Macrophage Stimulating Protein (MSP) Binds to Its Receptor via the MSP β Chain*

Macrophage stimulating protein (MSP) is a 78-kDa disulfide-linked heterodimer belonging to the plasminogen-related kringle protein family. MSP activates the RON receptor protein-tyrosine kinase, which results in cell migration, shape change, or proliferation. A structure-activity study of MSP was performed using pro-MSP, MSP, MSP α and β chains, and a complex including the first two kringles and IgG Fc (MSP-NK2). Radioiodinated MSP and MSP β chain both bound specifically to RON. The K d of 1.4 nm for MSP β chain is higher than the reportedK d range of 0.6–0.8 nm for MSP. Pro-MSP, MSP α chain, and MSP-NK2 did not bind. Only MSP stimulated RON autophosphorylation. Although the β chain bound to RON and partially inhibited MSP-induced RON phosphorylation in kidney 293 cells, it did not induce RON phosphorylation. Pro-MSP, MSP α chain, or MSP-NK2 failed to activate RON, consistent with their inability to bind to the RON receptor. Functional studies showed that only MSP induced cell migration, and shape change in resident macrophages, and growth of murine keratinocytes. Our data indicate that the primary receptor binding domain is located in a region of the MSP β chain, in contrast to structurally similar hepatocyte growth factor, in which the receptor binding site is in the α chain. However, full activation of RON requires binding of the complete MSP disulfide-linked αβ chain heterodimer.

Macrophage stimulating protein (MSP) 1 was originally purified from human plasma, based on its activity for murine resident peritoneal macrophages (1). It is a 78-kDa heterodimeric protein composed of a disulfide-linked 53-kDa ␣ chain and a 25-kDa ␤ chain (calculated from amino acid composition). The ␣ chain contains a N-terminal hairpin loop followed by four kringle domains. The ␤ chain has a serine protease-like domain but is devoid of enzymatic activity due to amino acid substitutions in the catalytic triad. MSP belongs to the kringle protein family that includes plasminogen (2) and hepatocyte growth factor/scatter factor (HGF/SF) (3,4). MSP is synthesized mainly by liver cells (5,6), circulates in blood as a biologically inactive single chain precursor (7), and is cleaved by members of the kallikrein family (8,9) or by trypsin-like enzymes located on macrophage surfaces (7). Recent functional studies have revealed that in addition to induction of macrophage shape change, chemotactic migration (10), and phagocytosis of C3bicoated erythrocytes (1), MSP has other activities. These include inhibition of expression of inducible nitric oxide synthase mRNA in endotoxin or cytokine-stimulated macrophages (11), induction of interleukin-6 production and differentiation of megakaryocytes (12), suppression of colony formation of human bone marrow cells induced by Steel factor plus granulocyte macrophage-stimulating factor (13), increase in beat frequency of nasal epithelium cilia (14), and stimulation in vitro of proliferation of certain epithelial cell lines (15)(16)(17).
The receptor for MSP was recently identified as the human RON gene product (18), a transmembrane receptor proteintyrosine kinase cloned from a human keratinocyte cDNA library (19). The murine STK gene cloned from hematopoietic stem cells of bone marrow is the homologue of human RON (20,21). The RON gene encodes a 190-kDa heterodimeric protein composed of a 40-kDa extracellular ␣ chain and 150-kDa transmembrane ␤ chain with intrinsic tyrosine kinase activity (21). This property places the product of the RON/STK gene into a subfamily of receptor tyrosine kinases that includes proto-oncogene MET and SEA (22,23). These receptors share many unique structural properties including a putative proteolytic cleavage site, similar location of cysteine residues in their extracellular domain, and two conserved tyrosines in the Cterminal tail (19,20,22,23). Studies of the signaling pathways of RON have shown that tyrosine-phosphorylated RON associates in vivo with intracellular signal transducers, including Grb-2-Sos and phosphatidylinositol 3-kinase (17,24).
In this work, we initiated a structure-activity study of MSP to identify functionally important domains that interact with the RON receptor. Five purified recombinant proteins were used, including pro-MSP, MSP, MSP ␣ and ␤ chains, and the MSP N terminus (including the first two kringles) fused to human IgG Fc. We report the binding capacity of MSP and its subunits to RON receptor in intact cells. We also analyzed the capacity of MSP and its subunits to induce receptor phosphorylation and consequent cellular responses.

MATERIALS AND METHODS
Reagents-Human mature plasma MSP was purified as described (1). Human recombinant single chain pro-MSP was derived from CHO * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. cells transfected with human MSP cDNA and purified in two steps by S-Sepharose and anti-MSP IgG affinity column chromatography. MSP ␣ and ␤ chains were obtained from kallikrein-treated pro-MSP and purified on a CM-Sepharose column. An N-terminal segment of recombinant MSP that included the first two kringles (MSP-NK2) fused with human IgG Fc (16) was produced at Genentech, Inc. (San Francisco, CA). The purity of the above reagents was evaluated by SDS-PAGE under reducing and nonreducing conditions (Fig. 1). Rabbit IgG antibodies against a synthetic C-terminal peptide of RON ␤ chain were as described (18). Mouse monoclonal antibody to phosphotyrosine (4G10) was from Upstate Biotechnology Inc. (Lake Placid, NY). Goat antimouse or rabbit IgG conjugated with horseradish peroxidase and enhanced ECL detection reagents were from Amersham Corp. RPMI 1640 and Dulbecco's modified Eagle's medium were from Life Technologies, Inc. Bolton-Hunter reagent was from NEN Life Science Products. Protein G-Sepharose was from Pharmacia Biotech Inc.
Radiolabeling of MSP, MSP ␣ and ␤ Chains, and MSP-NK2-10 g of each protein in 15 l of 0.1 M borate buffer, pH 8.5, were added to 250 Ci of 125 I-labeled Bolton-Hunter reagent (25) and equilibrated on ice for 60 min. The reaction was terminated by the addition of 0.2 M, pH 8.5, glycine borate buffer. The reaction mixture was then applied to an Excellulose GF-5 desalting column (Pierce) equilibrated with phosphate-buffered saline containing 0.25% gelatin. Iodinated protein was eluted with 2 ml of phosphate-buffered saline-gelatin buffer and counted in a gamma counter (Gamma 5500, Beckman). The specific activity of the labeled proteins was about 200 Ci/mmol.
Absorption of Radiolabeled MSP ␤ Chain by MDCK-RE7 and 3T3/ STK Cells-CHO-MSP18 cells were incubated with 100 Ci of [ 35 S]cysteine in cysteine-free Dulbecco's modified Eagle's medium without fetal bovine serum for 56 h. Under these conditions, all the recombinant protein is 35 S-labeled single chain pro-MSP (8). 35 S-pro-MSP in culture supernatants was then converted into two chain mature MSP with 50 nM kallikrein, a serine protease that specifically cleaves pro-MSP at the Arg 483 -Val 484 bond (8). The concentrations of MSP were determined by a specific sandwich enzyme-linked immunosorbent assay (26). Immunoprecipitation and SDS-PAGE under nonreducing conditions of cleaved 35 S-pro-MSP showed not only disulfide-linked mature MSP but also free ␣ and ␤ chain. We concluded that about 30% of the recombinant pro-MSP preparation did not have a disulfide link between its ␣ and ␤ chain, which results in free ␣ and ␤ chain after specific R-V bond cleavage. For the absorption assay, MDCK-RE7 or 3T3/STK cells (6 ϫ 10 6 in 0.5 ml of RPMI 1640 medium) were equilibrated with 0.5 nM 35 S-MSP mixtures at 0°C for 2 h. Supernatants were collected, and rabbit anti-MSP IgG was added to precipitate the remaining MSP as well as MSP ␣ and ␤ chains; this was followed by the addition of protein G-Sepharose. After extensive washing with 0.1 M Tris buffer, pH 7.6, containing 0.15 M NaCl and 0.5% Tween 20, samples were separated on a 12% gel by SDS-PAGE under reducing conditions. The gel were treated with Enlightning for 20 min, dried at 75°C, and exposed to film with an intensifying screen.
Binding of 125 I-MSP or Its Subunits to Cells-Binding of 125 I-MSP, MSP ␣, MSP ␤, or MSP-NK2 protein to MDCK-RE7, BK-1, MK308, and other cells was carried out as described (15). In steady-state binding assays, 3 ϫ 10 5 cells were equilibrated in duplicate with increasing amounts of 125 I-labeled MSP, MSP ␤, or MSP-NK2 in binding buffer (RPMI 1604 medium, pH 7.4, with 20 mM Hepes, and 100 g/ml cytochrome c) in a total volume of 200 l. Nonspecific binding was determined in parallel equilibrations with a 30-fold excess of unlabeled MSP, MSP ␤, or MSP-NK2. After 3 h at 0°C, cells were pelleted through an oil cushion (18). The tips of tubes containing cells were cut. Radioactivity in supernatants and tips was counted in a gamma counter. For estimation of K d , a we used a linear regression to generate a straight line through Scatchard plot data points.
Detection of Tyrosine Phosphorylation of RON-A suspension of 5 ϫ 10 6 cells in 1 ml of binding buffer was incubated with 5 nM MSP at 37°C for different time intervals. Cell were then equilibrated for 30 min in 200 l of lysis buffer (50 mM Tris buffer, pH 7.4, 1% Triton X-100, 1% Nonidet P-40, 150 mM NaCl, 2 mM EDTA, 100 M vanadate, 20 g/ml leopeptin, 20 g/ml aprotinin, and 50 g/ml soybean trypsin inhibitor). Lysate proteins were precipitated with monoclonal antibody ID2 to RON or rabbit anti-STK serum coupled to protein G-Sepharose beads. Samples were dissolved in sample buffer with 2-mercaptoethanol, separated on a 7.5% polyacrylamide gel by SDS-PAGE, and transferred to Immobilon-P (Millipore). Membranes were blocked with 1% bovine serum albumin in 0.15 M pH 7.6 Tris buffer with 0.5% Tween 20, then incubated with 0.2 g/ml anti-phosphotyrosine antibody overnight, followed by goat anti-mouse IgG conjugated with horseradish peroxidase. The horseradish peroxidase reaction was developed with ECL detection reagents. In some experiments, the membrane was treated with SDS/ 2-mecaptoethanol erasure buffer and reprobed with rabbit anti-RON serum as described (18).
Assay for Macrophage Shape Change-Murine peritoneal resident macrophages (5 ϫ 10 5 /ml) were incubated in 1 ml of serum-free RPMI 1640 medium in 24-well tissue culture plates. MSP, MSP subunits, or their different combinations were added. After incubation at 37°C for 45 min, cells were photographed.
Cell Migration Assay-The assay was done as described (18). Bottom wells of a chemotaxis chamber were filled in triplicate with 30 l of RPMI 1640 medium containing different amounts of MSP or MSP subunits and then covered with a polycarbonate membrane coated with mouse collagen IV. Upper wells were filled with 45 l of cell suspension (2 ϫ 10 6 /ml in RPMI 1640 medium). To see the effect of MSP subunits on MSP-induced migration, cells were first mixed with 5 or 30 nM of MSP ␣ or ␤ chain or MSP-NK2 and then added to top wells. After a 3-h incubation at 37°C, the chamber was disassembled, and the membranes were dried in air. The migrated cells were stained and counted with an image analyzer. The results were expressed as the percentage of input cells that migrated.
Cell Proliferation Assay-The experiments were performed as described (15). BK-1 cells at a concentration of 10 5 /ml of a serum-free medium (equal volumes of keratinocyte serum free-medium, Eagle's minimum essential medium, and CHO-SF medium) were seeded at 100 l/well in a 96-well culture plate. MSP, MSP ␣ or ␤ chains, MSP-NK2, or their different combinations were added. Cells without stimulation served as control. After incubation for 5 days, cells were stained and lysed in 1% SDS buffer. Color intensity was measured at 570 nM in an enzyme-linked immunosorbent assay plate reader. Absorbance was converted into cell number by reference to a standard curve derived from stained cell concentration.

Absorption of Free MSP ␤ Chain by MDCK-RE7 or 3T3/STK
Cells-In the course of studying pro-MSP conversion into mature MSP, we noticed that about 30% of our metabolically 35 S-labeled recombinant pro-MSP lacked the disulfide link between its ␣ and ␤ chain, which resulted in free ␣ and ␤ chain after specific cleavage by kallikrein of the pro-MSP R-V bond at the ␣␤ chain junction (data not shown). We took advantage of this finding to determine if free ␣ or ␤ chain binds to the MSP receptor (human RON or murine STK) using an absorption assay. When 35 S-labeled pro-MSP was cleaved by kallikrein and then equilibrated with RON-expressing MDCK-RE7 cells or 3T3/STK cells as absorbents, MSP ␤ chain in recovered supernatants from both MDCK-RE7 and 3T3/STK cells was significantly reduced, as analyzed by SDS-PAGE (Fig. 2). By densitometric comparison with nontransfected control cells, about 80% of fluid phase MSP ␤ chain was absorbed by MDCK-RE7 cells, and about 50% was absorbed by 3T3/STK cells. In contrast, the level of MSP ␣ chain did not change. Absorption by these cells suggested that the MSP ␤ chain might bind to the RON receptor.
Assay for Binding of 125 I-pro-MSP, MSP, and Its Subunits to Cells Expressing RON or STK-We tested for binding of radiolabeled pure MSP and its subunits to murine keratinocyte BK1 and MK308 cells, which express 10,000 -15,000 STK receptors/ cell (15). Fig. 3 shows that in both cell lines, specific binding of 125 I-MSP was inhibited in a concentration-dependent manner by unlabeled MSP ␤ chain but not by MSP ␣ chain. On a molar basis, MSP is more potent than the free ␤ chain as a competitive inhibitor of labeled MSP binding. Pro-MSP did not compete with MSP for RON, as previously reported (15). Binding of 125 I-MSP ␤ chain to MDCK-RE7 cells is shown in Fig. 4. Binding of the MSP ␤ chain to the RON receptor was specific; either unlabeled MSP or MSP ␤ chain inhibited binding of 125 I-MSP ␤ chain. From Fig. 4C, we estimated a K d for binding of the MSP ␤ chain of about 1.7 nM, higher than the K d values of 0.6 to 0.8 for binding of MSP to the RON receptor (17). On the other hand, we did not detect specific binding of the MSP ␣ chain to the RON receptor (data not shown). The relatively low binding of 125 I-MSP-NK2 to the cell surface was unaffected by unlabeled MSP or MSP-NK2, indicating that the interaction of labeled MSP-NK2 with MDCK-RE7 cells is nonspecific (Fig. 5).
Induction of RON Tyrosine Phosphorylation by MSP and Its Subunits-We next studied tyrosine phosphorylation of RON induced by pro-MSP, MSP, and its subunits in kidney 293 and MDCK-RE7 cells. After precipitation of proteins with ID2 anti-RON, Western blot with monoclonal antibody 4G10 to phosphotyrosine showed that only MSP-induced tyrosine phosphorylation of the 150-kDa RON ␤ chain (Fig. 6A) Fig. 6B shows that high con- Effect of MSP Subunits on MSP-induced Cell Shape Change and Migration-To see if MSP subunits can induce cell shape change or migration, mouse peritoneal resident macrophages were used. Fig. 7 shows that MSP ␣, MSP ␤ or MSP-NK2 protein did not induce morphological changes in resident macrophages. In combination with MSP, none of these three subunits inhibited the biological effect of MSP on macrophages. Likewise, except for a statistically insignificant effect of MSP ␤, the addition of MSP subunits to macrophages did not inhibit their migration toward MSP as a chemoatractant (data not shown).
Effect of MSP Subunits on MSP-stimulated Cell Proliferation-BK-1 keratinocytes were used to assess if MSP ␣ chain, MSP ␤ chain, or MSP-NK2 protein at high concentration could affect MSP-induced cell proliferation. Table I shows that only MSP increased cell number after 5 days in culture, compared with the medium control. In combination experiments, none of the fragments affected MSP-induced proliferation, except for a small inhibition by 100 nM MSP ␤ chain. DISCUSSION We have presented several lines of evidence that the MSP ␤ chain binds to RON. 1) Metabolically labeled free ␤ chain, but not ␣ chain, was specifically absorbed by cells expressing the RON receptor. 2) 125 I-␤ chain bound to RON in intact cells in a specific and saturable manner. 3) Not only unlabeled MSP but also ␤ chain competitively inhibited binding of 125 I-MSP to RON in intact cells. Thus, in contrast to the ␤ chain of HGF/SF, which does not bind to its receptor (Met) (27), the ␤ chain of MSP appears to contain the primary binding site for the RON receptor. The ␤ chain is the serine protease domain of kringle proteins (28). In HGF/SF and MSP, amino acid substitutions in the catalytic triad have eliminated the protease activity. It is possible that residues in the modified substrate pocket of the MSP ␤ chain might form the binding site for RON. To obtain clues about the MSP binding domain, we have begun modeling of the MSP ␤ chain for comparison with the published model of the HGF ␤ chain (29). MSP has Asp and Asn in the binding pocket, the corresponding locations of which are Gly in HGF. 2 Substitution of these candidate residues, as well as two surface loop arginines, are now being made to evaluate their significance for receptor binding.
There is one report that in RON cDNA-transfected COS-1 cells, MSP-NK2 stimulated RON phosphorylation (16). MSP-NK2 is a recombinant protein comprising the first two kringles of the MSP ␣ chain, fused to IgG Fc. We found that neither MSP-NK2 nor MSP ␣ chain bound to RON on intact cells including kidney 293 and MDCK-RE7 cells. We cannot explain the reported activity of MSP-NK2, especially because the source of the MSP-NK2 was the same. However, our ␤ chain data combined with the fact that free ␣ chain does not bind to RON support the conclusion that MSP binds to its receptor via the ␤ chain.
We have shown that although the MSP ␤ chain binds to RON, it does not cause biological activity or induce phosphorylation of the receptor, except for a small amount at high ligand concentrations. It is generally accepted that ligand binding to growth factor receptors is associated with receptor oligomerization and autophosphorylation (30). Receptor oligomerization may be mediated by interaction of ligand pairs. In this context, our results indicate that receptor oligomerization requires an intact ␣␤ chain disulfide-linked heterodimeric ligand. Although HGF differs from MSP in that the primary binding site resides in the ␣ chain, two or three amino acid substitutions in the ␤ chain are sufficient to reduce biological activity to less than 2% that of wild type HGF (31). Thus, for both MSP and HGF the ␣␤ chain heterodimer is required to fully activate their respective receptors. An HGF dimer, formed by noncovalent interactions between kringles 2 and 3 of the protein pair, has been suggested as the moiety that induces dimerization and activation of MET (29). This idea is supported by a report of nonconvalent kringle-kringle interactions (32). Although MSP and HGF have different primary receptor binding regions, they may have a similar structural basis for receptor activation by dimer formation through kringle interactions. We plan to express and purify selected kringle regions of MSP and to determine their effects when added together with intact MSP to target cells. If receptor activation requires ligand dimerization by kringle interaction, the result could be no effect on MSP binding but inhibition of receptor phosphorylation.
An interesting alternative mechanism for ligand-induced receptor dimerization is suggested by the crystal structure of human growth hormone and the extracellular domain of its receptor (33). The complex comprises one ligand molecule per two receptors. Two structurally unrelated regions of the ligand interact with similar binding surfaces of the two receptors. It has been proposed that receptor dimerization occurs by a sequential mechanism, because human growth hormone binds to a second receptor only if it has bound to the first receptor (34). This is consistent with the fact that the contact surface for the binding site of the receptor I is about 30% larger than that for receptor II. The authors suggest that ligand binding to receptor II is stabilized by interaction between the two receptor domains near their C terminus. If this model applied to the RON receptor, the candidate region for binding to receptor I might be a cluster of ␤ chain surface loop arginines 2 ; a single arginine on the N domain hairpin loop of the ␣ chain (29) might mediate binding to receptor II. The Arg cluster density for the corresponding regions of HGF is reversed, which is consistent with receptor binding by the ␣ chain. This model would account for primary binding by MSP ␤ or HGF-␣, and the requirement for binding by the ␣␤ chain heterodimer for optimal receptor activation. The model should be readily testable by mutagenesis studies.