Concomitant expression of hepatocyte growth factor/scatter factor and the receptor c-MET in human myeloma cell lines.

Myeloma cell line supernatants were screened for their ability to inhibit the activity of transforming growth factor-β (TGFβ) in the mink lung cell (Mv-1-Lu) bioassay. Supernatant from the human myeloma cell line JJN-3 contained potent TGFβ antagonistic activity. This activity was isolated and found to be associated with a 72-78-kDa glycoprotein. Specific polyclonal and monoclonal antibodies were generated toward the 72-78-kDa protein, and these antibodies precipitated the TGFβ inhibitory activity from JJN-3 supernatant. Upon amino acid sequencing the protein appeared to be identical to hepatocyte growth factor (HGF), and some of the generated antibodies directly blocked the action of recombinant HGF in various assays. By HGF-specific polymerase chain reaction we demonstrated that HGF mRNA was expressed in five out of five tested myeloma cell lines. The level of HGF protein in supernatants showed great variation from >500 ng/ml in JJN-3 supernatant to a few ng/ml in the supernatants from other myeloma cell lines. The same five cell lines were also screened for expression the HGF receptor c-MET. Four of them expressed the receptor as shown by reverse transcriptase-polymerase chain reaction and Western blot. The receptor was shown to be constitutively phosphorylated in the human myeloma cell line JJN-3. This receptor could be dephosphorylated by anti-HGF antibodies, indicating the existence of an autocrine HGF loop in this cell line. We propose that HGF/c-MET may play a role in multiple myeloma.

Multiple myeloma, a malignant disease that involves proliferation of monoclonal plasma cells, is associated with several clinical manifestations such as skeletal pathology, anemia, hypercalcemia, and renal dysfunction. The cause of these features of the disease is only partly understood, but production of soluble factors by the myeloma cells is likely to be involved. In screening for unknown myeloma-produced factors, we noticed that several supernatants from myeloma cell lines counteracted the growth inhibiting effect of TGF␤ on the mink lung cells Mv-1-Lu. Here we report the purification and characterization of this activity from the supernatant of the human myeloma cell line JJN-3. The activity was associated with a 72-78-kDa protein, which upon sequencing appeared to be identical to hepatocyte growth factor (HGF). 1 HGF is a pleiotropic cytokine of mesenchymal origin (for a recent review see Ref. 1). It is known to be produced by fibroblasts, macrophages, and smooth muscle cells (2,3). Recently HGF production by B-lineage cell lines was also reported (4). Among the known effects of HGF is growth stimulation of hepatocytes (5), endothelial cells (6,7), and various epithelial cells (8). HGF also causes spread of epithelial cells, a property of HGF that led to the designation "scatter factor" (3). This name is still used synonymously with HGF. Because of HGF's ability to promote cell proliferation and blood vessel formation (7,9) and to disperse cohesive epithelial cells, it is proposed to be involved in the process of cancer growth and metastasis. The receptor for hepatocyte growth factor is a transmembrane tyrosine kinase encoded by the proto-oncogene c-Met (10). Concomitant expression of HGF and its receptor by cancer cells has been shown to be associated with increased malignancy (11)(12)(13).
By reverse transcriptase polymerase chain reaction (RT-PCR), we showed that five out of five human myeloma cell lines express HGF mRNA, and HGF protein was found in supernatants from four of these cell lines. To determine whether HGF has the potential to exert autocrine effects on the myeloma cells, we looked for HGF receptor expression on the same cells. By RT-PCR and Western blots we showed that four out of the five cell lines also expressed the HGF receptor, c-MET. Autocrine HGF-mediated tyrosine phophorylation of c-MET in the cell line JJN-3 could be blocked by anti-HGF. We propose that HGF is one of the long sought cytokines with pathophysiological functions in multiple myeloma.

MATERIALS AND METHODS
Cell Lines and Materials-The human myeloma cell line JJN-3 (14) was a gift from Jennifer Ball, Department of Immunology, University of Birmingham, UK. The OH-2 (15) and JW cell lines were established in our laboratory from pleural effusions of two myeloma patients. The Mv-1-Lu (CCL-64) (16), SW-480, RPMI 8226 (17), and U-266 (18) cell lines were purchased at American Type Culture Collection (Rockville, MA). Cells were grown in RPMI 1640 (Life Technologies, Inc.) supplemented with 10% fetal calf serum, 2 mM L-glutamine, and 40 g/ml gentamicin (complete medium (CM)), except for the OH-2 cell line, where the fetal calf serum was replaced by human A ϩ serum (The BloodBank, Regionsykehuset, Trondheim, Norway). For estimation of HGF content in supernatants from various cell lines by enzyme-linked immunosorbent assay as described below, cells were seeded at a concentration of 7.5 ϫ 10 5 (OH-2 and U-266 cells), 5 ϫ 10 5 (JJN-3) or 2.5 ϫ 10 5 (RPMI 8226 and JW) cells/ml, and the supernatants were harvested after 72 h of culture and stored at Ϫ20°C. NSO cells were generously provided by Z. Eshhar, Weizmann Institute of Science, Rehovot, Israel (19). Porcine platelet-derived TGF␤ 1 and ␤ 2 , human platelet-derived TGF␤ 1 , and recombinant human HGF were from R & D Systems (Minneapolis, MN).
Mv-1-Lu Bioassay-The amount of TGF␤-inhibiting activity was estimated by ability to counteract the TGF␤ inhibition of DNA synthesis in Mv-1-Lu cells (16). Cells were seeded at 10 4 cells/well in microplates and incubated in the presence of 350 pg/ml porcine TGF␤ 1 with or without serial dilutions of samples in a final volume of 0.2 ml. After 20 h the cells were pulsed with 1 Ci/well of methyl-[ 3 H]thymidine (Amersham Corp.). They were harvested 4 h later with a Micromate 196 cell harvester (Packard, Meriden, CT), and ␤-radiation was measured. In some experiments porcine TGF␤ 2 or human TGF␤ 1 was used.
Purification of TGF␤-inhibiting Activity-After washing JJN-3 cells three times in Hanks' balanced salt solution to eliminate serum proteins, the cells were seeded in protein-free hybridoma medium (Life Technologies, Inc.) at a concentration of 5 ϫ 10 5 /ml in 225-cm 2 cell culture flasks (Costar, Cambridge, MA) in a total volume of 45 ml. The protein-free hybridoma medium was supplemented with 2.5 ϫ 10 Ϫ5 M mercaptoethanol, 2 mM L-glutamine, and 40 g/ml gentamicin. We incubated the cultures at 37°C in 5% CO 2 humidified atmosphere for 72 h, at which time the cells were pelleted and supernatant was frozen at Ϫ20°C. The cell pellet was dispersed in protein-free hybridoma medium, and cells were reused for supernatant production up to five times.
JJN-3 conditioned protein-free hybridoma medium was concentrated on a Q-Sepharose HP anion exchange column (Pharmacia Biotech Inc.) that was equilibrated with 10 mM Tris/HCl, pH 7.4. Bound material was eluted with 1 M NaCl. The activity was precipitated within 27-50% saturation of (NH 4 ) 2 SO 4 and reconstituted in 10 mM Tris/HCl, pH 7.4. After filtration through a 0.2-m filter (FP 030/3, Schleicher & Schuell), this fraction was loaded on a Resource Q (RQ) anion exchange column (Pharmacia) and eluted with a gradient of NaCl in 10 mM Tris/HCl, pH 7.4, at room temperature. The appropriate fractions from the RQ column were diluted 1:4 in 0.1% trifluoroacetic acid in H 2 O and loaded on a reverse phase column C2/C18 that was run on a Smart System high pressure liquid chromatography (Pharmacia). The mobile phase was 0 -100% acetonitrile containing 0.1% trifluoroacetic acid. Acetonitrile and trifluoroacetic acid was removed by lyophilization after chromatography.
Alternatively, the activity was purified by affinity chromatography by the following method. 10 mg of the 3F4 mouse anti-HGF monoclonal antibody (see below) was covalently immobilized on a Hitrap column (Pharmacia) by methods described by the manufacturer. The column was loaded with JJN-3 conditioned medium that had been concentrated on a Q-Sepharose HP anion exchange column as described above. After washing with 0.5 M NaCl, 10 mM Tris/HCl, pH 7.4, bound proteins were eluted in 0.05% trifluoroacetic acid buffered with NaH 2 PO 4 to pH 3.8. Final purification was done by reverse phase chromatography as described above. Protein concentrations were estimated by a dye fixation method (Bio-Rad) using bovine serum albumin (Sigma) as standard.
Development of Antibodies against the TGF␤-inhibiting Activity-Mice were immunized with 1 g of purified TGF␤-inhibiting protein emulsified in complete Freund's adjuvant in the footpads. After 2 weeks, a new immunization was performed subcutanously with protein in incomplete Freund's adjuvant. Six weeks later boosting injections were given intraperitoneally at 4 and 3 days before the removal of the spleen and fusion. Fusion of spleen cells with NSO cells was performed as described in (20), and selection of clones was performed on immunoplates (Maxisorp, Nunc, Denmark) coated with purified antigen. Binding antibodies in hybridoma supernatants were detected using peroxidase-labeled anti-mouse goat antibodies (Dako, Glostrup, Denmark) and ortho-phenylene diamine (Dako) as a substrate. The antibodies were isotyped using a commercial isotyping kit (Zymed, San Francisco, CA) and purified from hybridoma supernatants using goat anti-mouse Sepharose (Zymed). Epitope mapping was performed with a BiaLite instrument (Pharmacia) and with competitive binding immunoassays. Briefly, these sandwich TGF␤-inhibiting protein immunoassays were performed using various capture antibodies, and the secondary antibodies were labeled with digoxigenin (digoxigenin antibody labeling kit, Boehringer Mannheim). The binding of digoxigenin antibodies was detected with peroxidase anti-digoxigenin Fab fragments (Boehringer Mannheim) and ortho-phenylene diamine substrate reaction. Another sandwich enzyme-linked immunosorbent assay used to detect amount of the antigen (which eventually turned out to be HGF) consisted of two monoclonal antibodies, denoted 3F4 and 2B5, as catching antibodies. Detection of bound antigen was done by a rabbit polyclonal serum toward the protein. The sensitivity of this assay was approximately 500 pg/ml HGF, and the assay was not affected by the presence of 10% normal human or mouse serum nor by the presence of 1 mg/ml of plasminogen, which has about 38% amino acid similarity to HGF (21) (data not shown).
Metabolic Labeling and Immunoprecipitation of HGF-JJN-3 cells were seeded in 24-well culture plates (Costar) at a concentration of 2 ϫ 10 6 cells/ml medium. Each well contained 0.5 ml of RPMI medium without methionine or cysteine but supplemented with 2 mM glutamine, 0.05 mCi of Tran 35 S-label (ICN Biomedicals, Irvine, CA), and 10% fetal calf serum that had been dialyzed against PBS. In some experiments the cells were grown in the presence of tunicamycin as indicated. After 4 h the cells were collected by centrifugation, and the supernatant was incubated for 1 h at 4°C, either without antibodies or with 2 g/ml of the monoclonal antibody 3F4, which specifically recognizes HGF, or the control antibody 6H8 (mouse monoclonal antibody, which recognizes a 180-kDa surface antigen of activated NK cells). 2 Aliquots of 0.5 ml were further incubated for 30 min with 50 l (50% v/v) of anti-mouse Igcoated Sepharose (Pharmacia). The Sepharose particles were subsequently washed twice in 10 mM Tris, 1 mM EDTA, 0.5 M NaCl, 0.1% Nonidet P-40, pH 7.5. and once in the same buffer without NaCl. After the addition of 50 l of SDS sample buffer, the suspension was boiled for 1 min and analyzed by SDS-polyacrylamide gel electrophoresis. Protein bands were visualized by fluorography.
In other experiments designed to remove TGF␤-inhibiting activity from JJN-3 supernatant, 100 l of anti-mouse Ig-coated Sepharose was preincubated with 50 g of 3F4 or 6H8 for 2 h, washed in PBS containing 0.1% bovine serum albumin, and rotated overnight with 1 ml of JJN-3 supernatant. The Sepharose was pelleted, and the JJN-3 supernatants were tested for TGF␤-inhibiting activity in the Mu-1-Lv bioassay.
SDS-Polyacrylamide Gel Electrophoresis and Amino Acid Sequencing of Proteins-SDS-polyacrylamide gels were fixed and silver-stained by the method described (22). For sequencing purposes, proteins separated by gel electrophoresis were electrophoretically transferred to Immobiline-PSQ membranes (Millipore, Burlington, MA). Membranes were stained in Coomassie Blue before bands were cut out and subjected to amino acid sequencing using an Applied Biosystems A471 automatic sequencer coupled to an on-line amino acid identification system (Applied Biosystems, Foster City, CA). Alternatively, the purified protein was subjected to proteolytic degradation by V8 (Endo-Glu) protease (Pierce) at 37°C over night before separation of fragments by electrophoresis and sequencing as described above.
HGF PCR fragments were cloned into the pCR TM II vector (Invitrogene Corp., San Diego, CA). Inserts were sequenced on both strands applying M13 universal forward and reverse primers and analyzed on an Applied Biosystems automatic sequencing machine according to methods provided by the manufacturer (Perkin-Elmer). The HGF-specific fragment was transcribed in vitro by SP6 RNA polymerase (Promega) in the presence of [␥-32 P]CTP, and the transcribed RNA was used as a probe on Northern blots of total RNA or mRNA from myeloma cell lines by standard procedures (26). mRNA for this purpose was isolated using a Dynabeads mRNA DIRECT kit (Dynal, Oslo, Norway).
Detection of c-MET Protein by Western Blot-Cells were washed in PBS and extracted with 10 mM Tris/HCl, pH 6.8, 1% Triton X-100. Nuclei and insoluble material were pelleted, and the protein content of the extracts was measured with the Bio-Rad protein assay. 40 g of protein was solubilized in SDS sample buffer containing 1% ␤-mercaptoethanol, and run on 10% polyacrylamide gels. The gels were blotted onto nitrocellulose filters (Bio-Rad), and the filters were developed by rabbit polyclonal antibodies toward c-MET and goat anti-rabbit HRP conjugates and Enhanced Chemiluminescence (ECL) detection (Amersham Corp.). As anti-c-MET we used two different antibodies raised toward synthetic peptides comprising of either the 28 or the 12 COOHterminal amino acids of the c-MET ␤-chain (Santa Cruz Biotechnology Inc., Santa Cruz, CA). The appropriate synthetic peptide was added in excess to control reactions to check the specificity of the antibodies.
Immunoprecipitation of c-MET and Detection of Tyrosine-phosphorylated c-MET-JJN-3 cells were removed from culture bottles after 3 days of culture in CM and washed once in Hanks' balanced salt solution with 100 g/ml heparin. Samples of 2.5 ϫ 10 7 cells were used in immunoprecipitaion of c-MET, either without further incubation or after pretreatment for 30 min at 37°C in CM with 1:100 dilution of rabbit anti-HGF serum or 100 ng/ml HGF. The cells were washed in PBS and lysed in 70 l of modified RIPA buffer containing 50 mM Tris, pH 7.4, 0.25% sodium deoxycholate, 150 mM NaCl, 1 mM NaF, 1 mM Na 3 VO 4 , 1 mM phenylmethylsulfonyl fluoride, 1 mM EDTA, 1% Nonidet P-40, and 1 g/ml aprotinin. Nuclei were pelleted, and the supernatants were diluted 1:3 in RIPA buffer with 0.1% Nonidet P-40. The supernatants were further incubated at 4°C for 1 h with a mouse monoclonal antibody directed against the extracellular domain of c-MET (Upstate Biotechnology Incorporated, Lake Placid, NY) and subsequently for 1 h with 50 l (50% v/v) of anti-mouse Ig-coated Sepharose. The Sepharose particles were washed three times in RIPA with 0.1% Nonidet P-40 and pelleted. Sample buffer with 1% mercaptoethanol was added. SDSpolyacrylamide gels were run and blotted, and tyrosine-phosphorylated proteins were detected by ECL essentially as described above. HRPconjugated anti-phosphtyrosine antibodies were purchased from Transduction Laboratories (Lexington, KY).

RESULTS
Purification and Identification of the TGF␤-inhibiting Activity-Supernatant from the JJN-3 cells was antagonistic to the growth inhibiting effects of TGF␤ on Mv-1-Lu cells (Fig. 1). In order to purify this inhibitory activity, JJN-3 cells were adapted to grow in the absence of serum. After 3 days of culture, the cell supernatants contained approximately 500 units/ml of this activity (Table I). We defined one unit as the amount of activity that in 200 l restored the thymidine incorporation in Mv-1-Lu cells treated with 350 pg/ml TGF␤ to 50% of the thymidine incorporation in untreated cells. Porcine TGF␤ 2 and platelet-derived human TGF␤ 1 were inhibited in the same manner by the supernatant as porcine TGF ␤ 1 (data not shown).
The substance inhibiting TGF␤ was isolated by a series of chromatographic steps as shown in Table I. With respect to protein, the overall purification was 84-fold, and about 10% of the total activity in the supernatant was recovered. The reason for the apparent increase in activity during the ammonium sulfate precipitation step is unclear. From results obtained in the Mv-1-Lu assay, it can be seen that 1 unit corresponded to about 1 ng of purified material (Table I), indicating that the activity was due to a highly potent protein. Thus, the TGF␤ antagonist had a specific activity of similar order as the specific activity of TGF␤ in the Mv-1-Lu cells. However, binding experiments indicated that the purified inhibitor of TGF␤ did not inhibit the binding of iodinated TGF␤ to these cells (data not shown), suggesting that the purified protein had its own activity on the cells.
The final purification step consisted of reverse phase chromatography, and a typical chromatogram is shown in Fig. 2. The activity eluted as a peak at about 40% acetonitrile. When the peak from the reverse phase column was analyzed by SDS-polyacrylamide gel electrophoresis, the protein eluted from the RPC column appeared as a broad band at 72-78 kDa under nonreducing conditions (Fig. 3, lane E). Under reducing conditions, three main bands could be detected, migrating in the gels corresponding to molecular masses of approximately 90, 63, and 32-34 kDa (Fig. 3, lane F), suggesting that the broad 72-78-kDa band consisted of at least two subunits linked by disulfide bonds. After the generation of monoclonal antibodies against the 72-78-kDa protein, the inhibitory activity was purified by affinity chromatography before a final reverse phase column as above. This purification scheme resulted in the isolation of inhibitory activity with similar electrophoretic mobility (Fig. 3

, lane G versus lane E).
Initial attempts to identify the TGF␤-inhibiting activity by NH 2 -terminal amino acid sequencing of the protein comprising the 72-78-kDa band were unsuccessful. However, treatment of the purified protein by V8 protease (Endo-Glu) generated several fragments that could be sequenced after separation by SDS-polyacrylamide gel electrophoresis and transferal to PVDF membranes. Two of the fragments generated NH 2 -terminal sequences that were identical to sequences in the human hepatocyte growth factor/scatter factor (21,23). Thus, in one of the fragments, 17 consecutive amino acids were identified that all were identical to the reported sequence of human HGF for amino acids 45-61 in Ref. 21. It also became clear that the electrophoretic mobility of the purified TGF␤ antagonist corresponded closely to the earlier published data for HGF, both under reducing and nonreducing conditions (27).
To demonstrate that the TGF␤-inhibiting activity produced by the JJN-3 myeloma cells was due to the purified 72-78-kDa protein and that this protein was identical to HGF, monoclonal antibodies were raised toward the protein. As shown in Fig. 3, these antibodies recognized a protein of 72-78 kDa from the supernatants of JJN-3 cells. Furthermore, as shown in Fig. 4, these antibodies removed the TGF␤-inhibiting activity from the supernatants of JJN-3 cells, establishing that the TGF␤antagonistic activity was associated with the 72-78-kDa band. And finally, these antibodies precipitated recombinant human HGF, which in itself had identical TGF␤-inhibiting activity in the MV-1-Lu cells (data not shown). Taken together, these data demonstrate that the potent TGF␤-antagonistic protein purified from the cell supernatant of the myeloma cell line JJN-3 was HGF.
Biological Activity of the Purified Protein-As shown above, HGF produced by the JJN-3 myeloma cells appeared as a broad band at 72-78 kDa under nonreducing conditions in gel electrophoresis. HGF is known to be produced as a glycoprotein (28), and Fig. 5 shows the effect of treating the JJN-3 cells with tunicamycin before immunoprecipitation of HGF. Inhibition of N-linked glycosylation by tunicamycin resulted in an increase in electrophoretic mobility, indicating a reduction of molecular mass of approximately 10 kDa (Fig. 5). These results show that the broad band of 72-78 kDa probably consists of HGF with a heterogeneous glycosylation, in accordance with earlier reported results (28). However, it has been shown that glycosylation is not necessary for the biological action of HGF (28). In contrast to this, it has been reported that the cleavage of HGF into the ␣and ␤-chains is necessary for the biological activity of HGF (27). As judged from silver-stained gels run under reducing conditions, our preparations of HGF from JJN-3 supernatants resulted in material with 20 -60% cleavage in various preparations ( Fig. 3 and data not shown). Proteolytic enzymes capable of cleaving HGF have been characterized (29,30). Our results indicate that JJN-3 cells, even when grown in the absence of serum for several days, contain similar proteolytic activity and process HGF into bioactive protein.
Expression of HGF in Myeloma Cell Lines-By RT-PCR specific mRNA encoding HGF was detected in all tested myeloma cell lines (Fig. 6). Although we did not optimize the PCR for quantitation, there was a clear difference in specific band intensity obtained with cDNA from the various myeloma cell lines, suggesting differences in HGF mRNA expression. In contrast, we were not able to demonstrate the existence of  HGF-specific mRNA in normal human peripheral B-cells (data not shown). Molecular cloning and sequencing of the 749-base pair HGF PCR fragment revealed that the sequence of this fragment, which encodes most of the ␣-chain, was identical to the previously published HGF cDNA sequence obtained from human placenta (23). When the mRNA from the two myeloma cell lines JJN3 and U-266 was analyzed by Northern blot, applying in vitro transcribed HGF antisense RNA as a probe, four species of mRNA with the approximate sizes of 6.2, 3.0, 2.3, and 1.5 kilobases could be detected in both cell lines as shown in Fig. 7. The presence of these mRNA species, as well as the relative abundance of the bands as apparent in Fig. 7, corresponds closely with what has earlier been shown for human placenta mRNA (31), indicating that the transcription mechanism and RNA processing for the HGF gene in myeloma cells resembles the process taking place in placenta.
The ability of myeloma cells to produce HGF was further analyzed by assaying cell supernatants for the presence of HGF by sandwich enzyme-linked immunosorbent assay. Table II shows the content of HGF in the supernatants from five myeloma cell lines. JJN-3 produces up to 0.5 g/ml of HGF and is unique among these cell line in its ability to secrete very high amounts of HGF. However, significant amounts of HGF could also be detected in the supernatants of three out of four other myeloma cell lines (Table II).
Expression of c-met in Myeloma Cell Lines-RT-PCR was performed with c-met-specific primers amplifying an mRNA sequence coding for part of the intracellular portion of the ␤-chain of c-met. c-met-specific mRNA could be detected in four of the five tested cell lines (Fig. 6, lane B). The only exception was the RPMI 8226 cell line, in which no PCR amplification was achieved.
Protein extracts from the human myeloma cell lines U-266, JJN-3, JW, and RPMI 8226, as well as from the adenocarcinoma cell line SW-480 (positive control), were run in SDSpolyacrylamide gels and electroblotted onto nitrocellulose membranes. Rabbit polyclonal antibodies against the carboxylterminal end of the ␤-chain of the c-met product were used to detect the protein. U-266, JJN-3, SW-480 (Fig. 8), and JW (data not shown) cells all contained c-met protein. As shown by other groups, both the precursor protein, p170 met , and the ␤-chain, p145 met , were detected by this antibody (32). In RPMI 8226 cells the protein could not be detected (Fig. 8). Three other bands, present in all three myeloma cell lines but not in SW-480, were also recognized by the antibody (Fig. 8). When excess soluble antigen was added, not only the antibody binding to the two proteins of expected size was reduced, but also the binding of the three other bands. However, any relationship between these three bands and the p170 met or p145 met has not yet been found.
Autocrine Tyrosine Phosphorylation of c-MET in the Myeloma Cell Line JJN-3-The HGF receptor c-MET was found to be constitutively tyrosine-phosphorylated in JJN-3 cells after 3 days of culture in CM (Fig. 9, lane A). Some cell samples were taken from the culture bottles and washed in heparin-containing solution to deplete surface-bound HGF. Subsequent incubation with exogenously added HGF maintained the phosphorylation (Fig. 9, lane B), whereas incubation with antibodies to HGF caused dephosphorylation of c-MET (Fig. 9, lane C). The lack of tyrosine phosphorylation in the anti-HGF-treated samples was not due to a general down-regulation of the receptor, because in parallel blots probed with anti-c-MET, the receptor was detected to approximately the same level in cells treated with anti-HGF as in cells stimulated with HGF (Fig. 9, lane E versus lane F).  a Estimated by sandwich immunoassay. The detection limit of the assay was about 0.5 ng/ml. ND, not detected.

DISCUSSION
The most important finding presented here is that human myeloma cell lines produce HGF and its receptor c-met and that the receptor is stimulated in an autocrine manner. All but one cell line expressed both HGF and c-MET, suggesting that concomitant expression of these proteins is the rule in myeloma cell lines. In healthy tissue c-MET and HGF expression seems largely to be restricted to epithelial cells and mesenchymal cells, respectively. Whereas concomitant expression of HGF and its receptor seems to be uncommon in normal cells, it has been observed in tumor specimens such as nonsmall cell lung cancers, various sarcomas, and pancreatic adenocarcinomas (33)(34)(35)(36). Furthermore, murine NIH 3T3 fibroblasts that were co-transfected to express human HGF and the protooncogene c-MET became highly tumorigenic in nude mice (12,13). Transfection of human NBT-II rat bladder carcinoma cells with HGF created an autocrine loop, and the cells became more tumorigenic and invasive than their normal counterparts (11). Simultaneous expression of HGF and its receptor thus seems to correlate with increased malignancy. Interestingly, high levels of HGF (Ͼ1 ng/ml) were found in blood plasma of patients with hematological malignancies, notably 4 out of 21 patient with acute myeloblastic leukemia and 1 out of 2 multiple myeloma patients (37). To our knowledge, the present study shows the first cancer cell lines of hematological origin where simultaneous expression of this ligand receptor pair has been detected, and these findings raise the possibility of an autocrine effect of HGF in multiple myeloma. Further experiments showed that both HGF and c-MET were also simultaneously expressed in highly purified fresh patient samples of myeloma cells, suggesting that the results presented here are not due to a trait that developed after prolonged in vitro culture of the cell lines. 3 The JJN-3 cell line was unique among the tested cell lines in the amount of HGF produced. JJN-3 cells grow easily under serum-free conditions, a feature that simplifies the isolation of secreted proteins. Thus, HGF can be purified with high yield and purity by a simple two-step chromatographic procedure. HGF is synthesized as a single polypeptide chain that has to be proteolyticaly processed into the respective ␣and ␤-chains to become biologically active (27,38). Recently, the isolation and characterization of serum proteases, which specifically cleaves the HGF precursor, was reported (29,30). However, it is still unclear whether these particular proteases are expressed in myeloma cells. When JJN-3 cells were grown under serum-free conditions, a substantial part of the HGF in the supernatant could be reduced into ␣and ␤-chains, suggesting that a protease with the ability to cleave HGF precursor is associated with these cells.
In this paper we also show that HGF is able to counteract the biological effect of TGF␤ on Mv-1-Lu mink lung cells. This release from TGF␤-induced growth arrest by HGF has recently also been observed by others (39). The inhibition of TGF␤ by HGF can be used as a sensitive and reliable bioassay for HGF (40). There are several possibilities for intracellular crosstalk between the TGF␤ and HGF signaling pathways. One of the intracellular effects of c-MET is activation of a Ras protein, resulting in a shift in the equilibrium between Ras-GTP and Ras-GDP toward the active, proliferation-inducing GTP-bound state (41). From studies of Mv-1-Lu cells, it is known that TGF␤ maintains p21 ras in the GDP-bound state (42). This prevention of Ras activation was crucial for TGF␤-induced growth arrest of Mv-1-Lu cells and activation of p21 ras was required for progression into S-phase after cell cycle arrest by TGF␤ (42). Ras proteins are therefore a possible point of intersection for the intracellular pathways of HGF and TGF␤ signaling.
Although HGF counteracted the activity of TGF␤ on the Mv-1-Lu cells, further experiments showed that these cytokines have unrelated or cooperative effects in other cell systems. Thus, when the human FS-4 fibroblastic cell line was cultured for 4 days, both TGF␤ and HGF stimulated the DNA synthesis in these cells in an additive manner (data not shown). Furthermore, TGF␤ causes inhibition of the IL-2-induced proliferation of the murine T-cell line HT-2 (43). We tried to reverse this TGF␤-mediated activity, but HGF did not seem to interfere with this effect of TGF␤ (data not shown). c-MET activation is known to favor cell invasion and migration and to cause angiogenesis and proliferation, all of which are essential in tumor progression. Several of these aspects could be important in multiple myeloma. In vivo, myeloma cells are often found in the bone marrow of most of the skeleton, usually associated with destruction of the bone substance surrounding the cells. This bone destruction is thought to be due to a combination of enhanced bone resorption and diminished 3 Børset, M., Hjorth-Hansen, H., Seidel, C., Sundan, A., and Waage, A. (1996) Blood, in presss. bone formation (44,45). TGF␤ has been proposed as a factor which preserves the balance in bone remodeling by controlling that increased resorption is counteracted by both a rise in bone formation and a decrease in bone resorption (46). HGF was recently shown to promote formation of osteoclasts from hematopoietic precursor cells (47), to attract osteoclasts to sites of bone resorption (48), and, in co-culture with osteoblasts, to increase the level of resorption (47,48). However, whether HGF affects bone remodeling in multiple myeloma directly or as an opponent to TGF␤ remains to be tested. We have tested whether the HGF/c-MET receptor-ligand interaction is involved in proliferation in an autocrine or paracrine manner in our myeloma cell lines. However, in a series of experiments with various concentrations of HGF or anti-HGF we have not been able to show such effect on cell proliferation. 4 Nevertheless, the fact that concomitant expression of the ligand and receptor seems to be the rule in multiple myeloma cell lines, whereas it probably is an exception in other hematological cancers, certainly merits further investigation of the role of this cytokine-receptor pair in multiple myeloma.