5 M NaCl. Successful biotinylation of oligosaccharides was always confirmed by a dot-blot procedure.
Immobilization of Oligosaccharides-Streptavidin was immobilized on planar aminosilane surfaces according to the recommendations of the manufacturer (Affinity Sensors, Cambridge, UK). Biotinylated heparin was immobilized as described elsewhere (37). GAGs give a variable and poor response in optical biosensors (37,38). Therefore the amount of oligosaccharide immobilized was titrated indirectly by measuring the response to a concentration of HGF/SF above the K d value. Sufficient amounts of each size of oligosaccharide were immobilized to yield a response of around 85 arc s (600 arc s corresponds to 1 ng of protein/ mm 2 of the cuvette surface) when 24 nM HGF/SF was added. This is equivalent to less than 10% coverage of the sensor surface when more than half the oligosaccharides have bound HGF/SF. The distribution of the bound HGF/SF, and by inference the immobilized heparin or oligosaccharide on the surface of the biosensor cuvette, was inspected by examination of the resonance scan in the course of the association phase of binding reactions. These showed that at all times, HGF/SF was distributed uniformly on the sensor surface and therefore was not microaggregated (results not shown). Controls showed that in PBST (PBS supplemented with 0.02% (v/v) Tween 20), HGF/SF did not bind to this surface or to a streptavidin surface derivatized with biotin hydrazide (result not shown).
Competitive Binding Assays-In these experiments, we used cuvettes with heparin-derivatized aminosilane surfaces and a customized program for the IAsys Autoϩ instrument (available from the manufacturer, Affinity Sensors). The cuvette was equilibrated at 20°C in 40 l of PBST, and then 5 l of the relevant dilution of an oligosaccharide in PBST was added. Once the base line was stable, 5 l of HGF/SF was added to initiate the association phase. The binding reaction was continued for 5 min. The surface was then washed three times with 50 l of PBST and 2 min later was regenerated with 2 M NaCl. All experiments were performed with reagents stored in the instrument's reagent trays at 4°C. No change was detected in the amount of HGF/SF bound to immobilized heparin in the absence of competing polysaccharide at the start and the end of each series of the experiments (result not shown). The extent of binding was calculated by fitting the association curve to a single site binding model using the non-linear curve fitting FastFit software (Affinity Sensors).
Measurement of Binding Kinetics-A binding assay consisted of adding the HGF/SF, at a known concentration, in 5 l of PBST to a cuvette containing an oligosaccharide-derivatized aminosilane surface equilibrated in 45 l of PBST. The association reaction was followed over 210 s. The cuvette was then washed three times with 50 l of PBST, and the dissociation of bound ligate into the bulk PBST was followed over time. The surface was regenerated by washing twice with 50 l of 2 M NaCl, 10 mM phosphate, pH 7.2. Binding parameters were calculated from the association and dissociation phases of the binding reactions using the non-linear curve-fitting FastFit software (Affinity Sensors). A single binding assay yielded four binding parameters as follows: the initial rate of association, the on-rate constant (k on ), and the extent of binding, all calculated from the association phase, and the off-rate constant (k off equivalent to the dissociation rate constant, k d ), calculated from the dissociation phase. In these kinetic experiments, k on was only determined at low concentrations of HGF/SF, whereas k off was measured both at these lower concentrations of HGF/SF and, in separate experiments, using higher concentrations of HGF/SF and competing heparin (100 g/ml) in the dissociation buffer to avoid any rebinding artifacts (38,39). The equilibrium dissociation constant (K d ) was calculated from the ratio of the dissociation and association rate constants.
A single site binding model fitted the data at least as well as a two-site binding model in both the competitive binding assays and the kinetic experiments. Therefore the binding reaction between HGF/SF and the oligosaccharides was deemed to be monophasic, and a single site model was used to calculate all binding parameters.

RESULTS
Competition by Heparin-derived Oligosaccharides for HGF/SF Binding to Heparin-The ability of soluble oligosaccharides to compete with immobilized heparin for HGF/SF binding was determined in an optical biosensor. Disaccharides (dp 2) in solution failed to inhibit the binding of HGF/SF to the heparin immobilized on the biosensor cuvette ( Fig. 1), even at 333 g/ml (result not shown). Surprisingly, tetrasaccharides competed efficiently for HGF/SF binding to the immobilized heparin and inhibited 50% of the binding of HGF/SF (IC 50 ) at 8 g/ml (Fig. 1). The competition by oligosaccharides for HGF/SF binding to heparin exhibited a size-dependent gradation. The shortest competing oligosaccharides (dp 4) were the least potent (IC 50 8 g/ml), and as the oligosaccharides progressively increased in length by a disaccharide unit to dp 10, there was an increase in the potency of their inhibition of HGF/SF binding to heparin (dp 10 with an IC 50 of 1 g/ml). Heparin was slightly more potent (IC 50 0.7 ng/ml) than the decasaccharide.
Kinetics of HGF/SF Binding to Heparin-derived Oligosaccharides-The kinetics of the interaction of HGF/SF with re-ducing end biotinylated oligosaccharides immobilized on a streptavidin-derivatized cuvette were then determined in an optical biosensor (Fig. 2A). The binding of HGF/SF to immobilized tetrasaccharide ( Fig. 2A) was typical. Binding was detected at low concentrations of HGF/SF (0.4 nM), and the binding reaction was fast, since it reached a maximum extent within 2-3 min. Importantly, the binding of HGF/SF to tetrasaccharides was always monophasic, and there was no evidence for secondary binding sites. For example, plots of k on , the observed on-rate, determined using a one-site binding model, against concentration of HGF/SF were linear (Fig. 2B). The interaction of HGF/SF with the longer oligosaccharides was similarly monophasic (result not shown), and thus a one-site binding model was used for all data analyses. The association rate constant (k a ) of HGF/SF for tetrasaccharides was 3.6 ϫ 10 6 M Ϫ1 s Ϫ1 . Increasing the length from dp 4 to 14 did not change significantly the value of k a . The dissociation rate constant (k d ) was measured independently using a high concentration of HGF/SF (100 g/ml) and including heparin (100 g/ml) in the dissociation buffer to prevent re-binding (38). Overall, k d did not vary appreciably with the length of the oligosaccharides (Table I), and therefore, the equilibrium dissociation constants (K d ) for the interaction between HGF/SF and the oligosaccharides was similar and in the range of 9 -16 nM (Table I). Thus, the binding parameters of HGF/SF for the oligosaccharides are the same, even as the length of the oligosaccharides is increased from dp 4 to 14, which suggests that tetrasaccharides represent the binding site of HGF/SF in heparin/HS. However, because only larger natural (18) and semisynthetic (40) oligosaccharides (dp 6) have previously been found to be active, it was important to determine whether such small oligosaccharides functioned as agonists or antagonists of HGF/SF-stimulated cell proliferation. Effect

FIG. 2. Binding interactions of HGF/SF to heparin-derived tetrasaccharides.
The binding kinetics of HGF/SF to immobilized reducing end biotinylated tetrasaccharides was measured as described under "Experimental Procedures." A, representative set of binding interactions of HGF/SF to tetrasaccharides. B, plot of the k on , calculated by non-linear regression from the curves in A using a one-site binding model, against concentration of HGF/SF obtained with dp 4 immobilized cuvette.
immortalized keratinocytes that maintain the ability to differentiate in culture (27,(41)(42)(43). HGF/SF strongly stimulated DNA synthesis in these cells (Fig. 3A), which possess the MET tyrosine kinase receptor and heparan sulfate (data not shown). Addition of PD098059, a well established inhibitor of mitogenactivated protein kinase/extracellular signal-regulated protein kinase kinase-1, 15 min before the addition of HGF/SF reduced DNA synthesis to the level seen in untreated cells. Moreover, after addition of PD098059, a near complete inhibition of the stimulation of dual phosphorylation of p42/44 MAPK on Thr 183/202 /Tyr 185/204 in response to HGF/SF was observed by Western blotting (Fig. 3B). The kinetics of phosphorylation of p42/44 MAPK and of p90 RSK , a downstream target of p42/44 MAPK , were then examined (Fig. 3C). In HaCaT keratinocytes, the dual phosphorylation of p42/44 MAPK was stimulated 5 min after the addition of 30 ng/ml HGF/SF and reached a maximum level at 10 min. The level of phosphorylation decreased slightly by 30 min, and the level of phosphorylation was then maintained as a plateau until at least 60 min (Fig. 3C). Phosphorylation of p90 RSK was observed within 10 min of addition of HGF/SF and increased slowly to reach a maximum after 30 min and then slightly decreased to a plateau maintained to the end of the experiment. Thus, as found in other epithelial cells (19,20), the proliferation of HaCaT cells is stimulated by HGF/SF through a mechanism dependent, at least in part, on p42/ 44 MAPK and its downstream targets.
Effect of Chlorate Treatment of HaCat Cells on HGF/SFstimulated Cell Proliferation-Chlorate is a potent inhibitor of sulfation, widely used to address sulfated GAG function (19, 20, 28 -32, 44). When HaCaT keratinocytes are grown in the presence of 15 mM sodium chlorate, DNA synthesis induced by HGF/SF is strongly reduced (Fig. 4). Chlorate itself did not affect the intrinsic ability of the cells to trigger a growthstimulatory response, because dialyzed serum stimulated DNA synthesis in chlorate-treated HaCaT cells (Fig. 4). Moreover, addition of 7.5 mM Na 2 SO 4 , which will relieve the inhibition of 3Ј-phosphoadenosine 5Ј-phosphosulfate synthesis by 15 mM chlorate (30,31) and hence enable the synthesis of sulfated glycosaminoglycan chains by the cells, restored the growthstimulatory effect of HGF/SF (Fig. 4). The addition of soluble heparin (10 ng/ml) simultaneously with HGF/SF also restored the growth-stimulatory effects of HGF/SF in chlorate-treated keratinocytes (Fig. 4), whereas heparin alone had no effect on DNA synthesis (result not shown). Therefore, in HaCaT keratinocytes, HGF/SF required the presence of heparan sulfate receptor to fully trigger a proliferative response.
Effect of Heparin-derived Oligosaccharides on HGF/SF Stimulation of DNA Synthesis-The stimulation of proliferation of HaCaT by HGF/SF is dependent on sulfated GAGs, and the growth stimulatory activity of HGF/SF in chlorate-treated Ha-CaT cells is restored by the addition of soluble heparin. Thus, these cells provide a sensitive assay for studying the dependence of the stimulation of cell proliferation by HGF/SF on GAGs. Heparin-derived oligosaccharides from dp 2 to 24 were used with chlorate-treated cells to determine the minimum size of oligosaccharides able to restore HGF/SF-stimulated DNA synthesis (Fig. 5). The trisulfated disaccharide was without effect at all concentrations tested (Fig. 5). In contrast, tetrasaccharides were able to restore partly the growth stimulatory activity of HGF/SF, and their ED 50 was ϳ100 ng/ml (Fig. 5). a The S.E. is derived from the deviation of the data from a one-site binding model, calculated by matrix inversion using the FastFit software provided with the instrument (see "Experimental Procedures"). No evidence was found for a two-site model of association, and so the HGF/SFbinding sites in each oligosaccharide were homogeneous in this respect. Four independent sets of k on were measured, and the four resulting values for k a and their errors were combined.
b The correlation coefficient of the linear regression through the k on values. c The k d is the mean Ϯ S.E. of at least six values. No evidence was found for a two-site model of dissociation, and so the HGF/SF-binding sites of oligosaccharide were homogeneous in this respect. d The K d value was calculated from the ratio of k d /k a , and the S.E. is the combined S.E. of the two kinetic parameters. Oligosaccharides of dp 6 and 8 had the same potency in this assay as the tetrasaccharides. However, increasing the length of the oligosaccharide by successive disaccharide units to dp 14 resulted in a substantial increase in potency, with an ED 50 of 20 ng/ml for dp 10, 4 ng/ml for dp 12, and around 1 ng/ml for dp Ն 14, which was the maximal potency observed in this assay (Fig. 5). DISCUSSION HGF/SF binds to two distinct types of receptors, a transmembrane tyrosine kinase, MET, which corresponds to the product of the c-met protooncogene, and proteoglycans bearing chains of the glycosaminoglycans HS or DS. HGF/SF, like many GAGbinding growth factors, must interact with both types of receptor to promote a cellular response (2, 11, 18 -20, 45). However, the mechanism by which the HS receptors or the more recently described DS receptors contribute to the delivery of growthstimulatory signals by HGF/SF is unclear. To determine some of the limiting structural features in HS/heparin that enable HGF/SF to stimulate cell proliferation, we used optical biosensor-based binding assays to analyze the interaction between HGF/SF and either soluble or reducing end immobilized oligosaccharides in competition and direct binding studies, respectively. The competition assays show that tetrasaccharides contain all the necessary information required for HGF/SF binding, because they clearly inhibit the binding of HGF/SF to heparin. However, as the length of the oligosaccharide increases by successive disaccharide units to dp 10, the efficiency of inhibition of HGF/SF binding to heparin increases. The measurement of the kinetics of HGF/SF binding to immobilized oligosaccharides provides a quantitative analysis of these interactions. Surprisingly, in contrast to the competition experiment (Fig. 1), the results (Table I) show that there is no difference in the association rate constant, k a , the dissociation rate constant, k d , and the equilibrium dissociation constant, K d , of HGF/SF for the oligosaccharides of dp 4 -14. It seems unlikely that the longer oligosaccharides, which are competing more effectively for HGF/SF, are able to bind more than one molecule of HGF/SF, because the steric hindrance attributable to multivalent binding would have resulted in biphasic binding kinetics at higher concentrations of HGF/SF (38). An explanation for the higher efficiency of HGF/SF binding by the longer oligosaccharides in the competition assay is that they may present, in solution, more opportunities for a collision with HGF/SF to be productive and result in a binding event. However, when the same oligosaccharides are immobilized on the planar surface of the biosensor, they all have the same orientation, and so this effect of length would no longer be apparent. Moreover, because the HGF/SF binding kinetics do not vary with the length of the oligosaccharides, this suggests that the binding site of HGF/SF in heparin is equivalent to the shortest oligosaccharide, dp 4. However, the non-reducing terminal hexuronic acid of the oligosaccharides will be 4,5-unsaturated because of the action of heparinase I, whereas the biotinylation reaction will cause the reducing terminal GlcN unit to lose its ring structure. Therefore, only part of the oligosaccharides, including the minimal tetrasaccharide, used in this work represent the native structure.
To gain insights into the relationship of oligosaccharide length and biological potency, we then developed a model based on human HaCaT keratinocytes. The proliferation of these cells is stimulated by HGF/SF (Fig. 3). In the presence of chlorate, which inhibits sulfation on proteins and on carbohydrate residues in intact cells without inhibiting cell growth or protein synthesis (29 -32), the growth-stimulatory response of HaCaT cells to HGF/SF is strongly reduced. Moreover, the addition of soluble heparin restores the growth stimulatory activity of HGF/SF. Thus, as observed in other cell types (18 -20, 40), the cellular response to HGF/SF in HaCaT keratinocytes depends on the presence of sulfated glycosaminoglycans such as HS and DS. Interestingly, in some other epithelial cell systems, the exogenously added HS must be anchored to the substratum to restore the cellular response to HGF/SF (19,20). The observation that soluble heparin is functional in the present assay makes HaCaT cells a more tractable model for studying HGF/ SF-dependent structure-function relationships in GAGs.
When soluble oligosaccharides of different lengths were tested for their ability to restore the growth stimulatory activity of HGF/SF in chlorate-treated HaCaT cells, we observed that tetrasaccharides were the shortest oligosaccharides active in this assay. Previous work has suggested that longer oligosaccharides are required. For example, it has been reported that the stimulation of migration of mutant Chinese hamster ovary cells devoid of HS and DS requires oligosaccharides of at least dp 6 (18). In addition, semisynthetic sulfated oligosaccharides of dp 6, but not dp 4, potentiated the activity of HGF/SF in cells containing GAGs (40). The ED 50 values of oligosaccharides of dp 4, 6, and 8 were similar in our study, but between dp 8 and 14 there was a clear increase in potency associated with the increasing length of the oligosaccharides. Therefore, at the sub-optimal concentrations of oligosaccharides used in this as- say, below dp 14 the length of the oligosaccharide becomes limiting with respect to the ability of HGF/SF to stimulate cell proliferation.
The structure of HGF/SF-binding sites in HS and DS have been partially elucidated. In HS, HGF/SF interacts within the S domains (17), and binding appeared to require at least a hexasaccharide by affinity chromatography (17,46). A combination of iduronates and 6-O-sulfation appears to be critical, although not N-sulfation (17) or 2-O-sulfation (47). The minimal binding sequence in HS was approximated to [IdoA-GlcNS(6-OSO 3 )] 3 , although the absolute number and positioning of critical residues are unknown. The analysis of the binding site of HGF/SF in DS has confirmed that N-sulfation of hexosamine and 2-Osulfation of iduronate play no role in HGF/SF binding (15). However, the latter study underlines the critical importance of the iduronate residues themselves, because chondroitin sulfate, similarly sulfated to DS but lacking iduronate, fails to bind HGF/SF. Thus the minimum binding sequence for HGF/SF in DS is likely to be [IdoA-GalNac(4-OSO 3 )] 3 . The question arises as to why the present quantitative measurements show that the HGF/SF-binding site in heparin is a tetrasaccharide. The 6-O-sulfate groups in HS are thought to be essential for HGF/SF binding (17). Sequencing of heparinase III-released S domains from 3T3 cell-derived HS has revealed that they are sparingly 6-O-sulfated, and a significant number are completely deficient in 6-O-sulfate groups. For example, only 12% of sequenced hexasaccharides and 29% of sequenced octasaccharides contained a 6-O-sulfate group (48). Therefore, it is likely that oligosaccharides derived from HS that bind HGF/SF will be longer than dp 4 to contain sufficient 6-O-sulfate groups. Moreover, a recent study (49) on a novel HGF/SFbinding PG isolated from endothelial cells, endocan, shows that it carries DS chains with a small number (1-2) of iduronates per chain. Although the sequential position of these iduronates is unknown, this observation supports the contention that HGF/SF may only require a very short sequence of iduronates possibly in oligosaccharides as small as dp 4.
A recent report (22) established that heparin binding does not induce a conformational change in NK1, but, as suggested by previous work (12,26), heparin promotes the formation of the NK1 dimer. However, such studies employ a subdomain of HGF/SF in isolation. The present results show that heparinderived oligosaccharides as short as dp 4 are able to replace cellular HS in terms of enabling HGF/SF to stimulate keratinocyte proliferation. Because the binding parameters of HGF/SF to the oligosaccharides of different length are the same, this suggests that a tetrasaccharide is the unit that is recognized by HGF/SF. Given the large size of HGF/SF (84 kDa, ϳ7 nm diameter for a globular protein) compared with other HS-binding growth factors, e.g. fibroblast growth factor-2 (18 kDa, ϳ4.5 nm diameter), it seems unlikely that a tetrasaccharide (length ϳ1.6 nm) can support dimerization of HGF/SF in cis mode. However, the observation that increasing the length of oligosaccharides by successive disaccharide units from dp 8 to 14 (length ϳ5.6 nm) increases their biological potency could be interpreted as representing the more efficient dimerization of HGF/SF by the oligosaccharide. There is also the possibility of HGF/SF dimerization in trans mode. However, a recent study using biologically active 1:1 covalent monomeric complexes of HGF/SF with HS/DS oligosaccharides of dp 8 -12 indicates that dimerization per se is unlikely to play a significant role in the mechanism whereby these GAGs enable the cellular response to HGF/SF (18). The binding of HGF/SF to the longer oligosaccharides (dp 10 and 14) was monophasic; if such oligosaccharides could support HGF/SF dimerization, biphasic kinetics would have been expected at the higher concentrations of HGF/SF due to steric hindrance (38). Taken together, these results suggest that, whereas small oligosaccharides can display activity, the maximal attainable biological activity of oligosaccharides may require an interaction with HGF/SF and another protein, which is facilitated when dp Ͼ8. One suggestion for the identity of the other protein is MET itself, which has been shown to bind to heparin (14,22), although another report queries whether the interaction of the native MET extracellular domain is of significant affinity by itself (18). However, given the high affinity of the interaction between HGF/SF and GAG, putative additional GAG-MET interactions would not need to be of great affinity. The structural features in HS required for this interaction and whether these requirements are compatible with those necessary for HGF/SF binding are currently unknown.