SCATTER FACTOR PROMOTES MOTILITY OF HUMAN GLIOMAAND NEUROMICROVASCULAR ENDOTHELIAL CELLSKatrin LAMSZUS1, Nils Ole SCHMIDT2, Liang JIN1, John LATERRA3, David ZAGZAG4, Dennis WAY5, Marlys WITTE5, Martin WEINAND5,Itzhak D. GOLDBERG1, Manfred WESTPHAL2 and Eliot M. ROSEN1*1Department of Radiation Oncology, Long Island Jewish Medical Center, New Hyde Park, NY, USA2Department of Neurosurgery, University Hospital Eppendorf, Hamburg, Germany3Departments of Neurology, Oncology, and Neuroscience, The Johns Hopkins School of Medicine and Kennedy-Krieger ResearchInstitute, Baltimore, MD, USA4Department of Surgical Neuropathology, New York University Medical Center, New York, NY, USA5University of Arizona College of Medicine, Department of Surgery, Tucson, AZ, USA

Malignant gliomas are characterized by rapid growth,infiltration of normal brain tissue, and high levels of tumor-associated angiogenesis. The genetic and local environmentaltissue factors responsible for the malignant progression fromlow to high grade gliomas and the highly malignant behaviorof glioblastomas are not well understood. In a study of 77human brain tissue extracts, high grade (III-IV) tumors hadsignificantly greater scatter factor (SF) content than did lowgrade tumors or non-neoplastic tissue. To investigate thepotential significance of SF accumulation in gliomas, wemeasured the effects of SF on DNA synthesis and motility ofcultured human glioma cell lines. SF stimulated DNA synthe-sis in 7/10 glioma cell lines and in 3/3 neuromicrovascularendothelial cell (NMVEC) lines, consistent with our previousreport that SF stimulated cell proliferation of a few humanglioma cell lines. SF markedly stimulated the chemotacticmigration of 10/10 glioma cell lines as well as 3/3 NMVEClines. In addition, SF stimulated the 2-dimensional migrationof glioma cells on culture surfaces coated with specificextracellular matrix molecules (collagen IV, laminin, andfibronection). As expected based on these biologic responsesto SF, 10/10 glioma lines and 4/4 NMVEC lines expressedmRNA for c-met, the SF receptor. To assess the possible invivo significance of these migration assays, we compared thechemotactic response of a glioma cell line to human braincyst fluids and tumor extracts that contained high or low SFconcentrations. Fluids and extracts with high SF contenttended to induce higher levels of chemotactic migration thandid fluids and extracts with low SF content. Addition ofanti-SF monoclonal antibody (MAb) inhibited migration in-duced by fluids and extracts with high SF content by about30–50%. Int. J. Cancer 75:19–28, 1998.r 1998 Wiley-Liss, Inc.

Astrogliomas are common primary brain tumors that displayvariable histologic characteristics and degrees of biologic aggres-siveness (Devauxet al.,1993; Thomas, 1990). Low grade astrocy-tomas are characterized by their slow growth and lack of prominentvascular proliferation relative to the high grade and malignantanaplastic astrocytomas and glioblastoma multiforme. The gliomacells that comprise these tumors typically invade surrounding brainto sites that are frequently very distant from the initial tumor. Whileglioma cell invasion occurs in all gliomas of astroglial andoligodendroglial lineage, the extent of invasiveness tends tocorrelate with other histologic criteria of tumor grade (e.g.,vascular proliferation, nuclear pleomorphism, necrosis). Excep-tions to this include: 1) the pilocytic astrocytomas that despitecertain aggressive histologic characteristics such as nuclear pleo-morphism and vascular proliferation remain well demarcated fromsurrounding brain; and 2) the entity of gliomatosis cerebri that canconsist of glioma cells of low histologic grade that aggressively anddiffusely invade brain, frequently without forming a discrete mass(3). It is this invasive characteristic that limits effective gliomatreatment using localized modalities such as surgery, radiosurgery,and interstitial chemotherapy (Devauxet al.,1993; Thomas, 1990).The biochemical determinants of glioma cell migration and inva-sion within the brain remain poorly defined.

In vitro model systems have identified cytokine growth factors,cell surface and extracellular matrix adhesion molecules (Izumotoet al., 1996; Okadaet al., 1996), and proteolytic enzyme systems(Gladsonet al.,1995; Mohanamet al.,1993) that influence gliomacell migration and invasion. Epidermal growth factor, fibroblastgrowth factor, and platelet-derived growth factor have been foundto stimulate glioma cell migration and invasiveness to variousdegrees, depending upon the tumor cell line examined (Lund-Johansenet al.,1992; Merzaket al.,1994; Pedersenet al.,1994).Scatter factor (SF) [hepatocyte growth factor (HGF)] is a multifunc-tional cytokine that stimulates angiogenesis (Grantet al.,1993) andthe motility and invasion of heterogeneous cell populations such asepithelial, carcinoma, and sarcoma cells bothin vivo and in vitro(Bhargavaet al., 1992; Gherardiet al., 1989; Rosenet al., 1989,1994; Stokeret al.,1987). These activities have led us to examineSF expression by human gliomas and its effects on properties thatare attributed to glioma malignancy (Rosenet al.,1996). Previousstudies show that human gliomas and glioblastoma cell linesexpress SF and/or its specific cell surface receptor protein, thec-met gene product (Moriyamaet al., 1995, 1996; Rosenet al.,1996). We have also shown that SF can stimulate cell proliferationof some human glioblastoma cell linesin vitro (Rosenet al.,1996).

This manuscript describes the effects of purified and glioma-derived SF on thein vitro migration of human glioblastoma andbrain microvascular endothelial cell lines. Our results provideadditional evidence that the SF/c-met signalling system pathwayenhances glioblastoma cell motility and, therefore, may play a rolein stimulating glioma cell migration through brain.

MATERIAL AND METHODS

Human brain tumor tissue and cyst fluid samples

Freshly obtained brain tumor biopsy or cyst fluid samples orfreshly thawed tissue and fluid samples were obtained from theJohns Hopkins Hospital Brain Tumor Bank, the Department ofSurgical Neuropathology at New York University Medical Center,and the Department of Neurosurgery, University Hospital Eppen-dorf, Hamburg, Germany. Tissue samples were cut into small

Contract grant sponsor: United States Public Health Service; Contractgrant numbers: R01-CA64869 and R01-NS32148; Contract grant sponsor:American Cancer Society; Contract grant number: EDT-102; Contract grantsponsor: Children’s Brain Tumor Foundation of New York.

Dr. Lamszus’ Current address: is: Department of Neuropathology,University Hospital Eppendorf, Martini Strasse 52, D-20246 Hamburg,Germany.

*Correspondence to: Department of Radiation Oncology, Long IslandJewish Medical Center, 270-05 76th Avenue, New Hyde Park, New York11040, USA. Fax (718)-470-9756.

Received 24 May 1997; Revised 29 July 1997

Int. J. Cancer:75,19–28 (1998)

r 1998 Wiley-Liss, Inc.

Publication of the International Union Against CancerPublication de l’Union Internationale Contre le Cancer

pieces; homogenized in extraction buffer (20 mM Tris, pH 7.5, 0.5M NaCl, 0.1 mM phenylmethylsulfonyl fluoride) (4–6 ml ofbuffer/g of tissue); sonicated; and clarified by microfuging. Theprecipitates were re-extracted in the same buffer with 1 M NaCl.High salt buffers were used to extract SF bound to heparan sulfateproteoglycans in the matrix and on cells. The 2 clarified superna-tants were pooled, ultrafiltered with a YM-30 Amicon (Beverly,MA) membrane to remove the excess salt, and made up to theoriginal volume with phosphate-buffered saline. The protein concen-trations of the extracts were determined using the BioRad (Rich-mond, CA) Coomassie blue dye-binding assay. All brain tumorspecimens were histologically classified by a board-certified patholo-gist at the institution of origin.

Scatter factor ELISAImmunoreactive SF was quantitated using a sensitive double

antibody ELISA, as described before (Grantet al.,1993; Rosenetal., 1996). The assay was specific for SF. It did not detectplasminogen, albumin, or various other growth factors and cyto-kines. The standard curve was linear from 0.2 to 4.0 ng/ml of SF(useful measuring range). All test samples were serially diluted sothat one or more of the measurements was within the linearmeasuring range. For tissue samples, SF content values wereexpressed as ng of SF/mg of protein. For cyst fluids, SF values wereexpressed as ng/ml.

Cell linesTen different glioma cell lines (G28, G84, G111, G112, G118,

G121, G122, G124, G260, U373MG) were studied. The first 9glioma lines were isolated by us from malignant grade IV gliomas(glioblastomas), as described before (Ankeret al., 1993). Gliomacell line U373MG was obtained from the ATCC (Rockville, MD).These cell lines were utilized between passages 13 and 90. Fourdifferent lines of neuromicrovascular endothelial cells (NMVEC3,NMVEC4, NMVEC5, and NMVEC9) were isolated from dis-carded temporal lobe tissue from patients who had undergonetemporal lobectomy for intractable epileptic seizures, as describedearlier (Rosenet al., 1996). These cells were studied betweenpassages 6 and 10.

Cell cultureAll cell lines were cultured in Dulbecco’s modified Eagle’s

medium (DMEM) supplemented with FCS (10% v/v), non-essential amino acids (0.1 mM), L-glutamine (5 mM), penicillin(100 U/ml), streptomycin (100 µg/ml), and fungizone (0.25 µg/ml)(all from Bio Whittaker, Walkersville, MD), as described before(Rosenet al., 1996). Cells were subcultured at weekly intervalsusing trypsin, and seeded into fresh medium at 1:4 (NMVECs) or1:10 (glioma cells). Cultures were incubated at 37°C in a humidi-fied atmosphere of 5% CO2 and 95% air.

SF and anti-SF antibodyBiologically active two-chain recombinant human SF was

provided by Dr. R. Schwall, Genentech, South San Francisco, CA.Anti-SF neutralizing mouse MAb MO294 was purchased fromR & D Systems (Minneapolis, MN).

Chemotaxis assaysGlioma and NMVEC cell chemotaxis in response to SF, glioma

cyst fluids, or tumor extracts was assayed using a modifiedmicrowell Boyden chamber assay. The chemoattractant to be testedwas added to the lower wells of a 96-well Boyden chamber(Neuroprobe, Cabin John, MD). The wells were covered with an8-µm pore size Nucleopore filter that had previously been coatedwith 100 µg/ml of Vitrogen (Celtrix, Palo Alto, CA). Cells weredetached with trypsin, washed twice by centrifugation, counted byhemocytometer, and resuspended at 13 106 cells/ml in DMEMcontaining 0.1% BSA (DMEM-BSA). Cells (23 104 cells in 50 µlof DMEM-BSA) were seeded into the upper wells, and thechamber was incubated at 37°C for 5 hr. The chamber was thendisassembled, and the non-migrated cells were scraped off the

upper side of the filter. The filter was stained using Diff QuickChemicals (Baxter, W. Sacramento, CA). Migrated cell nuclei werecounted using a 403 objective and a calibrated ocular grid, andvalues were recorded as migrated cells per ten 4003 grids. Allvalues were assessed in triplicate.

DNA synthesis assaysCells were seeded into 24-well plates at 43 104 cells per well

and allowed to grow to about 80% of confluence in completeculture medium. Cells were then washed 3 times with DMEM andincubated for 48 hr in serum-free DMEM (0.5 ml per well). Themedium was then replaced with 0.5 ml of fresh DMEM containingdifferent concentrations of recombinant human SF or containing10% FCS (positive control). The plates were incubated for 24 hr,after which (3H)-thymidine [methyl-3H] (NEN, Boston, MA) wasadded to each well to a final concentration of 2 µCi/ml. Cells wereincubated for 3 hr to allow incorporation of (3H)-thymidine intocellular DNA. They were then washed twice with ice-cold phos-phate-buffered saline (PBS); incubated in 5% TCA at 4°C for 30min, washed once with TCA; washed again with PBS, and lysed byaddition of 0.2 ml of 0.1 N NaOH. Cell lysate (0.15 ml) wascounted using a liquid scintillation counter, and the remaininglysate was used for determination of protein content using theBioRad Coomassie blue dye-binding assay. Incorporation of TCA-insoluble counts into cell DNA was expressed as CPM per µgprotein, mean6 SEM of quadruplicate determinations.

Fence migration assayMigration of glioma and NMVEC cells on different substrates in

the absence and presence of SF was determined using a fencemigration assay similar to that described earlier (Gieseet al.,1994;Prattet al.,1984). Forty-eight well tissue culture plates were coatedwith BSA or with the following human extracellular matrixproteins: collagen IV (Collaborative Research, Bedford, MA),fibronectin (Sigma, St. Louis, MO), laminin (Sigma), or tenascin(Sigma). Wells were coated by incubation with 100 µg/ml ofprotein in sterile PBS for 90 min at 37°C. Plates were then washedwith PBS, blocked with 1% BSA in PBS for 1 hr at 25°C; andrinsed twice with PBS. Sterile glass cloning cylinders with an outerdiameter of 6 mm (Bellco, Vineland, NJ) were placed in the centerof each well. Cells were detached with trypsin, washed, andresuspended in complete culture medium at 13 106 cells/ml.Thirty microliters of cell suspension (33 104 cells) was added toeach glass cylinder. Plates were placed on ice for 30 min tofacilitate attachment of the cells, and then incubated for 7 hr at37°C. The cylinders were removed, and 0.5 ml of fresh culturemedium containing FCS (5% v/v), hydroxyurea (10 mM) [toinhibit cell proliferation], and either 20 ng/ml of SF or no SF wasadded to the wells. Three different diameters of the area covered bythe cells were measured after 0, 24, 48, and 72 hr, using a gradedrecticule (with 100 recticule bars equaling about 2.5 mm on thewell). The distance migrated by the cells was calculated as theradius of the cell area minus the initial radius at time 0. All assayconditions were tested in triplicate.

RT-PCR analysisExpression of c-metmRNA by glioma and NMVEC cell lines

was assayed by RT-PCR analysis. Total cellular RNA (1 µg) wasreverse transcribed using 10,000 U/ml Superscript RNaseH2Moloney murine leukemia virus reverse transcriptase (GIBCOBRL, Gaithersburg, MD), 2.5 mM oligo dT, 5 mM MgCl2, and 13PCR buffer consisting of 50 mM KCI, 10 mM Tris-HCl, pH 8.3, 1mM dNTPs, and 1 U/ml RNAse inhibitor. The mixture wasincubated at 42°C for 1 hr, heated to 99°C for 5 min to denaturereverse transcriptase, and cooled at 5°C for 5 min. cDNA from thereverse transcription reaction was subjected to PCR in the presenceof 0.15 mM and each of 58 and 38 primers, 1.25 U of Taqpolymerase (Perkin-Elmer, Norwalk, CT), 2 mM MgCl2, and 13PCR buffer. PCR was performed in a DNA thermal cycler(Perkin-Elmer) for 35 cycles, each consisting of 95°C for 1 min and60°C for 1 min. After the 35 cycles, there was a time delay for 7

20 LAMSZUS ET AL.

min at 72°C. The reaction products were visualized by ethidiumbromide staining following 1.5% agarose gel electrophoresis. Thesense and antisense primers, respectively, and the predicted sizes ofthe RT-PCR reaction products were as follows:

c-met:58-ACAGTGGCATGTCAACATCGCT-38,

58-GCTCGGTAGTCTACAGATTC-38, 655 bp.

b-actin: 58-TTGTAACCAACTGGGACGATATGG-38,

58-GATCTTGATCTTCATGGTGCTAGG-38, 764 bp.

Method of data analysisFor experimentalin vitro assays, results were calculated as

means6 standard errors of the mean (SEMs) and compared usingthe two-tailed Student’st-test. For human brain tumor tissuesamples, the SF content values did not appear to follow a normaldistribution, but rather showed a tail containing some extremelyhigh values, as is commonly the case with measurements made onhuman biologic samples. Therefore, comparison of SF contentwithin these tumor samples was performed using a non-parametrictest (Mann-Whitney U-test) (Freund, 1992).

RESULTS

Expression of SF in human tumor samplesBrain tumor tissue.To determine if the SF content of human

brain tumors was correlated with the biologic aggressiveness of thetumor, we measured SF titers in extracts of brain tumor tissue andnormalized these values to tissue protein content (ng of SF/mg ofprotein). Among 74 tissue samples from tumors of glial, astrocytic,or ependymal type, high grade (III-IV) tumors had about 3-foldhigher SF content than did low grade (I-II) tumors: 1.736 0.43ng/mg vs. 0.58 6 0.15 ng/mg (Table I). This difference wasstatistically significant (p , 0.03, Mann-Whitney U-test). Thehigh grade tumors exhibited more heterogeneity of SF content,indicated by a larger range of values and a higher standard error,than did the low grade tumors. Four temporal lobe specimens frompatients who had undergone surgical resection for epilepsy werealso examined. The mean SF content of these non-neoplasticsamples was slightly lower (0.496 0.12 ng/mg) but not signifi-cantly different from that of the low grade tumors.

Brain cyst fluid.We also assayed SF concentration in a numberof samples of fluids obtained from cysts associated with varioustypes of brain tumors and in cyst or cerebrospinal fluid from

non-neoplastic conditions. Fluids from these non-neoplastic condi-tions contained much less SF than did fluids from oligodendroglial,astroglial, ependymal, and other tumor types (Table II). However,the SF concentrations were not significantly different in cysts fromhigh gradevs. low grade tumors (Table II). For samples in whichsufficient fluid remained after assay of SF, the protein concentrationwas measured and used to normalize SF content (ng SF/mgprotein). Even after normalization to fluid protein, there was nosignificant difference in the cyst fluid SF content of these highvs.low grade tumors (data not shown). Interestingly, some of thehighest SF titers were observed in juvenile pilocytic astrocytomas,which are low grade tumors that have a very good prognosisfollowing surgical resection.

Chemotactic activity of SF for human gliomaand neuromicrovascular endothelial cell lines

To assess the biologic activity of SF on cell types found inhuman brain tumors, we utilized standard Boyden chamber assaysto determine if purified SF stimulates the migration of humanglioma and NMVEC cell lines. In these assays, SF showedsignificant chemotactic activity for 10/10 glioma cell lines and 3/3NMVEC cell lines (Fig. 1). Maximum migration responses wereobserved at 5 or 20 ng/ml. In some cases, levels of migrationdecreased significantly below the peak level at the higher doses ofSF (e.g.,G111, G112, G121, G122). Basal levels of migration inthe absence of SF varied among the different cell lines, with somelines showing little or no unstimulated migration and othersshowing high basal rates of migration (e.g.,G118 and G260). Theseresults establish that SF is a potent chemoattractant for both humanglioma cells and human brain capillary endothelial cells.

Two-dimensional migration of human glioma and NMVEC celllines on different substrates

We utilized a fence migration assay to measure the outwardmovement of glioma and endothelial cells on wells coated withBSA, collagen IV, fibronectin, or laminin, as described earlier(Gieseet al.,1994; Prattet al.,1984). Assays were also attemptedusing wells coated with tenascin, but could not be completedbecause the cells adhered very poorly to this substrate. Assays wereperformed in the presence (20 ng/ml) or absence of SF, withhydroxyurea added to prevent cell proliferation. In these assays, SFstimulated the migration of 4/4 glioma cell lines (G28, G122,G124, and U373) (Figs. 2, 3, Table III). However, SF did notstimulate 2-dimensional migration of NMVECs on any substrate.For the glioma cell lines, SF-induced migration was observed on all

TABLE I – SF CONTENT VALUES OF BRAIN TUMOR EXTRACTS1

Tumor typeSF content (ng SF/mg protein)

Mean6 SEM (N) Median Range

High grade (III–IV) tumorsGlioblastoma multi-

forme1.836 0.67 (27) 0.60 0.0–16.5

Anaplastic astro/malig-nant glioma

1.636 0.89 (10) 0.77 0.26–9.5

Malignant oligodendro-glioma

1.546 0.56 (10) 1.16 0.01–5.6

All high grade 1.736 0.43 (47) 0.72 0.0–16.5Low grade (I–II) tumors

Low grade astrocytoma/glioma

0.816 0.29 (8) 0.54 0.0–2.6

Benign oligodendro-glioma

0.496 0.16 (16) 0.20 0.0–2.1

Ependymoma 0.326 0.32 (2) 0.32 0.0–0.63All low grade 0.586 0.15 (26) 0.37 0.0–2.6

Non-tumor tissueTemporal lobe 0.496 0.12 (4) 0.35 0.27–1.0

1Tissue extracts were prepared and SF content was measured asdescribed in Material and Methods. Tabulated values represent means6SEMs (number of samples). Statistical comparison of the SF content of‘‘all high grade’’ vs. ‘‘all low grade’’ tumors gavep , 0.03 (MannWhitney U-test).

TABLE II – SF CONTENT OF BRAIN CYST FLUIDS1

Sample tested SF concentration (ng/ml)

Non-tumor 0.26 0.1 (9)High grade (III–IV) tumors

Glioblastoma multifome 7.46 2.5 (17)Malignant astocytoma/mixed glioma 6.26 1.4 (17)Malignant oligodendroglioma 7.06 1.626)Malignant ependymoma 4.66 1.8 (3)

Low grade (I–II) tumorsJuvenile pilocytic astrocytoma 10.86 3.5 (12)Low grade astrocytoma 6.56 3.4 (5)Ependymoma 1.06 0.6 (6)

Other tumor typesHemangioblastoma 3.56 1.4 (5)Metastasis 3.16 0.6 (4)Ganglioglioma 5.4 (1)Acoustic neuroma 0 (1)

1Values tabulated are means6 SEMs (number of samples). The‘‘non-tumor’’ diagnoses were as follows: arachnoid cyst (N5 2);subdural hygroma (N5 2); ruptured arterio-venous malformation(N 5 1); cervical cavernoma (N5 1); cerebrospinal fluid from CTmyelogram (N5 1); cerebrospinal fluid from shunt infection (N5 1);and cerebrospinal fluid from patient with degenerative joint disease ofthe spine (N5 1).

21SCATTER FACTOR PROMOTES GLIOMA MOTILITY

FIGURE 1 – Chemotactic response of human glioma and NMVEC cells to SF. Chemotaxis assays were performed in microwell modifiedBoyden chambers using recombinant human SF, as described in Material and Methods. Values shown are means6 SEMs of triplicatedeterminations.

22 LAMSZUS ET AL.

4 substrates. The absolute levels of migration in the absence orpresence of SF were usually highest on laminin, but the ability ofSF to stimulate glioma cell migration did not appear to be substratespecific.

Glioma cell chemotactic activity in cyst fluidsand brain tumor tissue

Our findings that SF is present in human glioma tumor tissue(especially in high grade gliomas) and that SF induces chemotacticmigration of cultured human glioma cells raise the possibility thatthe presence of SF may enhance the motility and invasiveness ofglioma tumor cellsin vivo.To further evaluate the potential role of

SF in stimulating the motile and invasive phenotype of gliomas, weperformed Boyden chamber migration assays to determine 1)whether biologic fluids and tissues from brain tumors and non-neoplastic conditions contained chemotactic factors for culturedglioma cells; and 2) if this chemotactic activity could be attributedto the presence of SF. These assays were performed usingSF-responsive glioma cell line G122. In each set of assays, theresponsiveness to purified recombinant human SF was also as-sessed in the same assay as a control.

Brain cyst fluids.Cyst fluids were assayed at a dilution of 1:4 or1:3, and the results were tabulated according to whether the cyst

FIGURE 2 – Fence migration assays showing the response of human glioma cell line G122 to SF. Assays were performed in wells coated withdifferent substrates, as indicated in the figure (see Material and Methods for details). Values plotted represent the radial distance migrated by thecells (arbitrary units) at the indicated time after removal of the barrier. Values are means6 SEMs of quadruplicate determinations.

23SCATTER FACTOR PROMOTES GLIOMA MOTILITY

FIGURE 3 – Photographs of fence migration assays showing the response of human glioma cell line G124 to SF. Assays were performed in wellscoated with different substrates, as described in the Material and Methods section. Wells were photographed 3 days after removal of the barrier.Panels on the left(a,c,e,g)show assays in the absence of SF; and panels on the right(b,d,f,h)show assays in the presence of SF (20 ng/ml). Thesubstrates tested were BSA(a,b); collagen IV(c,d); laminin (e,f); and fibronectin(g,h).

24 LAMSZUS ET AL.

fluid had low or high SF concentration. Most cyst fluids inducedglioma cell migration relative to the control medium (DMEM-BSAonly). On average, cyst fluids with high SF concentrations induceda 2.7-fold increase in chemotactic migration as compared withfluids with low SF content (p , 0.04; Table IV). More importantly,the addition of a neutralizing anti-SF monoclonal significantlyreduced migration levels for fluids with high SF content (p , 0.01)but had little or no effect on migration levels for fluids with low SFcontent (Table IV).

Brain tumor tissue.Six glioma tumor tissue extracts wereassayed for chemotactic activity for glioma cells, each at a fixedextract protein concentration of 100 µg/ml. All extracts inducedmigration levels about 2.6-fold or greater relative to the DMEM-BSA control (Table V). Addition of anti-SF MAb reduced themigration levels for glioma extracts with high SF content by 3366%, as compared with 56 4% for glioma extracts with low SFcontent (p , 0.02 for comparison of the effect of antibody onmigration induced by extracts with highvs.low SF content). Theseresults suggest that SF contributes to the glioma cell chemotacticactivity when present, but is not the only chemoattractant forglioma cells in tumor tissue or tumor-associated cystic fluid.

Expression of c-met receptor by human gliomaand NMVEC cell lines

Since all of the cell lines studied appeared to be biologicallyresponsive to SF in assays of cell migration, these cell lines shouldexpress c-met, which is the only known receptor for SF (Bottaroetal., 1991). We tested 10 glioma cell lines and 4 NMVEC lines forthe presence of c-metmRNA using RT-PCR analysis (see Materialand Methods). Each of the cell lines tested showed an easilydetectable band corresponding in size to the expected fragment ofc-met(Fig. 4). We had previously demonstrated that human gliomacells (Rosenet al., 1996) and vascular endothelial cells (Grantetal., 1993) express immunoreactive c-met proteinin vivo. Thus, itappears likely that both glioma and brain capillary endothelial celltypes are biologically responsive to SF through the canonical c-metreceptor.

Effect of SF on DNA synthesis by glioma and NMVEC cell linesSubconfluent cells were exposed to SF for 24 hr after 2 days of

serum-starvation, and then (3H)-thymidine incorporation into DNAwas assessed during a 3-hr pulse (Table VI). Among 10 glioma celllines, 7 showed some degree of SF-induced stimulation of DNAsynthesis, with maximum DNA synthesis levels ranging from 1.5to 4.5 times that of control: G28, G84, G111, G118, G121, G260,U373MG (Table VI). Three cell lines showed either no change or amodest decrease in DNA synthesis in response to SF: G112, G122,

and G124. For one of these lines (G124), the levels of (3H)-thymidine incorporation were essentially nil. All 3 NMVEC celllines showed stimulation of DNA synthesis induced by SF, withmaximal effects ranging from 1.9 to 3.6 times the control values.The maximum SF-induced DNA synthesis in responsive cell lineswas usually achieved at SF concentrations of 20 or 100 ng/ml.

Cell lines for which DNA synthesis was not stimulated by SFalso showed very little stimulation of DNA synthesis after additionof 10% FCS. Prior results suggested that glioma cells may continueto proliferate at low concentrations of serum (Rosenet al.,1996).In another experiment, DNA synthesis by G84 glioma cells wasmeasured with or without the DNA synthesis inhibitor hydroxyurea(10 mM). With hydroxyurea present, (3H)-thymidine incorporationlevels for cells treated with 0, 20, and 100 ng/ml of SF (1096 9,1146 4, and 1426 4 CPM per µg protein, respectively) werelower than the corresponding values obtained without hydroxyurea(2096 10, 2566 8, and 4466 18 CPM per µg protein, respec-tively). The reduction in DNA synthesis induced by hydroxyureasuggests that quiescence may not be fully achieved in these celllines under the conditions of the assay (i.e., after 2 days of serumstarvation).

DISCUSSION

We show in this report that SF can be found within primarygliomas and glioma cyst fluid regardless of histologic grade, aswell as within non-neoplastic cortex. Importantly, despite theheterogeneity of SF content of glioma extracts, a statisticallysignificant correlation between increasing glioma grade and SFcontent was found. Whereas no difference was found in the SFcontent of non-neoplastic brain and low grade gliomas, the meanSF content of high grade gliomas was approximately 3-fold higherthan that of low grade gliomas. This finding is consistent with amechanistic relationship between SF expression and the develop-ment of certain characteristics of the malignant phenotype inhuman gliomas. It is interesting that the SF content of cyst fluidswas not found to differ significantly between tumor grades. Thismay reflect: 1) different rates of turnover of proteins that aresecreted into cysts relative to those secreted into interstitial tumorfluid; or 2) biological differences between the cells that line tumorcystic cavities relative to cells that comprise the bulk of tumor. Thelatter possibility is supported by the finding of elevated SF in cystfluid of pilocytic astrocytomas, whose cyst walls are frequentlyhypervascular relative to other regions within this tumor.

We have shown that purified recombinant human SF stimulatesthe motility of 10 out of 10 human glioma cell lines using bothchemotaxis assays and fence migration assays that measure two-dimensional cell movement along surfaces coated with specificextracellular matrix proteins. These findings are consistent with theexpression by all of these glioblastoma cell lines of the SF receptorc-metmRNA (Moriyamaet al., 1995, 1996; and present results).Our findings are also in agreement with a previous report indicatingthat SF induces a dose-dependent chemotactic response in 3 humanglioma cell lines (Moriyamaet al.,1996).

Most important are our findings that glioma cyst fluids and tumorextracts with high SF content stimulated significantly greaterglioma cell migration than did fluids and extracts with low SFcontent. Furthermore, about 30–50% of the motility-inducingactivity of SF-containing glioma cyst fluids was specificallyattributed to SF. Thus, SF retains this biological activity within thecontext of more biochemically complex fluid endogenous to humangliomasin vivo.

While the mean motility-inducing activity was significantlygreater in cyst fluids with highvs.low SF content, we did find somefluids with high SF content and low motility-inducing activity. Thisfinding might reflect the presence of an inhibitor of SF and/ordenaturation or partial degradation of SF. Our ELISA does notdistinguish the inactive single-chain precursor SF (pro-SF) fromthe biologically active heterodimeric (2-chain) SF, which is gener-

TABLE III – EFFECT OF SF ON FENCE MIGRATION ASSAYS1

Cell lineSF

concentration(ng/ml)

Distance migrated by day 3(arbitrary units)

BSA Coll IV Laminin Fibronectin

G28 0 226 0 396 1 536 1 406 020 406 1 546 1 746 2 536 1

G122 0 276 1 216 3 346 3 286 120 436 1 426 3 516 0 436 0

G124 0 256 1 326 1 426 2 306 120 486 0 556 1 716 2 496 1

U373MG 0 456 2 286 1 426 1 366 120 526 1 456 2 616 2 526 2

NMVEC3 0 No Att. 276 1 276 1 296 220 No Att. 286 2 296 3 276 2

NMVEC4 0 No Att. 386 0 326 3 396 320 No Att. 426 1 346 0 386 1

1BSA 5 bovine serum albumin; coll IV5 collagen type IV; SF5scatter factor. Assays were performed as described in Material andMethods. NMVEC cells did not attach to wells coated with BSA (‘‘NoAtt.’’), and thus could not be assayed on BSA. All values are means6SEMs of triplicate determinations.

25SCATTER FACTOR PROMOTES GLIOMA MOTILITY

ated by proteolytic cleavage during tissue injury (Miyazawaet al.,1994). Thus, it is also possible, in these cases, that the SF detectedby the ELISA was predominantly the inactive single-chain form.Conversely, in a few cyst fluid samples, we found relatively highlevels of motility-inducing activity despite the absence of SF. Thesefluids probably contain glioma cell motility-inducing factors otherthan SF.

The ability of SF to stimulate glioblastoma cell migrationequally well on laminin, fibronectin, and collagen IV substratessuggests that SF has the potential to regulate migration along tissuepathways such as blood vessels, subependyma, and white matter

tracts that are relevant to thein vivo spread of glioma cells. It is ofinterest that SF was also found to be a potent chemoattractant forbrain capillary endothelial cells in Boyden chamber assays but hadno effect on endothelial cell migration in the fence two-dimensional migration assay. These findings are consistent with arole for glioma cell-derived SF in the chemoattraction of brainendothelial cells during tumor neovascularization, in addition to theautocrine and/or paracrine stimulation of glioma cell migration.

Several distinct properties of SF might explain its effect onglioma malignancy. We previously reported that purified recombi-nant human SF can stimulate the proliferation of human glioblas-

TABLE IV – CHEMOTAXIS OF G122 GLIOMA CELLS INDUCED BY CYST FLUID6 ANTI-SF ANTIBODY (AB)1

Sample testedSF

content(ng/ml)

Chemotactic migration(cells/10 grids)

No AB 1AB % Inhibition

Cyst fluids with low SF contentZ34 Arachnoid cyst 0.9 186 2 176 1 6Z40 Low grade astryocytoma 0 426 2 486 2 0Z42 Subdural hygroma 0.7 556 4 366 3 35

Z119 CT myelogram 0 506 8 386 3 24Z157 Arachnoid cyst 0 0 0 0Z160 Ependymoma 0 646 2 656 4 0S32 Hemangioblastoma 0.4 106 1 106 2 0

Mean values 0.36 0.15 346 9 316 9 9 6 5Cyst fluids with high SF content

Z28 Recurrent GBM 9.8 1896 10 376 3 80Z36 Recurrent GBM 6.7 1486 3 246 3 84

Z148 Recurrent GBM 7.4 616 1 426 1 31Z160 Malignant ependymoma 7.0 146 2 146 2 0Z169 GBM 8.6 1006 14 516 1 49S35 GBM 14.0 706 4 466 3 34S36 Grade III astrocytoma 16.3 796 6 176 2 78S37 Grade III glioma 18.1 86 1 2 6 1 (75)S49 GBM 40.2 1696 9 516 3 68

Mean values 14.26 3.5 936 21 326 6 536 11Control assays SF (ng/ml)

0 216 4 186 4 141 1746 7 566 3 685 5296 6 556 4 90

20 7516 28 996 2 87

1GBM 5 glioblastoma multiforme. Cyst fluids were assayed at a 1:4, except for Z119, Z157, Z148, andZ169, which were assayed at a 1:3 dilution. Anti-SF MO294 MAb was used at 20 µg/ml. Migration valuesare means6 SEMs of triplicate assays. Cyst fluids with high SF content gave higher levels of chemotacticmigration than did cyst fluids with low SF content cyst:p 5 0.037 (two-tailedt-test). Addition of anti-SFMAb inhibited migration to a significantly greater extent for cyst fluids with high SF content than for thosewith low SF content:p , 0.01.

TABLE V – CHEMOTAXIS OF G122 GLIOMA CELLS INDUCED BY GLIOMA EXTRACTS6 ANTI-SF ANTIBODY1

Sample tested SF content(ng/mg)

Migration (cells/10 grids)

No AB 1AB % inhibition

Glioma extracts with low SF content20 Grade II oligodendroglioma 0 2406 6 2576 18 024 Grade II mixed glioma 0 2276 23 2206 21 333 Grade II astrocytoma 0 2606 20 2306 6 12Mean values 0 2426 10 2366 11 56 4

Glioma extracts with high SF content12 Grade IV oligodendroglioma 5.6 2736 22 2106 21 2315 Grade III astrocytoma 9.5 4436 20 2536 26 4323 Recurrent GBM 16.5 5636 9 3756 5 33Mean values 10.56 3.1 4266 84 2796 49 336 6

Control assays SF (ng/ml)0 876 3 956 5 00.5 1436 12 976 8 335 4176 3 976 9 77

20 7306 25 1636 9 78

1GBM 5 glioblastoma multiforme. Tumor extracts were assayed at concentration of 100 µg/ml ofextract protein. Anti-SF MAb MO294 (‘‘AB’’) was used at 20 µg/ml. Migration values are means6 SEMsof triplicate assays. Comparison of the mean percentage of inhibition of SF-induced chemotactic migrationof glioma extracts with highvs.low SF content yieldedp , 0.02 (two-tailedt-test).

26 LAMSZUS ET AL.

toma cells in vitro by autocrine and/or paracrine mechanisms.Consistent with these prior findings are our present data showingthat SF stimulates DNA synthesis in a different series of humanglioblastoma cell lines. Thus, SF might function as a humanglioblastoma cell mitogenin vivo. SF is known to be a potentangiogenic factorin vivobased on its ability to induce neovascular-ization in the rat cornea and in mouse subcutaneous tissue (Grantet

al., 1993). Our present finding that SF stimulates DNA synthesisand chemotaxis in multiple human brain microvessel endothelialcell lines further suggests that SF can induce angiogenic processesin human brain and in brain neoplasms. The biological relevance ofthesein vitro responses of endothelial cells and glioma cells topurified SF are supported by our recent finding in a rat gliomamodel that SF gene transfer to glioma cells increases intracranial

FIGURE 4 – Expression of c-metmRNA by human glioma and NMVEC cell lines. Semi-quantitative RT-PCR analysis of total cell RNA (1 µgper lane) was performed as described in Material and Methods. Upper panel shows ethidium bromide stained bands corresponding to the 655 bpamplified fragment of c-met; and lower panel shows stained bands corresponding to the 764 bp amplified fragment ofb-actin. The cell linesstudied were as follows: lane 1, G28; 2, G84, 3, G111; 4, G112; 5, G118; 6, G121; 7, G122; 8, G124; 9, G260; 10, U373MG; 11, NMVEC3; 12,NMVEC4; 13, NMVEC5; 14, NMVEC9.

TABLE VI – EFFECT OF SF ON3H-THYMIDINE INCORPORATION BY GLIOMA AND NMVEC CELLS1

Cell line3H-thymidine incorporated into cell DNA (CPM per µg protein)

0 1 5 20 100 ng/ml SF 10% FCS

G28 2796 13 2546 2 3916 14 406 15 4926 8 9726 27G84 1816 10 3196 29 3626 0 3806 15 8236 64 7896 54G111 3266 2 3536 8 3716 4 4296 6 5926 10 7316 8G112 4026 8 3356 4 2946 30 3336 16 3256 61 5606 10G118 2126 10 2296 12 516 10 3416 28 2756 17 2566 11G121 4506 15 7386 61 9496 49 12766 53 11096 28 28336 103G122 6056 13 7006 34 7306 8 5686 9 5566 9 6856 12G124 196 2 166 1 136 2 116 1 9 6 1 226 2G260 1856 2 2606 8 2666 8 3796 7 3576 10 10686 27U373MG 3116 7 2606 15 4176 42 4666 14 4166 39 4816 23NMVEC3 1566 1 1906 6 2486 2 2936 13 2946 4 4716 8NMVEC4 2196 1 2466 9 2646 4 3256 5 4456 8 7016 8NMVEC5 1336 4 2456 18 3206 16 4716 23 4826 12 7236 19

1FCS 5 fetal calf serum; SF5 scatter factor. Assays were performed as described in Material andMethods. Values listed are means6 SEMs of quadruplicate detrminations.

27SCATTER FACTOR PROMOTES GLIOMA MOTILITY

glioma growth and glioma-associated angiogenesis when com-pared to SF-negative control tumors (Laterraet al., 1997a).Similarly, U373MG human glioma cells transfected with SF cDNAgrew more rapidly than control cells in immunocompromised mice(Laterraet al.,1997b).

Based on these results, we postulate that the expression of c-metreceptor by glioma cells and brain endothelial cells in conjunctionwith the accumulation of SF promotes the malignant progression ofgliomas by promoting: 1) the loco-regional dissemination ofglioma cells; 2) glioma cell proliferation; and 3) glioma-associatedangiogenesis. Previous studies using cultured glioma cells, animaland human xenograft glioma models, and human glioma specimenshave identified various biological factors that contribute to themalignant phenotype of these tumors (Weingartet al.,1996). Theseinclude factors with direct effects on glioma cell proliferation andindirect effects secondary to the stimulation of glioma-associatedangiogenesis. Variable subsets of these factors, including SF, arelikely to co-exist and function either additively or synergistically

to, influence the malignant behavior of these heterogeneoustumors.

ACKNOWLEDGEMENT

This work was supported, in part, by United States Public HealthService research grants R01-CA64869 (EMR) and R01-NS32148(JL), by American Cancer Society research grant EDT-102 (EMR),and by a research grant from the Children’s Brain Tumor Founda-tion of New York (EMR). KL was supported in part by a grant fromthe Monika Kutzner-Stiftung, Berlin; and KL and MW weresupported by the Deutsche Forschungsgemeinschaft (WE 928/2-1).We thank Dr. P. Burger, Department of Pathology, Johns HopkinsSchool of Medicine, for his help in obtaining and reviewing humanbrain tumor specimens; and Dr. F. Lenz, Department of Neurosur-gery, Johns Hopkins School of Medicine, for help in obtainingtemporal lobe specimens. This work contains major parts of adoctoral thesis by NOS, to be submitted to the FachbereichMedizin, Hamburg.

REFERENCES

ANKER, L., OHGAKI, H., KUDEKE, B.I., HERRMANN, H-D., KLEIHUES, P. andWESTPHAL, M., p53 protein accumulation and gene mutations in humanglioma cell lines.Int. J. Cancer,55,982–987 (1993).

BHARGAVA, M. and 13OTHERS, Scatter factor and hepatocyte growth factor:Activities, properties, and mechanism.Cell Growth Differentiation,3,11–20 (1992).

BOTTARO, D.P., RUBIN, J.S., FALETTO, D.L., CHAN, A.M-L., KMIECIK, T.E.,VANDE WOUDE, G.F. and AARONSON, S.A., Identification of the hepatocytegrowth factor receptor as the c-met proto-oncogene product.Science,251,802–804 (1991).

DEVAUX, B.C., O’FALLON, J.R. and KELLY, P.J., Resection, biopsy, andsurvival in malignant glial neoplasms.J. Neurosurg.,78,767–775 (1993).

FREUND, J.E.,Mathematical Statistics(5th ed.), Prentice Hall, EnglewoodCliffs, NJ (1992).

GHERARDI, E., GRAY, J., STOKER, M., PERRYMAN, M. and FURLONG, R.,Purification of scatter factor, a fibroblast-derived basic protein whichmodulates epithelial interactions and movement.Proc. nat. Acad. Sci.(Wash.),86,5844–5848 (1989).

GIESE, A., RIEF, M.D., LOO, M.A. and BERENS, M.E., Determinants ofhuman astrocytoma migration.Cancer Res.,54,3897–3904 (1994).

GLADSON, C.L., PIJUAN-THOMPSON, V., OLMAN , M.A., GILLESPIE, G.Y. andYACOUB, I.Z., Up-regulation of urokinase and urokinase receptor genes inmalignant astrocytoma.Amer. J. Pathol.,146,1150–1160 (1995).

GRANT, D.S., KLEINMAN , H.K., GOLDBERG, I.D., BHARGAVA, M., NICKOLOFF,B.J., POLVERINI, P. and ROSEN, E.M., Scatter factor induces blood vesselformation in vivo.Proc. nat. Acad. Sci. (Wash.),90,1937–1941 (1993).

IZUMOTO, S., OHNISHI, T., ARITA, N., HIRAGA, S., TAKI , T. and HAYAKAWA ,T., Gene expression of neural cell adhesion molecule L1 in malignantgliomas and biological significance of L1 in glioma invasion.Cancer Res.,56,1440–1444 (1996).

LATERRA, J., NAM, M., ROSEN, E.M., RAO, J.S., LAMSZUS, K. and JOHNSTON,P., Scatter factor/hepatocyte growth factor gene transfer enhances gliomagrowth and angiogenesisin vivo. Lab. Invest.,76,565–577 (1997a).

LATERRA, J., ROSEN, E., NAM, M., RANGATHAN, S., FIELDING, K. andJOHNSTON, P., Scatter factor/hepatocyte growth factor expression enhanceshuman glioblastoma tumorigenicity and growth.Biochem. biophys. Res.Comm.,235,743–747 (1997b).

LUND-JOHANSEN, M., FORSBERG, K., BJERKVIG, R. and LAERUM, O.D.,Effects of growth factors on a human glioma cell line during invasion intorat brain aggregates in culture.Acta Neuropathol. (Berlin),84, 190–197(1992).

MERZAK, A., MCCREA, S., KOOCHECKPOOR, S. and PILKINGTON, G.J.,

Control of human glioma cell growth, migration, and invasion in vitro bytransforming growth factor beta 1.Brit. J. Cancer,70,199–203 (1994).MIYAZAWA , K., SHIMOMURA, T., NAKA , D. and KITAMURA , N., Proteolyticactivation of hepatocyte growth factor in response to tissue injury.J. biol.Chem.,269,8966–8970 (1994).MOHANAM, S., SAWAYA , R., MCCUTCHEON, I., ALI-OSMAN, F., BOYD, D. andRAO, J.S., Modulation of thein vitro invasion of human glioblastoma cellsby urokinase-type plasminogen activator receptor antibody.Cancer Res.,53,4143–4147 (1993).MORIYAMA , T., KATAOKA , H., SEGUCHI, K., TSUBOUSHI, H. and KOONO, M.,Effects of hepatocyte growth factor (HGF) on human glioma cellsin vitro:HGF acts as a motility factor in glioma cells.Int. J. Cancer,66, 678–685(1996).MORIYAMA , T., KATAOKA , H., TSUBOUSHI, H. and KOONO, M., Concomitantexpression of hepatocyte growth factor (HGF), HGF activator and c-metgenes in human glioma cellsin vitro. FEBS Lett.,372,78–82 (1995).OKADA , H., YOSHIDA, J., SOKABE, M., WAKABAYASHI , T. and HAGIWARA, M.,Suppression of CD44 expression decreases migration and invasion ofhuman glioma cells.Int. J. Cancer,66,255–260 (1996).PEDERSEN, P.H., NESS, G.O., ENGEBRAATEN, O., BJERKVIG, R., LILLEHAUG,J.R. and LAERUM, O.D., Heterogeneous response to the human growthfactors [EGF, PDGF (bb), TGF-alpha, bFGF, IL-2] on glioma spheroidgrowth, migration, and invasion.Int. J. Cancer,56,255–261 (1994).PRATT, B.M., HARRIS, A.S., MORROW, J.S. and MADRI, J.A., Mechanisms ofcytoskeletal regulation. Modulation of aortic endothelial spectrin by theextracellular matrix.Amer. J. Pathol.,117,349–354 (1984).ROSEN, E.M., GOLDBERG, I.D., KACINSKI, B.M., BUCKHOLZ, T. and VINTER,D.W., Smooth muscle releases an epithelial cell scatter factor which bindsto heparin.In Vitro Cell Develop. Biol.,25,163–173 (1989).ROSEN, E.M., LATERRA, J., JOSEPH, A., JIN, L., FUCHS, A., WAY, D., WITTE,M., WEINAND, M. and GOLDBERG, I.D., Scatter factor expression andregulation in human glial tumors.Int. J. Cancer,67,248–255 (1996).ROSEN, E.M., NIGAM, S.K. and GOLDBERG, I.D., Mini-Review: Scatterfactor and the c-met receptor: A paradigm for mesenchymal:epithelialinteraction.J. Cell Biol.,127,1783–1787 (1994).STOKER, M., GHERARDI, E., PERRYMAN, M. and GRAY, J., Scatter factor is afibroblast-derived modulator of epithelial cell mobility.Nature (Lond.),327,238–242 (1987).THOMAS, D.G.T. (ed.),Neuro-oncology,Johns Hopkins University Press,Baltimore (1990).WEINGART, J.D., LATERRA, J.J. and BREM, H., Growth factors and angiogen-esis. In: W.K.A. Yung (ed.),Clinical Neurology, International Practice andResearch(Vol. 5, Cerebral Gliomas), pp. 307–318, Baillie`re Tindall,London (1996).

28 LAMSZUS ET AL.