Heparin binding and oligomerization of hepatocyte growth factor/scatter factor isoforms. Heparan sulfate glycosaminoglycan requirement for Met binding and signaling.

Hepatocyte growth factor/scatter factor (HGF/SF) is a heparin-binding polypeptide that stimulates cell proliferation, motility, and morphogenesis by activation of its receptor, the c-Met tyrosine kinase. HGF/SF consists of a series of structural units, including an amino-terminal segment with a hairpin loop, four kringle domains, and a serine protease-like region. In this study, we demonstrate that the amino-terminal (N) domain retains the heparin-binding properties of full-length HGF/SF. In contrast to a previous hypothesis, selected basic amino acid residues in the hairpin loop are not critical for heparin binding, although alanine substitution at a subset of these sites markedly reduced the biological activity of the HGF/SF isoform, HGF/NK1. Covalent cross-linking experiments performed with wild-type and heparan sulfate glycosaminoglycan (HSGAG)-deficient Chinese hamster ovary (CHO) cells revealed that Met-HGF/NK1 binding was strongly dependent on HSGAG. Addition of heparin to HSGAG-deficient CHO cells not only restored ligand binding, but also increased ligand-dependent Met tyrosine phosphorylation and c-fos expression. Moreover, our results showed that heparin stimulated ligand oligomerization through an interaction with the N domain. These findings establish the importance of the N domain for heparin-ligand and ligand-ligand interactions, and demonstrate a crucial role for HSGAG in receptor binding and signal transduction.

During the past dozen years, several mitogens have been identified that possess the distinctive property of binding to heparin or closely related heparan sulfate glycosaminoglycan (HSGAG) 1 (1,2). Heparin-based affinity chromatography has been a convenient tool for purification of these factors, and several studies have shown that this binding phenomenon has functional relevance as well (2)(3)(4)(5)(6)(7)(8)(9)(10). HSGAG-dependent stabilization (3) and/or localization of the growth factor on the cell surface may promote interaction with less abundant, higher affinity tyrosine kinase receptors involved in signal transduction (2,4,5). Coupling of heparin or HSGAG either to the growth factor or its tyrosine kinase receptor may induce conformational changes in these molecules that augment signaling (3,6). Several laboratories have reported that heparin binding can provide a framework for ligand oligomerization, which may enhance signaling by stimulating dimerization of the tyrosine kinase receptor (5, 8 -10).
Hepatocyte growth factor/scatter factor (HGF/SF) is a heparin-binding polypeptide that can function as a mitogen, motogen, or morphogen on a broad spectrum of cellular targets (11)(12)(13). HGF/SF is related to plasminogen and macrophagestimulating protein, sharing with these molecules approximately 40 -45% amino acid sequence identity and several structural motifs (14,15). HGF/SF is synthesized as an inactive monomer which undergoes internal proteolysis to yield a biologically active, disulfide-linked heterodimer (16 -18). The heavy chain of the HGF/SF dimer (ϳ60 kDa) is derived from the amino terminus of the precursor and contains an aminoterminal segment followed by four kringle domains. A kringle consists of ϳ80 amino acids and has a characteristic folding pattern defined by three internal disulfide bonds and additional conserved residues (19). Two alternative transcripts have been identified that encode truncated variants of HGF/ SF, terminating after either the first or second kringle domain (20 -23). These smaller isoforms, designated HGF/NK1 and HGF/NK2, respectively, behave as antagonists or partial agonists, depending on the molecule and assay conditions (20,(22)(23)(24)(25). All of these HGF/SF isoforms bind with high affinity to the tyrosine kinase transmembrane HGF/SF receptor, the Met proto-oncogene product (20,22,23,26,27).
As with other heparin-binding growth factors, the higher capacity, lower affinity cell surface binding sites for HGF/SF were attributed to heparan sulfate proteoglycan (27)(28)(29)(30). Based on deletion analysis, kringle 2 (K2) and the hairpin loop in the amino-terminal (N) domain were implicated in heparin binding (31,32), although recent work raised doubt about the importance of K2 in this regard (23). The potential significance of HGF/SF-heparin interactions has been inferred from a variety of observations. HGF/SF-Met covalent cross-linking was diminished following treatment of primary hepatocytes with heparitinase, indicating that HSGAG might be required for binding (33). Administration of heparin to cells altered the biological response to HGF/SF, although the impact on DNA synthesis was either positive or negative depending on the assay system (34 -37). Recent evidence indicated that heparin stimulated dimerization of HGF/SF and HGF/NK1 in a cell-free setting, raising the possibility that a similar effect on the cell surface could facilitate Met dimerization and activation (25,37) as suggested for FGF-1 and FGF-2 (5,9,10).
In the present study, we have investigated the hypothesis that heparin binding of HGF/SF isoforms is dependent on K2 and specific basic amino acid residues in the hairpin loop of the N domain. We also provide new evidence that HSGAG is required for the binding of HGF/SF variants to Met and subsequent signal transduction. Finally, we demonstrate that heparin stimulates the oligomerization of HGF/SF isoforms in both cell-free and cell culture models, and that the N domain has an important role in this process.
Recombinant HGF/SF and Various HGF/SF Derivatives-Recombinant human HGF/SF was produced using a baculovirus expression system and purified as described previously (23). Recombinant human HGF/NK1 and HGF/NK2, as well as the individual N and K1 domains, were expressed in bacteria as described. 2 In brief, the DNAs encoding the various HGF/SF derivatives were generated as NdeI-BamHI fragments using the polymerase chain reaction (41). These DNA fragments were then cloned into expression vector pET11a, the Escherichia coli expression system developed by Studier (42). Site-directed mutants, designated 2A/NK1, 3A/NK1, and 5A/NK1, with alanine substitution in the NH 2 -terminal domain of HGF/NK1 were prepared using the technique of Higuchi et al. (43). The following positively charged residues were replaced with alanine: 2A/NK1, Arg-76 and Lys-78; 3A/NK1, Lys-91, Arg-93, and Lys-94; 5A/NK1, combined set of five. The mutant proteins were expressed in E. coli, refolded, and purified as described for wild type HGF/NK1. 2 Heparin-Sepharose Chromatography-HGF/SF and its derivatives (each at 1 mg/ml) were individually dissolved in phosphate-buffered saline (PBS), and 0.1 ml was applied to heparin-Sepharose CL-6B resin (0.3 ml bed volume, Pharmacia Biotech Inc.) equilibrated in PBS. Protein was eluted with a stepwise gradient of increasing NaCl concentration (1.5 ml/step, as indicated) and detected by immunoblotting.
Immunoblotting-For detection of HGF/SF derivatives eluted from heparin-Sepharose, 10 l of each fraction was resolved by SDS-PAGE in a 4 -20% polyacrylamide mini-gel (Novex) under reducing conditions. Immunoblotting was performed as described previously (44), with the following modifications: proteins were transferred to Immobilon-P (polyvinylidene fluoride) filters (Millipore), and probed with a mixture of GammaBind (Pharmacia Biotech Inc.)-purified rabbit polyclonal antisera raised against either full-length human HGF/SF (36) or a synthetic peptide corresponding to HGF/SF residues 158 -172 (numbering according to Ref. 36).
Met was detected by sequential immunoprecipitation and immunoblotting of protein isolated from equivalent amounts of whole cell lysates of serum-starved BALB/MK, CHO-WT, and CHO-745 cells prepared essentially as described previously (26). Samples were incubated with anti-mouse Met polyclonal antiserum (1 g/ml; Santa Cruz Biotechnology Inc., catalog no. sc-162) for 2 h at 4°C. After immunoprecipitation, SDS-PAGE and transfer to Immobilon filters, blotting was performed with anti-mouse Met polyclonal antiserum (0.5 g/ml; Santa Cruz, catalog no. sc-162).
Biological Assays-DNA synthesis by BALB/MK cells was measured by [ 3 H]thymidine incorporation as described (45), except that IGF-I (200 ng/ml; Pepro Tech Inc., catalog no. 100-11) was included in the assay medium. The scatter assay was performed with a subclone of MDCK cells, according to published methods (46). After 20 h, the cells were fixed for 20 min with 2.5% formaldehyde, and for an additional 20 min with methanol/formaldehyde/phosphate-buffered saline (7/2/1) prior to staining with Giemsa (pH 6.8).
HGF/NK1 Radioiodination and Covalent Affinity Cross-linking-Recombinant human HGF/NK1 was radiolabeled with [ 125 I]NaI (Amersham) by the chloramine-T method as described (26,44). Reactions performed with 10 g of protein and 1 mCi of [ 125 I]NaI yielded tracer with a specific activity of ϳ65 Ci/g.
Cross-linking was performed as described (44), with the following modifications. Confluent serum-starved monolayers were incubated for 45 min at room temperature in HEPES binding buffer (44) containing 0.9 nM of 125 I-HGF/NK1 (6.6 ϫ 10 6 cpm) in the presence or absence of heparin (Fisher, catalog no. H-19), unlabeled HGF/SF isoforms or IGF-I at the indicated concentrations. Cells were washed twice with HEPESbuffered saline (10 mM HEPES, 150 mM NaCl, pH 7.4) and then incubated for 16 min at room temperature with 0.1 mM disuccinimidyl suberate (Pierce) dissolved in dimethyl sulfoxide (Me 2 SO). The crosslinking reaction was terminated with quench buffer (200 mM glycine, 10 mM Tris-HCl, 2 mM EDTA, pH 7.4). Cell lysates, clarified by centrifugation (14,000 ϫ g, 15 min), were boiled in sample buffer containing 2% ␤-mercaptoethanol and resolved by 6% SDS-PAGE. Alternatively, Met protein was immunoprecipitated from cell lysates as described above, in the absence or presence of 10 g/ml of competing peptide (Santa Cruz, catalog no. sc-162 P). Following SDS-PAGE, radiolabeled cross-linked complexes were detected by autoradiography of dried gels exposed to Kodak X-Omat AR film at Ϫ70°C.
Cell-free Cross-linking/Oligomerization Analysis-Experiments were performed according to a protocol similar to one described in Ref. 25. 1 g HGF/NK1, N, or K1 domain was incubated in 20 l of PBS for 20 min at room temperature with or without cross-linking reagent bis(sulfosuccinimidyl) suberate (1 mM); Pierce) in the presence or absence of heparin (1 g). Next, 5 l of 1 M Tris-HCl, pH 6.8, was added, followed by 6 l of 5 ϫ SDS-sample buffer containing 10% ␤-mercaptoethanol. After boiling for 3 min, samples were resolved by SDS-PAGE in a 4 -20% mini-gel (Novex). Protein was detected by staining with colloidal Coomassie Blue (Novex).

N Domain Retains Heparin-binding Properties of Full-length
HGF/SF-To analyze the structural features of HGF/SF responsible for its interaction with heparin, we determined the NaCl concentration required to elute various HGF/SF isoforms and derivatives from heparin-Sepharose resin. Immunoblotting of fractions from each step in the salt gradient revealed that HGF/SF and the naturally occurring variants, HGF/NK1 and HGF/NK2, all were released from the resin with 0.8 and 1.0 M NaCl (Fig. 1). To further localize the heparin-binding site of HGF/SF, we produced the N and K1 domains separately in a bacterial expression system, and purified them to homogeneity. Analysis of their circular dichroism spectra suggested that the recombinant domains attained their native conformation. 2 When subjected to heparin-Sepharose chromatography, the K1 domain eluted at Յ0.2 M NaCl, while the N domain eluted at the same concentration as HGF/SF (Fig. 1). These data suggest that the determinants of HGF/SF heparin binding are confined to the N domain.

Modification of Basic Residues in the Hairpin Loop Has Only a Slight Effect on Heparin Binding, but a Significant Impact on
Biological Activity-In their investigation of the sites required for heparin binding, Mizuno et al. (32) observed that deletion of the hairpin loop region (amino acid residues 70 -96) in the N domain resulted in loss of heparin binding. After comparing the heparin-binding properties and hairpin loop sequences of HGF/SF and the structurally related molecules, plasminogen (14) and macrophage-stimulating protein (15), they noted a correlation between the net charge of the hairpin region and avidity for heparin. Specifically, the net charge of this region in HGF/SF was ϩ6 and the protein eluted from heparin-Sepha-rose with ϳ1.0 M NaCl, while the corresponding values of these parameters for plasminogen were Ϫ2.5 and ϳ 0.1 M NaCl, and for macrophage-stimulating protein were ϩ2 and ϳ 0.4 M NaCl. This led to the suggestion that clusters of basic amino acid residues in the HGF/SF hairpin loop mediated the interaction with heparin. We tested this hypothesis directly by substituting alanine for these residues in a series of site-directed mutants. As illustrated in Fig. 2, HGF/NK1 derivatives were generated that had 2, 3, or 5 alanine substitutions, as follows: R76A and K78A (2A/NK1), K91A, R93A, and K94A (3A/NK1), or a combined set (5A/NK1). When these reagents were chromatographed on heparin-Sepharose, they all bound tightly to the resin and eluted either with 0.6 -0.8 M (2A/NK1 and 5A/ NK1) or 0.8 -1.0 M (3A/NK1) NaCl (Fig. 1). Although replacement of R76 and K78 resulted in slightly earlier elution of protein, our findings indicated that the five basic residues did not have a critical role in heparin binding of HGF/SF isoforms.
The effect of alanine substitution on HGF/SF biological activity was evaluated by testing the 2A, 3A, and 5A/NK1 derivatives in a [ 3 H]thymidine incorporation bioassay using BALB/MK cells (Fig. 3A). 5A/NK1 was approximately 30-fold less potent than wild-type HGF/NK1 in this assay. This decrease in activity could not be attributed to the small reduction in heparin affinity associated with this molecule, because 2A/ NK1, which exhibited a comparable decline in affinity to heparin-Sepharose, was as active as wild type HGF/NK1. Moreover, the derivative that most closely retained the heparinbinding characteristics of HGF/SF, 3A/NK1, was markedly less potent than HGF/NK1 in the mitogenic bioassay. A similar pattern of relative potency was observed with these reagents in the standard scatter assay using MDCK cells (Fig. 3B). At a concentration where wild type HGF/NK1 and 2A/NK1 elicited a marked scatter response, 3A/NK1 and 5A/NK1 stimulated only modest cellular retraction and fewer signs of the spiky morphology characteristic of the scatter phenomenon (46). Preliminary experiments suggest that the diminished activity of 3A/NK1 and 5A/NK1 may result from reduced stability of these derivatives during overnight incubation at 37°C. 3 Nonetheless, regardless of the loss in biological activity, alanine substitution of the HGF/SF-specific basic residues in the hairpin loop had little impact on heparin binding.
Proteoglycan Is Important for Binding of HGF/NK1 to Met-In addition to characterizing the structural elements responsible for heparin binding of HGF/SF isoforms, we also explored the relevance of the heparin interaction for ligand binding to Met. To address this issue, we performed a series of experiments with wild type and glycosaminoglycan-deficient CHO cells (CHO-745). The latter are lacking in enzymatic activity required for synthesis of heparan sulfate (40), and have been used extensively to analyze the role of HSGAG in FGF- receptor binding (7,9,50). Using an antiserum raised against a carboxyl-terminal peptide sequence of murine Met, our initial studies established that a 145-kDa protein band corresponding to the Met ␤-subunit in BALB/MK cells was detected in CHO cells (Fig. 4). Moreover, the wild type and glycosaminoglycandeficient CHO strains expressed similar levels of hamster Met protein.
Covalent cross-linking experiments performed with 125 I-HGF/NK1 and BALB/MK or CHO cells revealed a high molecular mass band with an estimated mass of 165 kDa in lysates of BALB/MK and wild type CHO cells (Fig. 5A). However, no band was observed in lysates from the glycosaminoglycan-deficient CHO-745 cells. The ligand specificity of this binding interaction was verified by competition with an excess of unlabeled HGF/NK1 but not IGF-I (Fig. 5A). Immunoprecipitation of the cross-linked complex with mouse Met peptide antiserum in the absence but not presence of peptide confirmed that Met was in the complex (Fig. 5B). To demonstrate that the lack of HGF/NK1-Met cross-linking in CHO-745 cells was due to a deficiency in HSGAG, the experiment also was performed in the presence of varying concentrations of added soluble heparin. Consistent with earlier observations (33), when heparin was added to cells such as BALB/MK and wild type CHO, which have an ample supply of HSGAG on their surfaces, cross-linking of ligand to Met was diminished (Fig. 5A). In contrast, heparin at 0.3 g/ml (and to a lesser extent at 3 g/ml) facilitated cross-linking of 125 I-HGF/NK1 to Met in the glycosaminoglycan-deficient CHO-745 cells. These results established that the shortage of endogenous HSGAG in CHO-745 cells was specifically responsible for the paucity of their HGF/ NK1-Met binding.
Signal Transduction Is Enhanced by Addition of Heparin to Glycosaminoglycan-deficient Cells-To explore the functional significance of HGF/NK1 binding to Met observed in response to exogenous heparin, we measured tyrosine phosphorylation of Met following treatment of CHO cells with HGF/NK1 in the presence or absence of heparin. As shown in Fig. 6, incubation of wild type CHO cells with HGF/NK1 for 5 min resulted in a prominent Met protein band in the anti-phosphotyrosine immunoprecipitate. When heparin also was present in the HGF/ NK1-containing medium, there was little or no additional stimulation. In contrast, while HGF/NK1 and heparin each by itself caused only a slight increase in the amount of Met immunoprecipitated from CHO-745 cells by phosphotyrosine antibody, together they stimulated a large increase in the amount of immunoprecipitated Met protein. This result strongly suggested that the addition of heparin to glycosaminoglycan-deficient CHO cells not only enhanced HGF/NK1 binding to Met, but also triggered tyrosine kinase activity.
Further evidence of signaling was obtained by analysis of c-fos expression. c-fos induction is a sensitive measure of rapid cellular responses to mitogenic stimulation (51), which served as a convenient indicator of HGF/NK1 signaling. When wildtype CHO cells were incubated with HGF/NK1 at 100 ng/ml, an increase in c-fos transcript was observed after 15 min and was more pronounced at 60 min (Fig. 7). Heparin alone did not stimulate c-fos expression, and had only a modest positive effect when tested in combination with HGF/NK1 on CHO-WT cells. In contrast, HGF/NK1 alone stimulated only a slight increase in c-fos expression by CHO-745 cells after either 15 or 60 min. However, when heparin (3 g/ml) was included in the incubation medium, c-fos induction in CHO-745 was much stronger after 60 min, comparable to that observed with the wild-type cells. Heparin itself had no effect on c-fos expression in this assay. These findings indicated that the heparin-dependent interaction of HGF/NK1 with Met in CHO-745 cells was capable of activating the signal transduction process.

N Domain Mediates Heparin-induced Ligand
Oligomerization-Because of earlier reports that heparin enhanced the activity of FGFs (8 -10) and HGF/SF (37) by promoting ligand oligomerization, we investigated the possibility that the positive impact of heparin or cell surface heparan sulfate proteoglycan on Met-ligand binding and signaling might involve oligomerization of HGF/SF isoforms. We first examined the effect of heparin on HGF/NK1 oligomerization in a cell-free setting. In the absence of heparin, there was little evidence of crosslinked HGF/NK1 dimer as detected by Coomassie Blue staining after SDS-PAGE. However, when heparin was present during the incubation with cross-linking agent, subsequent analysis revealed that a substantial portion of HGF/NK1 protein migrated as a dimer (Fig. 8, left panel). Weak bands corresponding to higher order oligomers occasionally were also seen. When similar experiments were performed with separately prepared recombinant N and K1 domains, the ability to oligomerize correlated unequivocally with the N domain. As shown in the middle panel of Fig. 8, a small amount of putative N domain dimer was evident in the absence of heparin, and in the presence of heparin a set of bands were observed which represent a series of oligomers. Frequently the series included bands corresponding in size to hexamers and even octamers. In contrast, there was no evidence of oligomerization of the K1 domain either in the absence or presence of heparin (right panel, Fig. 8).
Ligand oligomerization also was observed when incubations were performed with cells in culture. Cross-linked 125 I-HGF/ NK1 dimer was readily detected when tracer was added to BALB/MK or CHO-WT cells, but was far less conspicuous with the glycosaminoglycan-deficient CHO-745 cells (Fig. 9). The addition of 25 nM unlabeled HGF/NK1 increased the amount of dimer in lysates of all three cell types, and even trimer and tetramer were seen with BALB/MK and CHO-WT cells. Exogenous heparin only promoted oligomer formation in CHO-745 lysates, while it inhibited oligomerization in cultures of cells having normal levels of proteoglycan. These findings suggested that a moderate amount of heparin or endogenous heparan sulfate proteoglycan could facilitate oligomerization on the cell surface, but that excessive heparin would block this process. DISCUSSION Several reports suggest that the interaction of heparin-binding growth factors with heparan sulfate proteoglycan has a major impact on their biological effects (2-5, 7-10, 25, 34 -37). In practical terms, the clinical utility of such growth factors may be strongly influenced by their binding to the proteoglycan on blood vessel walls and in extracellular matrix, which could hamper delivery of factor to cellular targets. The present study was undertaken to identify the heparin-binding site(s) of HGF/SF and investigate the role of heparin in growth factor binding and signal transduction. Mapping of the primary hep- arin-binding site of HGF/SF to the N domain is a first step in designing mutants with altered heparin affinity. Such derivatives would be useful in testing the importance of the proteoglycan interaction for growth factor activity and might have greater efficacy in therapeutic applications.
Previous analyses of HGF/SF deletion mutants had implicated the N domain (31) and more specifically the hairpin loop (32) in heparin interactions. However, in deletion studies loss of function could result not only from removal of the targeted domain but also from improper folding of the residual protein that could reduce its stability or alter tertiary structure. For instance, the K2 deletion mutant also reportedly bound poorly to heparin (32), suggesting a critical role for K2 in this interaction that was inconsistent with our results. By showing that the N domain retained the binding properties of full-length HGF/SF, we directly demonstrate that the heparin-binding site of HGF/SF resides in this domain.
Because the heparin-binding motifs of many proteins involve basic amino acid residues clustered in a linear sequence (52,53), the proposed role of two such sites in the hairpin region of the N domain was tested. The failure of alanine substitution at these positions to have much impact on the interaction with heparin-Sepharose suggested that other residues must be responsible. Two clusters of basic residues near the amino terminus of the N domain, RKRR (residues 33-36) and KKSAK (residues 43-47), probably can be excluded from consideration because corresponding sequences were absent from a truncated form of rat HGF/SF, which bound tightly to heparin-Sepharose (14,54). Two other segments, KIKTKK (residues 58 -63) and KK (residues 109 -110), are worthy of investigation, as well as other nearby residues that may contribute to heparin binding (52,53,55).
While our evidence that HSGAG is important for Met-ligand binding complements other recent work, it would appear to be in conflict with earlier observations made in a cell-free setting. When Mark et al. (56) detected high affinity binding of HGF/SF isoforms to a fusion protein containing the extracellular domain of Met and the heavy chain Fc portion of immunoglobulin G, no heparin or HSGAG had been added to the assay. These ostensibly contradictory observations could be reconciled in several ways. First, HSGAG may have inadvertently co-purified with the Met-IgG fusion protein, and consequently been present in the cell-free assay. Such co-purification is plausible given the evidence that the extracellular domain of another heparin-binding growth factor, tyrosine kinase receptor can itself bind heparin (6). Second, it is possible that nonspecific adsorption of HGF/SF isoforms in the solid phase binding assay used by Mark et al. (56) resulted in high local concentrations of ligand, obviating one of the potential mechanisms of HSGAG action in promoting high affinity receptor binding (5). Third, heparin-dependent protection of ligand from proteolytic degradation may not have been important in a cell-free binding assay, where proteases presumably are scarce. Finally, if the HSGAG requirement for Met-ligand binding involves receptor dimerization as has been suggested both for FGFs (5, 8 -10) and HGF/SF isoforms (25,37), this probably would have been unnecessary in the cell-free assay because immunoglobulin fusion proteins typically already exist as dimers.
The present findings indicated that heparin or endogenous HSGAG was required both for binding and activation of receptor signaling. The potent stimulation of tyrosine phosphorylation and c-fos expression in CHO-745 cells elicited by the combination of HGF/NK1 and exogenous heparin established the biological relevance of heparin-dependent binding reconstituted in these cells. Occasionally weak responses to HGF/NK1 alone were observed in CHO-745 cells, presumably due to a low level of receptor binding that was below the sensitivity of detection by cross-linking analysis. Such binding might arise either from small amounts of endogenous HSGAG on CHO-745 cells or occur independent of HSGAG. Similarly, heparin itself at times appeared to elicit a minor tyrosine phosphorylation signal. While not a consistent finding, this may reflect a capacity of the glycosaminoglycan to promote tyrosine phosphorylation, as has been reported in cells expressing the tyrosine kinase receptor FGFR-4 (57).
Our results emphasized the importance of heparin not only for Met binding but also ligand oligomerization. As recently reported for HGF/NK1 (25), cross-linking experiments performed in a cell-free setting suggested that HGF/NK1 and especially the N domain itself have an intrinsic tendency to oligomerize. This predisposition is markedly enhanced by the addition of heparin. For the N domain, a series of oligomers ranging up to octamers formed in the presence of heparin. When 125 I-labeled HGF/NK1 (Fig. 9) or N domain 3 was incubated with cells and cross-linking agent, oligomers again were observed if endogenous heparan sulfate proteoglycan or added heparin was present. Because the latter experiments were  8. Oligomerization of HGF/NK1, N, and K1 in a cell-free setting. Cross-linking reactions were performed as described under "Experimental Procedures." Protein was incubated alone or with crosslinking agent bis(sulfosuccinimidyl)suberate in the absence or presence of heparin. Subsequently, protein was resolved by SDS-PAGE and detected by staining with colloidal Coomassie Blue. In addition to oligomers of HGF/NK1 and the N domain, a doublet was observed when cross-linker was added either to HGF/NK1 or the K1 domain. This probably resulted from intramolecular cross-linking of compact, nonreduced protein, as earlier work indicated that HGF/NK1 migrated more rapidly in SDS-PAGE under non-reducing conditions (23). The positions of molecular mass markers are at the right.
performed with whole cells and ng/ml concentrations of tracer, the potential physiological relevance of ligand oligomerization was reinforced. Of note, the effect of exogenous heparin on ligand oligomerization varied with cell type: while heparin increased dimer formation on glycosaminoglycan-deficient CHO cells, it reduced oligomerization on BALB/MK and wild type CHO cells. In the latter instances, HGF/SF ligands probably were dispersed among the abundant heparin molecules such that ligand-ligand interactions were less likely to occur. Thus, the impact of heparin on oligomerization is dependent on the content of endogenous HSGAG. If ligand oligomerization is important for signal transduction, the effect of added heparin on the biological activity of HGF/SF ligands will vary with the quantity and perhaps composition of proteoglycan on the cell surface.