migration of brain tumor cells on extracellular matrix ...areexpressedin the basallamina surrounding...

[CANCER RESEARCH56. 1939-1947. April 15. 19961

migration and proliferation by mediating interactions between cellsand their environment. CAMs and ECM molecules are involved incritical cellular interactions during neural morphogenesis and histogenesis (I , 2) and are frequently reexpressed after injury or disease(3). CAMs can be divided into three major families of molecules (I,4, 5). These include the immunoglobulin superfamily, the calciumdependent family of CAMs termed cadherins, and integrins, a familyof heterodimeric (a, j3) receptors that require calcium for their binding. Integrins bind primarily to ECM molecules and in some instances

to other CAMs (6). Cells expressing the same cadherin are able tosegregateand remain separated from other cells differing from themin cadherin expression (7, 8), suggesting that loss of cadherin functionmay favor tumor invasiveness. Consistent with this view, tumormetastasis and invasiveness are often correlated with decreased adhesivity among tumor cells and reduced expression of cadherins (9).Lower adhesivity and enhanced migration rates have also been correlated with lower expression of the integrin a5@31(10). However,increased expression of other integrins, e.g., aj33 (1 1), can increasetumor cell invasiveness, indicating that there is no simple correlationbetween integrin expression and invasiveness. Integrins are involvedin various fundamental processesthat affect cell migration, includingcell growth and differentiation (6). Accordingly, integrin-mediatedmigration depends on complex interactions among different signaltransduction pathways that regulate cell behavior (12).

Components of the brain ECM including cytotactin/tenascin-C,laminin, thrombospondin, vitronectin, and fibronectin are synthesizedand secreted by astrocytes during development and are involved inadhesion and migration of neurons and glia during development (1,13â€”15).In addition to their interactions with integrins and otherCAMs, they can form complexes with other ECM molecules including proteoglycans. The controlled expression and assembly of ECMcomponents and their modulation by factors such as proteasesmay beimportant mechanisms for modulating cell migration in vivo (16).Several ECM proteins are expressed transiently at elevated levelsduring development, and subsequently they are found at relatively lowlevels in the mature brain (17).

Some ECM proteins are up-regulated in human brain tumors andhave been implicated in tumor cell migration and metastasis. Inaddition to their functional relevance, adhesion molecules that areselectively expressed in brain tumors are potential targets for therapeutic agents by virtue of being located on cell surfaces or in theextracellular space (18). Fibronectin, laminin, and type IV collagenare expressedin the basal lamina surrounding blood vessels in normaland neoplastic brain. Tenascin-C is present in human astrocytomaswhere its increased expression is correlated with higher grade tumors(19â€”21)as well as with angiogenesis (22, 23). In vitro studies usingglioma cell lines indicate that brain tumor cells can use some of theseECM moleculesas favorablesubstratesfor migration (24, 25).

Although it is clear that adhesion molecules play important roles incell migration, the identity of the major ECM proteins and cell surfacereceptors that promote human brain tumor migration in situ remains

1939

Migration of Brain Tumor Cells on Extracellular Matrix Proteins in VitroCorrelates with Tumor Type and Grade and Involves a, and

Integrins'

David R. Friedlander,2 David Zagzag, Bronya Shiff, Henry Cohen, Jeffrey C. Allen, Patrick J. Kelly, and

Martin Grumet

Department of Pharmacology [D. R. F.. B. S., M. G.J. Department of Neurology ID. R. F.. J. C. Al, Brai,: Tumor Research Center ID. R. F., D. Z.. J. C. A., P. J. K.. M. G.J.Department of Pathology, Division of Neuropathology ID. 1], and the Kaplan Comprehensive Cancer Center fD. Z. J. C. A.. P. J. K.. M. G.J. Department of Neurosurgery [B. S..P. J. K.J, and the Clinical Research Center and Environmental Medicine (H. C.], New York University Medical Center, New York, New York /0016

ABSTRACT

An important contributor to the malignancy of brain tumors is theirabifity to infiltrate the brain. Extracellular matrix molecules and celladhesionmoleculeson cell surfacesplay key roles in cell migration. In thepresent study, we used reaggregates of dissociated cells from freshlyexcised human brain tumors to analyze the migration ofcells from humanbrain tumors of different types and grades on many different adhesionproteinsadsorbedto glasssubstrates.Proteinswerechosenbasedontheirpresence in normal or neoplastic nervous tissue, and Included the extracellular matrix molecules fibronectin, collagens, fibrinogen, laminin, teanacin-C, thrombospondin, and the neuron-glia cell adhesion molecule,Ng-CAM. Cells from astrocytomas (n 24) migrated on a vanety ofsubstrates, in contrast to cells from primifive neuroectodermal tumorscells(n = 6), which only migrated well on laminin, fibronectin, or type IVcollagen but not on the other substrates. Typically, migrating cells fromastrocytomas of all grades had long, slender processes, were usuallybipolar, and their cell bodies did not spread well on any substrate.

Although there was variability in the migration ofcells from astrocytomasof the same grade, cells from high-grade astrocytomas tended to migratemore extensively (42.3 Â±4.7 pm/16 h; n 16) than cells from lower gradeastrocytomas (28.9 Â±3.9 jun/16 h; P 0.07; n 8); the most strikingdifferences were observed for collagen substrates, on which cells from

lower grade astrocytomas migrated at very low levels (7.6 Â±2.6 @.tm/16h)and cells from high-grade astrocytomas at higher levels (24.4 Â±5.2 @sm/16h; P 0.01). In contrastto primary cellsfrom glioblastomas(n = 13),glioblastoma cell lines (n = 10) consistently spread on various substratesand migrated at high levels (69.5 Â±7.6 versus 46.4 Â±5.7 pin/16 h;P = 0.03), in particular, on collagens (108.4 Â±20.2 versus 28.0 Â±61@zml16h; P = 0.001). Specific monoclonal antibodies to a@and@ integrin

monomers completely inhibited the migration of astrocytoma cells onmost substrates, suggesting that a@and I3@integrins play a crucial role inbrain tumor Infiltration. These studies also suggest that although a largenumber of extracellular matrix molecules may promote tumor cell migradon, disrupting the function of only a few tumor cell receptors may becritical for tumor infiltration In the bram

INTRODUCTION

A key contributor to the pathology of brain tumors is their ability toinfiltrate the parenchyma diffusely. Surgical excision of a malignantbrain neoplasm is not sufficient to eliminate neoplastic cells that haveinfiltrated normal adjacent brain. It is therefore of primary importanceto determine the mechanisms involved in brain tumor cell migration.

CAMs3 on cell surfaces and ECM molecules play key roles in cell

Received12/5/95;accepted2/16/96.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 with18 U.S.C. Section 1734 solely to indicate this fact.

I This work was supported by grants from NIH (NS 3 1088), The Children's Brain

Tumor Foundation, and The Long Island League Against Cancer.2 To whom requests for reprints should be Addressed, at Department of Pharmacology,

New York University Medical Center, New York, NY 10016.3 The abbreviations used are: CAM, cell adhesion molecule; ECM. extracellular

matrix; mAb, monoclonal antibody; GFAP, glial fibrillary acidic protein; Ng-CAM,neuron-glia CAM; PNET, primitive neuroectodermal tumor.

Association for Cancer Research. by guest on August 23, 2020. Copyright 1996 Americanhttps://bloodcancerdiscov.aacrjournals.orgDownloaded from

https://bloodcancerdiscov.aacrjournals.org

/@@x@/MIGR.ATIONOF BRAIN TUMOR CELLS

poorly understood. The present study was initiated to identify adhesion proteins and the mechanisms by which they promote migratorybehavior in the brain. We have approached the problem by identifyingadhesion proteins that are able to promote migration of human braintumor cells obtained from resected tumors of different types andgrades and using specific antibodies to integrins on tumor cell surfaces to inhibit migration of brain tumor cell lines.

MATERIALSANDMETHODS

Reagents. Chicken Ng-CAM was purified from 14-dayembryonic chickenbrains using immunoaffinity chromatography (26). Human tenascin-C was agenerousgift from Dr. Keith Langley (Amgen,ThousandOaks,CA). Cornmercial reagents included mouse laminin, rat type I collagen, and mouse typeIV collagen (Collaborative Biomedical Products, Bedford, MA), polylysine,

human fibrinogen (Sigma, St. Louis, MO), hurnan fibronectin (New York

BloodCenter,New York, NY), recombinantentactin(UpstateBiotechnology,Inc., Lake Placid, NY), and humanthrombospondin-1(Life Technologies,Grand Island, NY). Purified mAbs specific for human integrins included:anti-131 550034, clone mAb 13; anti-a2 550026, clone P1E6; anti-a3 550027,

cloneP1B5;anti-f34550037,cloneAA3 (CollaborativeBiomedicalProducts);and anti-a,, 1980 (Chemicon, Temecula, CA). Antibodies for immunohistochemistry included polyclonal anti-GFAP and anti-factor VIII (DAKO,

Carpinteria, CA) and anti-Ki-67 mAb MIB-l (AMAC, Westbrook, ME).

Tissue culture media were obtained frorn Life Technologies, CollaborativeBiomedical Products, and Biowhittaker (Walkersville, MD).

Tumor Specimens. Fragments from excised tumors or from stereotactic

biopsies were obtained at the operating room, placed on ice, and brought to the

laboratory.Sampleswereprocessedimmediately,andassayswereperformedwithout knowledge of the pathological diagnosis. Migration assays were performed on excess tissue that was not utilized for pathological examination. Theclassification scheme used for the grading of astrocytomas was based on the

new WHO guidelines (27): grade I, juvenile pilocytic astrocytoma (n 4);grade II, low-grade astrocytoma (n = 4); grade III, anaplastic astrocytoma(n 3); and grade IV, glioblastoma multiforme (n = 13). Juvenile pilocyticastrocytomas showed low to moderate levels of cellularity, often with â€œcornpactâ€•and â€œspongyâ€•areas. Numerous Rosenthal fibers were observed. Lowgrade astrocytomas showed hypercellularity and mild nuclear pleomorphismand no necrosis. Anaplastic astrocytomas showed moderate to high hypercellularity and moderateto marked nuclear pleomorphismand no necrosis.Glioblastomasrevealed moderateto markedcellular pleomorphism,hyperchromaticnuclei,andnumerousmitotic figures.Focalareasof necrosiswith orwithout pseudopalisading cells were seen in all cases.PNETs (n 6) exhibitedhighly cellular sheetsof undifferentiatedandmostlysmall tumorcells.Thesehad hyperchromatic, often irregular, nuclei and scant cytoplasm. Numerousmitotic figures and apoptotic cells were seen. The PNETs included fourcerebellarmedulloblastoma,one pineoblastoma(i.e., PNET arising in thepineal gland), and one cerebral neuroblastoma (i.e., PNET arising in the

cerebral hemisphere).

Cell Lines. GlioblastomasA-172, Hs683, SW 1088, SW 1783, U-87 MG,U-l18 MG, U-138 MG, U-251 MG, U-373 MG, and T98G, and medulloblastomasD283,D341,andDaoywereobtainedfrom theAmericanTypeCultureCollection and were cultured in DMEM/lO% FCS.

Migration Assay. In preliminary experiments, fragments offreshly excised

astrocytomatumorsplacedon substratescoatedwith adhesionmoleculesusuallyattachedto fibronectinandlaminin substrates,andcellsmigratedfromthefragmentsontothesesubstrates.However,thefragmentsattachedpoorlyornot at all to certain proteins, including BSA and entactin,which made itdifficult to evaluate cell migration on these substrates, given that migrationrequires attachment. To improve conditions for analyzing migration on varioussubstrates, tumor cell reaggregates were used as a source of migrating cells

(Fig. I). Tissues were minced with scalpels and incubated in 0.25% trypsinhl

mM EDTA for 20 mm at 37Â°C. FCS and DNase I were added to a final

concentration of 10% and 50 @g/ml,respectively, and the tissues were trite

rated with a Pasteur pipette. Remaining large fragments were pelleted bylow-speed centrifugation, and cells in the supernatant were washed three times

in DMEM and resuspended in DMEM/ITS@ (insulin, 6.3 @g/ml;transfemn,

6.3 @g/ml;seleniousacid,6.3ng/ml;BSA, 1.3mg/nil; linoleicacid.5.4 @Wml;

Tumor@Biopsy 9

0Â°000000

00 0

Collaborative Biomedical Products). To form reaggregates, 10 ml cell suspen

sion in 50 ml tissue culture flasks were shaken in a rotatory environmentalincubatorovernightat 37Â°C.Reaggregateswerecollectedby low-speedcentrifugation, centrifuged at low speed through a 3.5% BSA/PBS step gradient,

and resuspended in DMEM/ITS@. The diameter of these purified reaggregatesranged from 60 to 200 @m.Approximately 70% of processed tumors yielded

reaggregates; specimens that did not yield reaggregates were found to be

mostly infiltrated white matteror normaltissueuponhistopathologicalexamination. Trituration of these specimens generated large amounts of myein

fragments,which may havecontributedto hamperingthe formation and/orisolation of reaggregates from white matter tumors. Marginal regions of

resections that were judged to be normal by pathological analysis yielded veryfew cells.

To verify that reaggregates from astrocytoma tumors contained tumor cells,they were fixed and analyzed histologically (six cases). Tumor cells werereadily recognized in all cases (data not shown). H&E staining characteristics

of theneoplasticcellsincludednuclearhyperchromasia,highnuclear:cytoplasmic volume ratio, and scant or absent cytoplasmic processes. No cells withfeatures of reactive astrocytes (i.e., nuclear euchromasia, low nuclear:cyto

plasmic volume ratio, and elongatedand abundantcytoplasmicprocesses)were seen. It was apparent by H&E staining alone that the cells were neoplasticastrocytes, and immunohistochemistry performed in one case confirmed that

the cells were astrocytomas (data not shown). Cells in the reaggregates werepositive for GFAP like the astrocytoma cells themselves. The GFAP immunoreactivity showed scant processes, consistent with that of astrocytoma cellsand not reactive astrocytes. Approximately 30% of the cells were stained bythe antibody MIB-l, which recognizes the nuclear proliferation-related antigenKI-67. This percentageis characteristicof high-gradegliomaswhich normallyexhibita 10â€”40%rangeof MIB-l-stainedcells(22)andisconsistentwiththenotionthatreaggregatescontainedalargeproportionof tumorcells.Cellswere

aggregate

(16@@

dissociate

trypsin +trituration

Aggregates plated on wells coated with adhesion proteins

A

Fig. 1. Migration assay. In A, tumor biopsies were obtained at the operating room andcut into a small numberof fragments(I and II). Fragmentsusedfor migrationassaysweredissociated and reaggregated overnight to yield reaggregates, as described in â€œMaterialsand Methods.â€• Reaggregates were placed in wells of multiwell slides and incubatedovernight.Migration wasquantitatedas 0.5 times the differencebetweenthe diameter(arrowheads)ofthe region occupied by emigrating cell bodies at the final time(C, t = 18h) andthe reaggregatediameterat the initial time (B, t = 2 h). Bar, 100 @sm.

1940



MIGRATION OF BRAIN TUMOR CELLS

negative for factor VIII-related antigen, a marker for endothelial cells, mdicating that the reaggregates did not contain endothelial cells. These observations confirmed that the aggregates consisted primarily of tumor cells.

For migration assays, 15 pA of a 300/ml suspension of reaggregates inDMEMIITS@were addedto each2.5-mm well of 26-well slides (Cd-lineAssociates, Inc., Newfield, NJ) coated with proteins and incubated at 37Â°C.

Reaggregates and emigrating cells were analyzed through brightfield andphase microscopy. To determine migration from individual reaggregates, cultureswerephotographedafterthereaggregateshadattachedto thesubstrate(2h in vitro) and before cells from different aggregates intermingled (18 h in

vitro). Migration was quantitated as 0.5 times the difference between thediameter of the region occupied by emigrating cell bodies at the final time andthe reaggregate diameter at the initial time (Fig. 1). When fragments from braintumorsandreaggregatedcells from thesamebrain tumorswerecomparedfortheir behavior on fibronectin and laminin substrates, their cells resembled eachother both in their shape and extent of migration (data not shown), suggestingthat the proceduresusedin preparingreaggregatesdo not alter the migratorybehavior of tumor cells. However, when these reaggregates were placed on

BSA or entactin substrates, they attached much better than tumor fragments,

which facilitated our ability to monitor cell emigration on these substrates.Moreover, quantitation was easier, and the variability was lower for reaggregates than for fragments. Therefore, they were used for migration studies. Toprepare aggregates from cell lines, cells growing as monolayers were broughtto suspensionwith 0.25%trypsin/1mxiEDTA andprocessedfurtherusingthesamestepsfollowed for reaggregatingdissociatedtumor cells. All cell linesyielded aggregates.

To determinewhetheraggregatesize had an influenceon migration, themigration of reaggregatesof all primary glioblastomatumors(13 cases)onfibronectin substrates was analyzed. Scatter plots of migration versus aggregate size at the initial time showed no effect of aggregate size on migration(data not shown). To evaluate potential effects of cell proliferation, U-251 MG

cell aggregates were placed on fibronectin or type I collagen substrates in theabsence or presence of 10 mM hydroxyurea, which blocks proliferation without

affectingthemotility ofbrain tumorcellsin vitro (28).No effectsof hydroxyurea were observed, indicating that cell proliferation does not affect migrationas evaluated using the present assay (data not shown).

For inhibition experiments, stock solutions ofpurified mAbs were diluted inmedium,and2 @.tlofthe dilute solutionwereaddedperwell. mAbswereaddedafter taking the initial set of micrographs (t = 2 h). To determine the dosedependency of the mAb concentration, migration assaysusing U-25l MG andU-373 MG cell aggregates on fibronectin,type IV collagen, or fibrinogensubstrates were performed in the presence of different concentrations of anti-a@

and anti-/31mAbs, ranging from 0 to 10 @.tg/m1.Lowest concentrations thatexhibitedsaturatingeffectswere 5 @tg/mi(@3@)or 4 @.tg/ml(a@).Theseconcentrations were used for the subsequent experiments. The same mAb often

had strikingly different effects on the migration of brain tumor cells ondifferent substrates [e.g., compare effects of anti-@1 on fibronectin versuslamininaswell astheeffectsof anti-a@on lamininversusfibrinogenonU-373MG cells (Fig. 7)], indicating that inhibition of cell migration was not due

simply to a toxic effect of the mAbs. Furthermore, when U-251 MG cellaggregatesweregrown on laminin in the presenceof anti-@1mAb for 16h,which caused inhibition of migration, removal of soluble mAb resulted insignificant migration following an additional 16-h incubation period, showingreversibility of mAb inhibition. A similar experiment, using U-25l MG cell

aggregates on fibrinogen, showed reversibility of inhibition of migration byanti-a@mAb.

Substrates. Twenty-six-well slides were washed with deionized water,sterilizedin a microwaveovenfor 4 mm,andincubatedwith 5 p1/wellof 100,.tg/ml adhesion proteins in PBS for 30 mm. Wells were washed with PBS,blocked with 1% BSA/PBS for 1 h, and washed with DMEMIITS@.

Immunohistochemistry. Procedures were performed as previously described (22). In brief, 4-,.@msections were cut from paraffin blocks, and thendeparaffinized and rehydrated. Standard immunostaining methods using theavidin-biotin-horseradish peroxidase complex were used to localize the anti

body-bound antigen with diaminobenzidine as chromogen. Prior to incubationwith antibody, sections to be stained with MIB-l were microwaved to retrievethe antigen.

Statistics. Variables measured on a continuous scale were evaluated usingStudent'st testor the Mann-WhitneyU procedure.Whenevaluatingpropor

tions with an expected value 5 in any cell, Fisher's exact test was used.When more than two groups were tested, ANOVA or the Kruskal-Wallisprocedure was performed. When multiple comparisons were made, the TukeyKramer procedure was used. Significance was deemed to be at P = 0.05. Allprocedures used SAS v6.lO (SAS Institute, Cary, NC).

RESULTS

Reaggregatesfrom brain tumors of different types and grades weretested for their ability to migrate on many adhesion proteins adsorbedto glass substrates. Proteins tested included the ECM molecules fibronectin, type I collagen, type IV collagen, laminin, tenascin-C, andthrombospondin and the neuron-glia CAM, Ng-CAMIL1, which arenormally present in brain at least at certain periods during development or in neoplastic brain. Fibnnogen was also used because it isnormally present in circulating blood and may leak out from newlyformed vesselsof brain tumors which have an incomplete blood-brainbarrier (29, 30). In addition, entactin was tested due to its associationwith lamimn in basementmembranes (31), and polylysine was exammed because it commonly supports cell adhesion in vitro; BSA wasused as a nonadhesive protein control.

Morphology of Migrating Brain Tumor Cells in Vitro. Reaggregates prepared from different types of brain tumors attached to wellscoated with adhesion proteins but the shapeand extent of migration ofcells from aggregates varied among different tumors and adhesionproteins. In most cases, migrating cells from astrocytomas of allgrades exhibited long, slender processeswhich were usually bipolarand tended to be oriented radially with respect to the reaggregate(Fig.2A). Typically, a high proportion of cell bodies emigrated relative tothose remaining in the aggregates. In some cases, long processesradiated from reaggregatesbut cell bodies did not emigrate (Fig. 2B),indicating that cell migration and process extension are not directlyrelated. In most cases, cell bodies did not spread much on anysubstrate, including fibronectin (Fig. 2A), although they were somewhat spread on Ng-CAM (Fig. 2C) and polylysine (data not shown).Distinct cell spreading was observed only among cells from glioblastomas (5/13 analyzed, Table 1; Fig. 2D), in contrast to lower gradeastrocytomas, which were not spread (11 analyzed, Table 1). Theseobservations suggest that the ECM may interact more strongly withthe cytoskeleton of cells from glioblastomas than that of cells fromlower grade astrocytomas.

Reaggregates of PNET cells attached wealdy to most substrates,and cell emigration on these substrates was very poor. The onlysubstratesthat were found to stimulate significant levels of PNET cellmigration were laminin, fibronectin, and type IV collagen (seebelow).In thesecases,migrating cells were small, asexpected for PNET cells,and their processes were usually short (Fig. 2E). Spread cells wereobserved in only one of six casesanalyzed (Table 1).

Migration Patterns. To analyze tumor cell migration, the extentof cell migration on the different substrates was measured for eachtumor (Fig. 1) and plotted as a column graph. Thus, each tumor wasrepresentedby a pattern of columns (the â€œmigrationpattern,â€•e.g., Fig.3). Because migration assays were performed with reaggregated cellsfrom tumor fragments, we tested the extent to which migration patterns depended on sampling. When specimens from different regionsof the sametumor were analyzed, very similar migration patterns wereobserved (1 1/12 cases). For example, Fig. 3 shows pairs of specimens(A and B) from a glioblastoma and a PNET which exhibited verysimilar migration patternswithin the pairs. These observationsmdicate that the migration pattern was a reliable representation of tumorcell behavior.

Astrocytomas versus PNETs. Comparisons of cell migration patternsrevealedstriking differencesbetweendifferent typesof tumors

1941



-@ Ã˜@JGBM/FN

@@ .LGA/LM

GBM/NgCAM

4@

C-GBM/CoIIV

@.

@ E@Ã¸

@BM Iine/FN'

:Â° ::,â€”@

@..â€”.

@ .- F

Table 1 Spreading of cells from brain twnorreaggregatesTumor

type and/orgradeNo.Â°PNET

Juvenilepilocytic astrocytomaLow grade astrocytomaAnaplasticastrocytomaGlioblastomaGlioblastoma cell lines1/6

0/40/40/35/13b

10/10b


variety of substrates (Figs. 3 and 4), in contrast to cells from PNETswhich migrated very well (>80 @.am/l6h) on only one substrate (4/6cases,e.g., laminin in Fig. 3) or two substrates(1 case,PNET-73; datanot shown) but very poorly or not at all (<20 @.tm/l6h) on the othersubstrates.Laminin (three cases),fibronectin (two cases),and type IVcollagen (one case)were the only substratesthat stimulated significantmigration of cells from PNETs (Fig. 4). Cells from PNET-73 migrated>150 p.m/16 h on both laminin and type N collagen substrates,andwere also remarkable in that they had a spread shape (Table 1).Immunohistochemical analysis of formalin-fixed sections from thisparticular tumor revealedthat it was positive for GFAP (data notshown), indicating that it was a PNET with astrocytic differentiation.This casewas a recurrent tumor; the specimen analyzed for migrationwas from the third resection of the tumor which took place 3 monthsafter the previous surgery.

Cells from astrocytomas (pooled juvenile pilocytic astrocytomas,low-grade astrocytomas, anaplastic astrocytomas, and glioblastomas)and PNETs were compared for their ability to migrate on all substrates. Cells from astrocytomas migrated on average more extensively (31.9 Â±2.9 @m)than cells from PNETs (13.5 Â±4.1 @m;P < 0.001; Fig. 5A). Individual substrates that selectively stimulatedmigration of cells from astrocytomas at statistically significant levelsincluded fibrinogen (P 0.018), Ng-CAM (P = 0.023) and polylysine (P = 0.012). The combined results suggest that cells fromastrocytomas may expresson their surfacesa wider selection of activeadhesion molecule receptors than cells from PNETs.

Astrocytomas of Different Grades. Differences were also observed in the averagemigration patterns of cells from astrocytomas of

a Numerators represent the number of tumors whose emigrating cells exhibited a

spreadshapeon oneor moreof theculturesubstrates.Denominatorsindicatethenumberoftumors analyzedfor theindicatedcategory.Glioblastomacell lineswereSW 1088,SW1783, U-87 MG, U-118MG, U-138 MG, U-251 MG, U-373 MG, A-172, 1980, andHs683.

b Difference between glioblastoma and glioblastoma cell lines is statistically signifi

cant (P = 0.003;Fisher'sexacttest).

Fig. 2. Morphology of cells from primary brain tumors migrating in vitro. A, cellsemigrating from glioblastoma (GBM) reaggregates onto a fibronectin (FN) substrate,illustrating typical shape of primary astrocytoma cells in vitro. Note the slender longprocesses and the unspread phase-bright cell bodies. B, reaggregate of low grade astrocytoma (LGA) on laminin (LM) substrate as an illustration of cases in which extensiveprocessesradiatedfrom reaggregateswithout cell bodyemigration.C,emigrationof cellsfrom glioblastoma on Ng-CAM. Phase dark cell bodies are indicative of partial spreading.D, emigration of glioblastoma cells on type IV collagen (Col IV); this is an example ofthe few caseswhere primary tumor cells spread well in vitro. E, migrating PNET cells ona fibronectinsubstrate.F, emigratingcells from a glioblastomaline (U-138MG) reaggregate on fibronectin. These cells exhibit the typical spread morphology observed formigrating astrocytoma cell lines. The center of the aggregate is located to the left of thefield shown.Bar, 100 @m.

and to a lesser extent among tumors classified as being of the sametype and different grade (Fig. 4). Certain generalizations emergedwhen comparing the migratory behavior of tumors of different types.For example,cells from astrocytomaswere able to migrateon a wide

Fig. 3. Similarity of migration of cells from different regions of the same tumor.Migration assayswere performed on reaggregatesobtained from different fragments of aPNET (PNET35A and PNET35B) and a glioblastoma (GBM57A and GBM57B) specimen. Note the similarity betweenthe migration patternsobtained for fragmentsofindividual tumors. The negativemigration values observedfor the PNET representcompactionof reaggregates.Col I, type I collagen;FG, fibrinogen;EN. entactin; TN,tenascin-C;TSP,thrombospondin;PLYS,polylysine.Dataarethe means.Bar, SE.

1942



UPNET

80 D ACT


higher level of migration of cells from high-grade tumors, and thedifference was statistically significant (P = 0.007). These resultssuggest that high tumor grade correlates with higher level of migrationof cells from astrocytomas in vitro.

Glioblastoma Cell Lines. Because work on the involvement ofadhesion molecules in brain tumor migration has relied largely onstudies of cell lines, it was of interest to compare the migration of cellsfrom primary brain tumors with that of established brain tumor celllines. For this purpose, aggregatesof several well-studied brain tumor

A

B

C

100â€¢

8O@

00â€¢

4O@

2O@

160â€¢

@â€˜14o.Co

k::IC 80.

@60

o,40.

Fig. 4. Effects of extracellular matrix proteins on migration of cells from brain tumors.Migration assayswereperformedwith reaggregatedcells from juvenile pilocytic astrocytomas(IPA), low gradeastrocytomas(WA), anaplasticastrocytomas(AA), glioblastomas (GBM), and PNETs. The average migration was determined for each group oftumorson varioussubstrates.Numbersin parentheses,numbersof pooled tumorspergroup. Data are the means. Bar, SE.

different grades(Fig. 4). Juvenile pilocytic astrocytomas, for example,exhibited an overall lower level of migration than tumors of highergrade. The juvenile pilocytic astrocytoma case that exhibited thehighest level of migration in vitro was unusually aggressive for itsgrade, requiring surgery two additional times after the initial resection; this specimen was taken from the third resection that wasperformed 3 months after the second resection.

To evaluate differences in the migration of cells from differentgrades and due to the small number of astrocytomas of grades otherthan glioblastoma, astrocytomas were grouped as low grade tumors(pooled juvenile pilocytic astrocytomas and low grade astrocytomas;n = 8) or high-grade tumors (pooled anaplastic astrocytomas andglioblastomas; n 16). When these two groups were compared fortheir average migration, cells from high-grade tumors tended to migrate at higher levels (42.3 Â±4.7 @tm)than cells from low leveltumors (28.9 + 3.9 @.tm;P = 0.066; Fig. 5B). When individualsubstrates were compared for their ability to stimulate migration, itwas found that type IV collagen exhibited a tendency to stimulatehigher levels of migration of cells from high-grade tumors (21.4 Â±6.3tam) than cells from low grade tumors (5.0 Â±1.7 p.m; P = 0.078).Combined data on collagens (pooled results for type I collagen andtype IV collagen substrates) indicate that collagens stimulated a

Fig. 5. Comparison of migration of brain tumor cells. A, PNETs versus astrocytomas.Average migration of PNETs (n = 6) and pooled astrocytomas (ACT: pooled juvenilepilocytic astrocytomas, low grade astrocytomas, and anaplastic astrocytomas, and glibblastomas; n = 24) were compared. Data on PNETs are presented and are reproduced forcomparison;entactin,thrombospondin,andBSA werenot includedbecausethey stimulated minimal levels of cell migration in both groups. @,the difference in migrationbetweengroupswasstatisticallysignificantfor the indicatedsubstrates(FG, P = 0.018;Ng-CAM, P = 0.023; PLYS, P = 0.012; t test). Average migration on the varioussubstrateswas13.5Â±4.1pm forPNETs and31.9Â±2.9 @mforastrocytomas(P< 0.001;t test).Dataarethemeans.Bar,SE.B,lowgradetumorsversushigh-gradetumors.Theaverage migration of low grade (LG) tumors (pooled juvenile astrocytic astrocytomas andlow grade astrocytomas; n = 8) was compared with that of high-grade (HG) tumors(pooledanaplasticastrocytomasandglioblastomas;n = 16).*, thedifferencein migrationbetween groups was statistically significant for type I and type IV collagen substrates(P = 0.007). Average migration on the various substrates was 42.3 Â±4.7 p.m forhigh-gradetumorsand28.9 Â±3.9 @mfor low gradetumors(P = 0.066).C, cells fromglioblastomas versus glioblastoma cell lines. The migration of cells from glioblastomas(GBM; n = 13) was compared with that of eight glioblastoma cell lines (U-87 MG,U-l 18MG, U-138 MG, U-251 MG, U-373 MG, A-l72, T98G, and Hs683). @,thedifference between groups was statistically significant for type IV collagen (P = 0.0023)aswell aspooleddataon collagens(P = 0.001).Averagemigrationwas69.5 Â±7.6 @mfor glioblastomas and 46.4 Â±5.7 @.smfor glioblastoma cell lines (P = 0.025).

1943




cell lines were tested on the same substrates used for reaggregates ofprimary tumor cells.

In contrast to cells from glioblastomas which exhibited spreadshape in 5 of 13 cases,glioblastoma cell lines consistently exhibiteda spread shape (10/10 lines analyzed; P = 0.0027; Table 1 and Fig.2F). Glioblastoma cell lines resembled cells from glioblastoma tumors in that they migrated on a wide variety of substratesand in theiraverage migration patterns. However, marked differences were observed. When compared for their average migration on differentsubstrates, glioblastoma cell lines migrated more extensively(69.5 Â±7.6 p.m) than cells from glioblastomas (46.4 Â±5.7 @.tm;P = 0.025; Fig. 5C). Comparison of migration on individual substrates indicated that glioblastoma cell lines migrated on type IVcollagen at a strikingly higher level (126 Â±24 tim) than cells fromglioblastomas (35 Â±18 @.tm;P = 0.003). Combined data on collagens(pooled results for type I collagen and type IV collagen substrates)also indicated a higher level of migration of glioblastoma lines thatwas statistically significant (P 0.001).

PNET Cell Lines. Cells from aggregates of D-283 and D-341PNET cells migrated very poorly (data not shown). In contrast, DaoyPNET cells migratedwell and spreadon fibronectin, laminin, type Iand type IV collagens, tenascin-C and fibrinogen, which resembledPNET-73 (see â€œAstrocytesversus PNETsâ€•).However, Daoy cells didnot migrate on Ng-CAM, polylysine, entactin, or BSA (data notshown). Overall, our results with PNET cell lines are consistent withthose of previous adhesion studies (32) showing that PNET lines withneuronal differentiation (including D-283 and D-341) do not adheretofibronectin, laminin, or type IV collagen substrates, in contrast toDaoy cells which have glial differentiation and adhere to these substrates(32).

Antibodies to a, and@ Integrins Inhibit Brain Tumor CellMigration. The present studies showed that various ECM proteinsare able to stimulate migration of cells from astrocytomas. To gaininformation about receptors on astrocytoma cell surfaces that mayplay crucial roles in brain tumor cell migration, we performed experiments aimed at perturbing migration. Preliminary experiments on celllines were done with mAbs specific for several integrins, including a@and f3@,because they are components of a large number of integrinheterodimers (6), and they have been shown to be expressed byastrocytoma cells in vivo (33) as well as in astrocytoma cell lines (24,34).Thepreliminaryexperimentsshowedstrikinginhibitionbyanti-a@ and anti-f31 mAbs but other mAbs tested, including anti-a2,anti-a3, and anti-@4,were not particularly effective (data not shown).Given their effectiveness, we decided to focus on mAbs to a@and f3@used separately or in combination on substrates that stimulated significant migration of tumor cells, i.e., fibronectin, laminin, type I andtype IV collagens, and fibrinogen.

Because the migration patterns are less variable with astrocytomacell lines than with primary tumor cells, systematic analyses were onlymade using the cell lines. Several cell lines were chosen to identifyproperties that may apply broadly. U-87 MG, U-i 18 MG, U-25 1 MG,and U-373 MG cell lines were selected based on their ability tomigrate on the five different ECM molecules mentioned above, andthey were analyzed for effects of anti-a@ and anti-@1 mAbs onmigration (Figs. 6â€”8).

Typical effects of mAb treatment are illustrated by the behavior ofU-373 MG cells (Figs. 6 and 7). When applied to U-373 MG cellaggregates,anti-@31mAb strongly inhibited migration on type I and IVcollagens and laminin and fibrinogen, and partially inhibited migration on fibronectin. Anti-a2 and anti-@3 mAbs had no effect onmigration on fibronectin, confirming the specificity of the effects

(data not shown). Anti-a@ mAbs inhibited migration strongly onfibrinogen, partially on fibronectin and types I and IV collagen, and

Fig. 6. Inhibition of tumor cell migration by anti-integrin antibodies. U-373 MG cellswere tested for migration on fibrinogen substrates in the presence of buffer (control),anti-a@,anti-@1,or a combination of anti-a@and anti-(31 mAbs, as described in â€œMaterialsand Methods.â€•The phase micrographs were taken 16 h after adding the antibodies. Bar,100 gsm.

not at all on laminin. When applied in combination, anti-a@ andanti-f31 mAbs tended to have a stronger effect than either mAb whenapplied in isolation. Effects of anti-a@and/or anti-f31 mAbs on themigration of U-i 18 MG and U-25 1 MG on type IV collagen andlaminin (data not shown) were essentially the same as those observedfor U-373 MG.

Some differences on the effects of anti-a,@,and anti-@1mAbs wereobserved among cell lines migrating on fibronectin, type I collagen,and fibrinogen. Migration of both U-87 MG and U-118 MG cells onfibronectin was strongly inhibited by the combination of anti-as, andanti-@1mAbs (data not shown). In addition, U-ll8 MG cells differedfrom U-373 MG cells in that U-118 MG cell migration was onlypartially inhibited by anti-j31 mAb on fibrinogen substrates (Fig. 8)and strongly inhibited by anti-a@mAb on type I collagen (data notshown). Also, migration of U-251 MG cells on type I collagen wasstrongly inhibited by anti-a@ mAb, and, finally, U-25l MG cellmigration on fibrinogen was inhibited only partially by anti-(31mAb(Fig. 8). The results of the combined perturbation experiments mdicate that a@and (3@integnn components play a crucial role on themigration of glioblastoma cells on several different ECM molecules.

DISCUSSION

We have studied the migration of primary brain tumor cells onindividual ECM molecules in vitro. The combined results indicate thatcells from tumors of different types and grades differ in their migratory behavior, although variability was observed among tumors of thesame group. The major generalizations are: (a) cells from astrocytomas can migrate on a variety of substrates that are found in normalbrain and may be induced in and around brain tumors; (b) in contrast,cells from PNETs migratedonly on laminin, fibronectin,or type IV

1944



IUcontrolUanIH@1Dsntha@Dantkz,+antl@

FNFig. 7. Inhibition of tumor cell migration on various substrates by anti-@1 and anti-a@

integrin antibodies. U-373 MG cells were tested for migration on fibronectin, type I andtype IV collagens, laminin, and fibrinogen substrates in the presence of buffer (control),anti-a@,anti-@1,or a combination of anti-a,, and anti-f31mAbs, as described in â€œMaterialsand Methods.â€•For each substrate, values were normalized so that migration in the absenceof antibodieswas 100%.This figure is a representativeexampleof six independentexperimentsthat exhibitedsimilar results.Dataarethe means.Bar, SE.

collagen; (c) cells from high-grade tumors migrate more extensivelythan cells from low grade tumors on collagens; (d) glioblastoma celllines consistenfly migrated very well on collagens, in contrast to cellsfrom glioblastomas, which in about half of the cases migrated verypoorly on collagens; and (e) a@and@ integrmn subunits play a keyrole in brain tumor cell migration on several ECM proteins.

Migration Assay. The focus of the present work was to study themigratory response of cells from human brain tumors to stimuliprovided by individual adhesion molecules. It was judged to beimportant to test cells from excised tumors in a short-term migrationassay because protein expression and cell phenotype graduallychanges in vitro (35). To optimize conditions that facilitated scoringtumor cell migration, tumor cell reaggregateswere used, and thecytologicaland functionalobservationsindicatethat the reaggregatescontain mainly brain tumor cells that exhibit a migratory behaviorsimilar to that of cells from tumor fragments. Emigration of confluentcell lines from small circular regions has previously been used toinvestigatebrain tumor cell migration in vitro (25, 36).

Brain tumors of the same grade exhibit wide variability in theirbehavior in vivo, which contributes to uncertainties regarding patienttreatment and prognosis.Thus, any new information to improveclassification criteria may be of particular value. Consistent withtumorbehaviorinvivo,substantialvariabilitywasobservedinthemigrationpatternsof tumorsof the samegrade in vitro. In a prelimmary attempt to determine whether the migration patternsreflectedthe situation in vivo, certain clinical features were compared with themigration patterns. For both PNETs and juvenile pilocytic astrocytoma the tumor with the highest levels of migration was a recurrentcase, pointing at the possibility that recurrent brain tumors maygenerally exhibit higher levels of migration and malignancy. Additional studies are necessary to test this possibility.

PNETs. PNET cells tend to disseminate through the leptomeninges. Our results showing in vitro migration of cells from PNETs onlaminin, fibronectin, or type IV collagen (five of six cases) and therich expression of these ECM molecules in leptomeninges provide anexperimental basis to suggest that, in vivo, these ECM molecules arefavorable substratesfor disseminationof PNETs along leptomeningeal surfaces.

The lack of migration of cells from PNETs on Ng-CAM, which isa substratefor glial andglioma cell adhesion(37, 38) and is involved

colt COIN LU

MIGRATIONOF BRAINTUMORCELLS

in the normal migration of cerebellar granule cells (39), may berelated to low expression of Ng-CAM receptors on cells of PNETs.Indeed medulloblastomas (PNETs) show expression of both N-CAMand N-cadherin, which are expressed at very early stages of normaldevelopment, but lack expression of L1INg-CAM (40), which isnormally expressed in postmitotic neurons (41). Thus, the absenceofreceptors such as LuNg-CAM, which are on differentiating neurons,could explain the lack of migration of cells from PNETs on Ng-CAM.

Astrocytomas. That cells from astrocytomas migrate on fibronectin,lamimn,andcollagenIV,whicharerichlyexpressedinbasementmembranes of blood vessels, may explain why blood vessels are acommon route of migration for astrocytoma cells in situ (42, 43) andin experimental animal models (44). Furthermore, that transplantedexperimental astrocytomas elicit the production of fibronectin by hostcells (45) suggests that, in situ, reactive glia may deposit fibronectinwhich could provide a substrate for astrocytoma cell migration. Indeed, the consensusof studies on human tumors is that fibronectin insitu is restricted mainly to blood vessels and leptomeninges (21, 46).One report, however, indicated that fibronectin is expressed in malignant gliomas and adjacent regions with infiltrating tumor cells (47).

Given that Ng-CAMIL1 is expressed in normal brain (41), thepresent studies showing migration of astrocytoma cells on Ng-CAMindicate that Ng-CAM is a potential substrate for migration of astrocytomas in brain regions other than mature white matter, which lacksNg-CAM/Ll. Our results indicating low levels of migration of primary astrocytoma cells and cell lines on tenascin-C differ from thoseof recent experiments with cell lines which migrated at high levels ontenascin-C (25). This difference may be due to the presenceof serumin the medium used by the other laboratory, because tenascin-C maymodulate the promotion of migration by fibronectin or vitronectin inserum.

The observation that several adhesion proteins can be used by cellsfrom astrocytomas as substratesfor migration in culture suggeststhatthesecells are capable of using a variety of substratesfor migration invivo. Such versatility is likely to be due to the expression of variouscell surface receptors, and this view is supported by previous studiesshowing that integrmncomponents a3, a7, a@,@ and@ are upregulated in human brain tumors (33).

Although collagen is not found at significant levels in mature brain,migrationon collagensallowed us to discriminatebetweencells from

I@ control@@ @:[email protected]@+â€˜@â€œ@i

U87-MG U118-UG U251@MGU373-MGFig. 8. Inhibition of migration of various cell lines on fibrinogen substrates. U-87 MG.

U-118 MG, U-25l MG, and U-373 MG cells were assayed for their migration onfibrinogen substrates in the presence of the indicated anti-integrmn mAbs, as described inâ€œMaterialsand Methods.â€•Valueswere normalizedso that migration in the absenceofantibodies was 100%. For each cell line, this figure shows representative examples of twoto six independentexperimentsperformedon thedifferentcell lines.Dataarethemeans.Bar, SE.

FO

-. I 20-

2@.@ 80@

g

1945




low grade tumors (low levels of migration) versus cells fromhigh-grade tumors (intermediate levels) versus glioblastoma celllines (high levels). In the latter case, the distinction is likely to be

due, at least in part, to a gradual selection and adaptation ofdissociated glioblastoma cells to culture conditions. We suggestthat the contrasting migration in vitro of low grade tumors versushigh-grade tumors, as well as that of glioblastomas versus glioblastoma cell lines on collagens is a reflection of an increase in thenumber of active integrins. This view is consistent with studiesshowing increased integrin expression in glioblastomas versus lowgrade astrocytomas (33) and in glioblastoma cell lines versusglioblastomas (24, 33, 48, 49). That collagen substrates eliciteddistinct behavior of low grade versus high-grade tumors as well asglioblastomas versus glioblastoma cell lines may be related to theidentity of the particular receptors involved in stimulating astrocytoma cell migration on each substrate. It will be important todetermine whether these receptors are found in tumors and whetherappropriate ECM ligands are also present.

Adhesion Molecule Receptors. Integrin function has been investigated in adhesion of glioblastoma cell lines, but it is not clear thatthere is a direct relationship between cell adhesion and migration. Forexample, when Ng-CAM, N-cadherin, and lamimn substrates werecompared for their effect on axonal growth cone behavior, growthcone adhesiveness to substrate was a poor predictor of rate of growthcone migration (50). Adhesion of U-138 MG and U-373 MG cells totype I and type VI collagens, laminin, and the ECM molecule undulinwas strongly inhibited by mAb to f3@and was not affected by mAbsto a3 and ca@(33). In contrast, migration of these cells across filterscoated with Matrigel, a mixture of ECM components, was barelyinhibited by mAb to@ , although mAbs to other integrin componentshad stronger inhibitory (e.g., a@)or potentiating (e.g., a5) effects,depending on cell line and integrin component (34). These resultssupport the view that adhesion and migration may not be directlyrelated.

The present perturbation studies showed the involvement of a@and13@integrinsontumorcellmigrationonfibronectin,collagen,lamiin,and fibrinogen substrates. Our results are consistent with publisheddata on the binding specificity of these integrins. a@ligands includefibronectin, fibrinogen, and collagen (6, 51, 52), and f3@ligandsinclude fibronectin, collagens, laminin, and fibrinogen (6, 53). However, because several different integrins can bind to a single ECMmolecule, the complete inhibition of migration on collagens, lamiin,and fibrinogen achieved with these mAbs could not necessarily bepredicted in most of the cases. For example, the effects of anti-@1antibodies on the migration on type IV collagen was expected becausethe major receptor for this molecule is a1@1 (Col IV). However, theeffects of anti-/31 antibodies on the migration on lamrnm could not besimply predicted becausein addition to four @1integrmns(a1@1,a2@31,a3f31, and a7@1), laminin binds to aj34. Similarly, one could not

simply predict that a combination of anti-a@and anti-@1antibodieswould inhibit migration on fibrinogen because a11b133also binds tofibrinogen. In addition, our data are consistent with a role for at leasttwo separateintegrin heterodimers and not simply a@f31. For example,U-25l MG cell migration on collagens and laminin was stronglyinhibited by anti-@1mAb and only weakly or no at all by anti-a@,while their migration on fibrinogen was strongly inhibited by anti-a@and only weakly by anti-@1mAb.

The present studies focus on the effects of adhesion molecules onbrain tumor cell migration in vitro and do not addressother importantfactors contributing to tumor cell infiltration in vivo, including the roleof proteases released by tumor cells, cell division, and neovascularization. Thus, clearly, additional studies in vitro and in animal modelswill be necessary to investigate whether the conclusions of this work

are valid for brain tumor cell migration in vivo. Of particular interestwill be to determine whether fibronectin, laminin, and type IV collagen are in fact key players in the leptomeningeal dissemination ofPNETs, and to test the extent to which astrocytoma cell infiltrationdepends on the function of a@and (3@integrins. From a therapeuticpoint of view, it is relevant that antibodies specific for the mtegrina@f33disrupt angiogenesisby binding to endotheial cells, whichcausesregression of experimental tumors (54). Assuming that anti-a,,antibodies disrupt angiogenesis as well as anti-aj33 antibodies, ourstudies suggest that direct effects of a combined anti-a@ and anti-@1treatment on tumor cell migration should add to the effects of disrupting angiogenesis with anti-a@.

ACKNOWLEDGMENTS

We thank Boris Veksler for his expert technical assistance, Drs. Michael

Gruber,FilippoGiancotti,GeorgeJab, andJamesSalzerforvaluablediscussions, Dr. Keith Langley (Amgen) for providing us purified tenascin-C, and

CollaborativeBiochemicalProductsfor anti-integrinantibodies.

REFERENCES

1. Edelman,0. M., and Crossin,K. L. Cell adhesionmolecules:implications for amolecular histology. Annu. Rev. Biochem., 60: 155â€”190,1991.

2. Takeichi, M. Cadherincell adhesionreceptorsas a morphogeneticregulator.Science(Washington DC), 251: 1451â€”1455,1991.

3. Crossin, K. L. Cell adhesion molecules in embryogenesis and disease. Ann. N.Y.Acad. Sci., 615: 172â€”186,1991.

4. Jessell,T. M. Adhesionmoleculesandthehierarchyof neuraldevelopment.Neuron,I: 3-13, 1988.

5. Grumet, M. Cell adhesion molecules and their subgroups in the nervous system. Curr.Opin. Neurobiol., 1: 370â€”376,1991.

6. Hynes, R. Integrins. Versatility, modulation, and signaling in cell adhesion. Cell, 69:24711â€”24725,1992.

7. Friedlander, D. R., Mege, R-M., Cunningham, B. A., and Edelman, G. M. Cellsorting-out is modulated by both the specificity and amount of different cell adhesionmolecules(CAMs) expressedon cell surfaces.Proc. Nail. Acad. Sci. USA, 86:7043â€”7047,1989.

8. Nose, A., Nagafuchi, A., and Takeichi, M. Expressed recombinant cadherins mediatecell sorting in model systems. Cell, 54: 993-1001, 1988.

9. Takeichi, M. Cadherins in cancer@implications for invasion and metastasis. Curr.Opin. Cell Biol., 5: 806â€”811,1993.

10. Giancotti. F. G., and Ruoslahti, E. Elevated levels of the a@ fibronectin receptorsuppress the transfonned phenotype of Chinese hamster ovary cells. Cell, 60: 849â€”859, 1990.

11. Gehlsen, K. R., Davis, G. E., and Sriramarao, P. lntegrmn expression in humanmelanoma cells with differing invasive and metastatic properties. Clin. & Exp.Metastasis, 10: 111â€”120,1992.

12. Clark, E. A., and Brugge,J. S. Integrmnsand signal transductionpathways:the roadtaken. Science (Washington DC), 268: 233â€”239,1995.

13. Grumet, M., Hoffman, S., Crossin, K. L., and Edelman, G. M. Cytotactin. anextracellular matrix protein of neural and non-neural tissues that mediates glia-neuroninteraction. Proc. NatI. Acad. Sci. USA, 82: 8075â€”8079,1985.

14. Reichardt, L. F., and Tomaselli, K. J. Extracellular matrix molecules and theirreceptors:functions in neural development.Annu. Rev. Neurosci.,14: 531â€”570,1991.

15. Erickson, H. P., and Bourdon, M. A. Tenascin: an extracellular matrix proteinprominentin specializedembryonictissuesand tumors.Annu. Rev. Cell Biol, 5:71â€”89,1989.

16. Moonen, G.. Grau-Wagemans, M. P., and Selak, I. Plasminogen activator-plasminsystem and neuronal migration. Nature (Lond.), 298: 753â€”755,1982.

17. Venstrom,K. A., andReichardt,L F. Extracellularmatrix2: roleof extracellularmatrixmoleculesandtheirreceptorsin thenervoussystem.FASEBJ., 7: 996-1003,1993.

18. Schuster,J. M., andBigner,D. D. Immunotherapyandmonoclonalantibodytherapies. Curr. Opin. Oncol., 4: 547â€”552,1992.

19. Higuchi, M., Ohnishi, 1., Arita, N., Hiraga, S., and Hayakawa,T. Expressionoftenascinin humangliomas:its relationshipto histologicalmalignancy,tumordifferentiation and angiogenesis.Acta Neuropathol.,85: 481â€”487,1993.

20. Bourdon, M. A., Wikstrand, C. J., Funhmayer, H., Matthews, 1. J., and Bigner, D. D.Human glioma-mesenchymal extracellular matrix antigen defined by monoclonalantibody.CancerRes.,43: 2796â€”2805,1983.

21. McComb. R. D., and Bigner, D. D. Immunolocalizationof monoclonalantibodydefinedextracellularmatrix antigensin humanbrain tumors.J. Neuro-Oncol.,3:181â€”186,1985.

22. Zagzag, D., Fnedlander, D. R., Miller, D. C., Dosik, J., Cangiarella, J., Kostianovsky.M., Cohen, H., Grumet, M., and Greco, A. Tenascin expression in astrocytomascorrelateswith angiogenesis.CancerRes.,55: 907â€”914,1995.

23. Zagzag, D., Friedlander, D. R., Dosik, J., Chickramane, S., Chan, W., Greco, M. A.,Allen, J. C., Dorovini-Zis, K., and Grumet, M. Tenascin-C expression by angiogenic

1946




vesselsin humanastrocytomasandby humanbrainendothelialcells in vitro. CancerRes.,56: 182â€”189,1996.

24. Giese, A.. Rief, M. D., Loo, M. L., and Berens, M. E. Determinants of humanastrocytoma migration. Cancer Res., 54: 3897â€”3904,1994.

25. Giese, A., Loo, M. A., Rief, M. D., Tran, N., and Berens, M. E. Substrates forastrocytomainvasion.Neurosurgery,37: 294â€”302,1995.

26. Grumet, M., and Edelman, G. M. Heterotypic binding between neuronal membranevesicles and glial cells is mediated by a specific cell adhesion molecule. J. Cell Biol.,98: 1746â€”1756,1984.

27. Kleihues, P., Burger, P. C., and Scheithauer, B. W. The new WHO classification ofbrain tumors. Brain Pathol., 3: 255â€”268,1993.

28. Chicoine, M. R., and Silbergeld, D. L. Assessmentof brain tumor cell motility in vivoand in vitro. J. Neurosurg., 80: 615â€”622,1995.

29. Long, D. M. Capillaryultrastructureandtheblood-brainbarrierin humanmalignantbrain tumors. Neurosurgery (Baltimore), 32: 127â€”144,1970.

30. Zagzag,D., Goldenberg,M., and Brem, S. Angiogenesisand blood-brainbarrierbreakdownmodulateCT contrastenhancement:an experimentalstudy in a rabbitbrain-tumor model. Am. J. Neuroradiol., 10: 529â€”534,1989.

31. Paulsson,M., Aumalley, M., Deutzmann, R., Timpl, R., Beck, K., and Engel, J.Laminin-nidogencomplex:extractionwith chelatingagentsandstructuralcharacterization. Eur. J. Biochem., 166: 11â€”19,1987.

32. Wikstrand,C. J., Friedman,H. S.,andBigner,D. D. Medulloblastomacell-substrateinteractions in vitro. Invasion & Metastasis, 11: 310â€”324,1991.

33. Paulus. W., Baur, I., Schuppan, D., and Roggendorf, W. Characterization of integrmnreceptors in normal and neoplastic human brain. Am. J. Pathol., 143: 154, 1993.

34. Paulus, W., and loan, J. C. Basement membrane invasion of glioma cells mediatedby integrmnreceptors. J. Neurosurg., 80: 515â€”519,1994.

35. Paulus, W., Huettner, C., and Tonn, J. C. Collagens, integrmnsand the mesenchymaldrift in glioblastomas:a comparisonof biopsyspecimens,spheroidandearly monolayer cultures. hit. J. Cancer, 58: 841â€”846,1994.

36. Chicoine, M. R., and Silbergeld,D. L. The in vitro motility of humangliomasincreaseswith increasing grade of malignancy. Cancer (Phila.), 75: 2904â€”2909,1995.

37. Sakurai,T., Friedlander,D. R., andGrumet, M. Expressionof polypeptidevariantsofreceptorprotein tyrosinephosphatasef3: the secretedform, phosphacan,increasesdramatically during embryonic development and modulates glial cell behavior invitro. J. Neurosci. Res., 43: 694â€”706,1996.

38. Milev, P., Friedlander, D., Sakurai, T., Karthikeyan, L, Had, M., Margolis, R. K.,Onimet, M., and Margolis, R. U. interactions of the chondroitin sulfate proteoglycan

@@can,the extracellular domain of a receptor-type tyrosine phosphatase, withneurons, glia, and neural cell adhesion molecules. J. Cell Biol., 127: 1703â€”1715,1994.

39. Hoffman, S., Friedlander, D. R., Chuong, C-M., Grumet, M., and Edelman, 0. M.Differential contribution of Ng-CAM and N-CAM to cell adhesion in different neuralregions. J. Cell Biol., 103: 145â€”158,1986.

40. Figarella-Branger, D. F., Durbec, P. L., and Rougon, G. Differential spectrum of

expression of neural cell adhesion molecule isoforms and Ll adhesion molecules onhuman neuroectodermal tumors. Cancer Res., 50: 1102â€”1106, 1990.

41. Grumet, M. Structure, expression, and function of Ng-CAM, a member of theimmunoglobulin superfamily involved in neuron-neuron and neuron-glia adhesion.J. Neurosci. Res., 31: 1â€”13,1992.

42. Scherer, H. J. The forms of growth in gliomas and their practical significance. Brain.63: 1â€”35,1940.

43. Burger, P. C., Heinz, E. R., Shibata, T., and Kleihues, P. Topographic anatomy andCT correlations in the untreated glioblastoma multiforme. J. Neurosurg., 68: 698â€”704, 1988.

44. Laws, E. R., Goldberg, W. J., and Bernstein, J. J. Migration of human malignantastrocytoma cells in the mammalian brain: Scherer revisited. mt. J. 0ev. Neurosci.,II: 691â€”697,1993.

45. Pedersen, P-H., Marienhagen, K.. Mork, S., and Bjerkvig, R. Migratory pattern offetal rat brain cells and human glioma cells in the adult rat brain. Cancer Res., 53:5158â€”5165,1993.

46. McComb, R. D., and Bigner, D. D. Immunolocalization of laminin in neoplasms ofthe central and peripheral nervous systems. J. Neuropathol. Exp. Neurol., 44: 242â€”253, 1985.

47. Higuchi, M., Ohnishi, 1., Arita, N., Hiraga, S., Iwasaki, H., Mori, S., and Hayakawa,1. Immunohistochemical localization of fibronectin, laminin and fibronectin receptorin human malignant gliomas in relation to tumor invasion. Brain Nerve (Tokyo), 43:17â€”23,1991.

48. Gladson, C. L., and Cheresh, D. A. Glioblastoma expression of vitronectin and theavf33integrin:adhesionmechanismfor transformedglialcells.J. Clin.Invest.,88:1924â€”1932,1991.

49. Paulus, W., and Tonn, J. C. Interactions of glioma cells and extracellular matrix.J. Neuro-Oncol., 24: 87â€”91, 1995.

50. Lemmon. V., Burden, S. M., Payne, H. R., Elmslie, 0. J., and Hlavin, M. L. Neuritegrowth on different substrates:permissive versus instructive influence and the role ofadhesive strength. J. Neurosci., 12: 818â€”826,1992.

51. Stupack, D. G., Shen, C., and Wilkins, J. A. Induction of a@f3@integrin-mediatedattachmentto extracellularmatrix in f3@integrmn(CD29)-negativeB cell lines.Exp.Cell Res., 203: 443â€”448,1992.

52. Davis, G. E. Affinity of integrmns for damaged extracellular matrix: a@133binds todenatured collagen type I through RGD sites. Biochem. Biophys. Res. Commun., 3:1025â€”1031,1992.

53. Neugebauer, K. M., Venstrom, K. A., and Reichardt, L. F. Adhesion of a chickenmyeloblast cell line to fibrinogen and vitronectin through a @3l-classintegrin. J. CellBiol., 116: 809â€”815, 1992.

54. Brooks, P. C., Montgomery, A. M. P., Rosenfeld, M., Reisfeld, R. A., Hu, T., Klier,G., and Cheresh, D. A. Integrin aj33 antagonists promote tumor regression byinducing apoptosis of angiogenic blood vessels. Cell, 79: 1157â€”1164, 1994.

1947



migration of brain tumor cells on extracellular matrix ...areexpressedin the basallamina surrounding...

Documents