[CANCERRESEARCH54, 3897-3904, July 15, 19941

ABSTRACF

A unique characteristicof astrocyticmalignanciesis their frequentdissemination through the brain. Cellular determinants of migration inelude adhesion to the substratum, restructuring of the actin cytoskeletonto generate motion, and (in the setting of Invasion Into tissue) secretion ofenzymes for remodeling interstitial space to accommodate forward motionof the migrating cell. In order to better understand these features in thecontext of local brain invasionby astrocytomacells, the adhesionandmigratory propet@tlesof these cells have been investigated in an in vitromonolayer system. Adhesion of 8 dIfferent astrocytoma cell lines to different purified human extracellular matrix (ECM) proteins (collagen typeIv, cellularfibronectin,laminin,andvitronectin)revealedthatthereis no“astrocytoma-specific― ECM protein that consistently leads to high cellbinding. Similarly, migration of astrocytoma cells was found to be vanable and dependent on different ECM proteins. Laminin was frequentlythe most permissivefor adhesionand migration.Adhesionto collagen,fibronectin, and vitronectin was integnin dependent and could be blockedusing anli-@ integnin antibodies; m contrast, attachment to laminin couldnot be blocked using these antibodies. A comparison of adhesion withmigration for each ofthe cell lines on each ofthe 4 ECM proteins revealedthat poor adhesion was associated with minimal migration and thatfrequently, high adhesion was correlated with rapid migration. Whentested for migration on autologous, cell-derived ECM, none of the celllines were as migratory as they were on one ofthe punfied ECM proteins,with the exception of SF767 cells. Furthermore, it was found that ECMfrom SF767 cells promoted the migration of other astrocytoma cells. The

results from this study indicate that migration is a constitutive behavior ofglioma cells which is dependent on, or modified by, the presence orabsence of permissive ligands in the environmenL

INTRODUCHON

The local invasive behavior of astrocytic neoplasms into surrounding normal brain accounts for the difficult clinical management andpoor outcome of patients with these tumors (1). The ability to suc

cessfully permeate the brain is a unique feature of astrocytomas,which stands in contrast to the localized, well delineated lesionsformed by other tumors metastatic to the brain (2). The infrequencywith which astrocytomas successfully establish extracranial metastases (3) further highlights the unique interaction of astrocytomas withtheir local environment. The current study was undertaken to characterize astrocytoma cells for determinants of their migratory behaviorin a monolayer culture system.

The biological attributes of cell migration and invasion includeadhesion to the immediate tissue structure (either extracellular matrixor cell membranes), motility mechanisms to generate locomotion, andthe ability to influence the local environment so as to create space intowhich to move (4). The latter process is generally believed to includeselective remodeling of the extracellular matrix by the secretion ofproteases (5, 6).

The brain is known to include the basement membrane glycopro

Received 12/22/93; accepted 5/12/94.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.

1Supported by NIH Grant NS27030 (M. E. B., M. D. R.), Phi Beta Psi Sorority, andthe Deutsche Forschungsgemeinschaft Grant Gi 218/1-2 (A. G.).

2To whom requests for reprints should be addressed, at Neuro-Oncology Laboratory,Barrow Neurological Institute of St. Joseph's Hospital and Medical Center, 350 WestThomas Road, Phoenix, AZ 85013-4496.

teins fibronectin, laminin, and vitronectin, as well as collagen types I,III, and IV (7, 8); however, these structures are largely confined tomajor blood vessels and to the glial limitans externa. Conceptually,these are border structures isolating the brain from the rest ofthe bodyand do not account for the patterns of invasion seen in spontaneoushuman astrocytomas. In this context, the mechanisms used by astrocytoma cells to move through the brain parenchyma remain poorlyunderstood. It is uncertain whether invasive astrocytoma cells utilizeconstitutive brain ECM3 for migration or cell-cell interactions orwhether they elaborate their own matrix for subsequent invasion. Asan initial evaluation of the role of ECM proteins on the migratorybehavior of human astrocytoma cells, purified human collagen typeIv, laminin, cellular fibronectin, and vitronectin were used as substrates in this study.

Cell adhesion to ECM proteins is mediated by different types ofreceptors, among which is a class of transmembrane heterodimerproteins called integrins (9). The assortment of a and f3 subunits (ofwhich there are 16 and 8 presently described, respectively) (10) leadsto differential affinity to different ECM ligands, although there maybe some promiscuity for different ligands demonstrated by individualintegrins. Further complicating this binding activity is the diversity ofintegrins which are able to bind to the same ECM ligand (reviewed inRef. 11). The integrins serve to attach cells to the ECM and, in someinstances, also function in conjunction with the cytoskeleton to provide the force for cell movement (12—14).The signal transductionevents between adhesion to the ECM and cell migration are beingelucidated for selected integrmns and are revealing processes consistentwith signal transduction pathways driven by other transmembranereceptors (15—18).

This study was undertaken to identify features of the adhesive andmotility properties of human astrocytoma cells on defined ECMproteins or astrocytoma-derived ECM. The interrelationship betweenadhesion and migration was also investigated.

MATERIALS AND METhODS

Cells, Culture, and Extracellular Matrix. Humanastrocytomacell lines[NCE G-22, NCE G-28, NCE G-63, NCE G-112, and NCE G-118 (19, 20);U-251MG (21); and SF767 (22)] and one culture of normal human astrocytes(QG-4, kindly supplied by Joan R. Shapiro) were propagated in monolayerculture in minimal essential medium with 5—10% FCS (Hyclone Laboratories,

Logan, U'F)for tumor lines or Waymouth's MAB 87/3 (GIBCO) supplementedwith 20% FCS for the normal astrocytes. Cells were passaged using trypsinization at regular intervals depending on growth characteristics. Human extracellular matrix proteins and their sources were collagen type IV and vitronectin(GIBCO-BRL, Gaithersburg, MD) and laminin and fibronectin (Sigma Chemical Co., St. Louis, MO). Culture dishes or microtiter wells were coated withthe ECM proteins by preparing stock concentrations (100 @Wml)of theproteins and allowing these to coat the surface for 1 h at 37°C,rinsing the wellswith PBS, blocking the surface with 1% bovine serum albumin (Sigma) for 30

mm at 25°C,and then storing the vessels at 4°Cwith PBS. Production ofastrocytoma-derived ECM was done using cultures of the cells which wereallowed to persist at confluence for 10 days. After a thorough rinsing, the cellswere lysed from the ECM using treatment with 0.5% Triton X-100, followedby 0.1 M NH4OH and 3 rinses with PBS (23, 24). Protein-coated dishes or

3The abbreviations used are: ECM, extracellular matrix; FCS, fetal calf serum; PBS,phosphate-buffered saline; RT-PCR, reverse transcription-polymerase chain reaction.

3897

Determinants of Human Astrocytoma Migration'

A.IfGiese, Monique D. Rief, Melinda A. Loo, and Michael E. Berens2Neuro-Oncology Laboratory, Barrow Neurological Institute ofSt. Joseph ‘sHospital and Medical Center, Phoenix Arizona 85013-44%

Table 1 Primer sequences forRT-PCR analysis ofintegrinsGeneOligonucleotideAnnealing

temperatureSize

offragment

(basepairs)Refs.Integrin

@i5' AAT 000 AAC AAC GAG GTC ATG GU 3'5' UG TOG OAT UG CAC 000 CAG TAC 3'60°C300Bates

et aL, 1991(28)Integrin

1335' TOC TCA UG 0CC TFG CCG CCC TOC 3'5' TGA itT GAG OAT GAC TOC rFA TCA 3'60200Bates

et aL, 1991(28)Integrin

@5'GAC ACC ACG GOC ACC TAC AC 3'5' CCA CCF CCF COG CrC TCT TA3'601132Integrin

a25' TOG GOT GCA AAC AGA CAA GO 3'

5' GTA GOT CFG CI'G GU CAG C 3'60541Milamet aL, 1991(29)Integrin

a35' TOG GCA OAT OGA TOT OGA iDA GAA 3'5' OATOATOAT000 OCOGAOlTr GTC3'60406aIntegrin

a55' OCO CrC CAC TOT ACA OCr 0 3'5' CAO CAA OTC ATC CAO CCC 03'60564Milam

et aL, 1991 (29)

Integrin a,, 60 619 Milam et aL, 1991 (29)

DETERMINANTSOF ASTROCYFOMAMIGRATION

dishes covered with cell-derived ECM were used for adhesion and migrationstudies within 1 month of preparation. Preliminary studies showed that storage

interval did not affect the ability of the surfaces to support cell adhesion, as hasbeen reported by Jones (24).

Adhesion Assay (±Blocking Antibodies). Cells were tested for adhesionusing a method modified from that described by Kramer et aL (25). Briefly,cells were harvested from monolayer culture, suspended at a concentration of106 cells/mI, and then deposited into microtiter wells (50,000 cells/well). The

microtiter plates were incubated for 30 min on ice and then at 37°C for 60 min

to allow adhesion. The plates were then subjected to vigorous agitation (350rpm for 6 min on a horizontal rotator), after which nonattached cells wereremoved by aspiration and rinsing with PBS. Attached cells were fixed in 1%glutaraldehyde and stained using crystal violet (0.1% in H20). Absorbance ofstained nuclei was quantified using spectrometry at 540 nan (BioTek PlateReader; BioTek, Winooski, VT). Replicates from 6 wells were averaged. Thenumber of attached cells is reported as absorption units for each cell line,

which was linear over the range of cell numbers studied. The quantitativeuptake of stain by the different cell lines, however, was not the same, andsimilar absorbance values do not reflect the same number of cells for each line.

In experiments using anti-integrin antibodies, the antibody was added to thewells prior to the cells. Anti-@1integrin antibody was the hybridoma supernatant AIIB2 (kindly provided by Caroline Damsky, University of Californiaat San Francisco, San Francisco, CA) which is biologically active in blockingthe ligand-binding site of f31-containingintegnins (26).

RT-PCR Analysis of Integain Subunits. Total RNA was isolated frommonolayer cultures according to Chomczynski (27) and quantified by measuring absorbance at 260/280 nm. Reverse transcniptasesynthesis of complementary DNA was done using a First Strand Synthesis Kit (Stratagene, La Jolla,

CA) accordingto manufacturer'srecommendations;aliquotsof 2.5 @lwereused for PCR. Primer pairs for each of the integnin subunits are shown in Table

1.These were derived from published material (28, 29) or were designed basedon available integrin sequences with the assistance of Oligo 4.1 PrimerAnalysis Software (National BioSciences, Inc., Plymouth, MN). Amplificationof DNA [2.5 ILlof RT product, 2.5 @.dofeach primer, 0.5 @.dofTaq polymerase(Perkin Elmer/Cetus), and 2.0 p1 of nucleotides in 10 ,.d of lOX buffer, totalvolume of 100 @l]was allowed to run for 22—38cycles (1 min at 94°C,1 missat 60°C,and 2 min at 72°C)(MJ Research, Inc., Watertown, MA); aliquots of9 @lwere collected every 2 cycles. These aliquots were run in 2% agarose gels,stained with ethidium bromide, and photographed with Polaroid 665 film underUV illumination. The film negative was scanned for absorbance using visiblelight (Beckman DU-70). Absorbance values versus distance along the gel werecollected and analyzed for peak area using PeakFit (Jandal Scientific, San

Rafael, CA). The quantitative measures are reported as the interpolated number

of PCR cycles needed to generate a product of peak area equal to 0.1absorbance unit/mm. The molecular weights of the products were calculatedrelative to known size standards run in separate lanes on the same gels.

Migration Assay. A recently developed monolayer migration assay (30)was used to investigate the influence of different ECM proteins on astrocytomamovement. Briefly, surfaces of 8-well LabTek chamber slides (Nunc, Inc.,Naperville, IL) were coated with ECM proteins or astrocytoma-derived ECM

as described above. Custom produced sedimentation cylinders (G&S Glass,Gilbert, AZ) cut from 1-mi glass pipets with an internal dimaeter of 1 mm wereplaced into each chamber with 200 gd of culture media. A 1-pi cell suspensioncontaining 2000 cells was added to each cylinder; the slide was kept on ice for60 min to allow cell sedimentation and then incubated overnight. The cylinderswere removed and fresh media supplemented with FCS and autologous astrocytoma-conditioned media were added. Daily serial diameter measurements ofthe cell population area were made using inverted microscopy (Axiovert;Zeiss, Thornwood, NY) and image analysis (VIDAS; Kontron, Eching, Ocrmany). The diameter of the cell population increased linearly over time (30).

For contingency analysis (@), each ECM protein studied with each cell linewas ranked from 1 to 4 based on the extent of cell attachment and on the degreeof migration. Substrates to which the cells bound or moved most rapidly werescored as 1; 4 was assigned to the substrate of least adhesion or migration. In

instances where results were equivocal for different proteins, the rankingscores were averaged.

Actin Staining. Cells deposited within sedimentation cylinders as described above were allowed to migrate for 1—2days and then fixed for 30 mm

in 1% paraformaldehyde at 23°C.Actin was stained using bodipy phalloidin(Molecular Probes, Eugene, OR) according to manufacturer's recommendstions. Photomicrographs of the actin cytoskeleton were made under 546 nmexcitation-580 nm emission fluorescence microscopy (Zeiss Axioplan, rhodamine ifiter set).

RESULTS

Adhesion of Cells to ECM Proteins. Human astrocytoma cellsand normal cultured astrocytes adhered after a brief incubation todifferent purified human ECM proteins in a dose-dependent manner(Fig. 1). Relative to plastic “blocked―with bovine serum albumin,each cell line showed a preference for specific adhesion to at least oneECM protein. There was no “astrocytoma-specffic―ECM which wasconsistently most effective for adhesion. Laminin was frequently theoptimal adhesive substrate (5 of 7 cell lines); however, vitronectin andfibronectin were each optimal in separate instances (G-22 and G-118,respectively).

The best studied integrmns which mediate specific adhesion ofnonhematopoietic cells to ECM proteins are those containing a f3@subunit (11). When anti-@1 antibodies (AIIB2) were tested for theireffect on SF767 astrocytoma adhesion to different ECM proteins, it

5'GAG CAG CAA GGA CU TOG G 3'5' 006 TACACTTCAAGACCAGC3'

a Gene sequences were retrieved from the Rockefeller Gene Bank and primer sequences were designed using Oligo 4.1.

3898

0—22li_

i.i@HinIt

G—112@:

:•:

@ftfl@

1m_ui.I

251@ 00

30@I35L@40 L . __

SF—767

@ hull I @@•@ii ILaminin Collagen

DE'I'ERMINANTSOF ASTROCYIOMA MIGRATION

25

30

35

40

25

30

35

40

::@@ I

c—i1225

30

35

40

25

3035

40

0Ca

.0

0C,)@0

4).rC

E:Cz

4)0

U,4)

0>@

0

00@

0—118

I. _i@,.25

30

35

40

25

30

35

40

—“U,In @33@14o2 cs3 oc5

Integrin Subunit

Fig. 3. Expression of integrin subunits in astrocytoma cells determined by RT-PCR.interpolated PCR cycle numbers needed to produce a product of 0.1 absorbance unit/mmpeak area are shown for the integrin subunits from the different astrocytoma cell lines. Theabsence of bars indicates that a product did not appear after 38 cycles.

0.8

0.6

0.4

0.2

SF—767

I @.LIEEE

a :;@ -2@ :;;@@

Fig. 1. Effect of different ECM protein coatings on the adhesion of human astrocytomacells. Collagen type IV (Coil), fibronectin (FN), laminin (Lam) and vitronectin (Vil) wereincubated in microtiter plates at concentrations of 1, 10, and 100 @tg/mito coat the surface.Astrocytoma cells were deposited and incubated for 60 mlii to allow specific attachment.Nonattached cells were removed by shaking, aspirating, and rinsing. Cell number wasdetermined by crystal violet spectroscopy of stained nuclei. Columns, mean of triplicatedeterminations; bars, SD; BSA, bovine serum albumin.

1.6

1 .4

1.2

1.0

0.8

0.6

0.4

0.2

0.0

0a0

a.0Ez

a0

< LI) 0 0 0 0 0(11 C@.J e@ 0 0 0 0

@ . ‘- N@

Fig. 2. Effect of anti-@1 integrin antibody on astrocytoma cell attachment to purified ECM proteins.Antibody was added to microtiter wells prior toadding the cells for the adhesion assay. Columns,mean of triplicate determinations; bars, SD.

.0 @t) 0 0 0 0 0< C― @() 0 0 0 0

Dilution of Anti j@1Antibody3899

iiiiniii @iI@

U—251

I, i,,u

G-22

0 24 48

G-28

480

0 24

Hours in Culture

DETERMINANTh OF ASTROCYTOMA MIGRATION

f g he

Fig. 4. Human astrocytoma cells on different ECM proteins. 0-28 cells (a-d) and 0-112 (e-h) were deposited in sedimentation cylinders onto ECM-coated slides and fixed after3 days of migration. Cells were plated on laminin (a and e), collagen type IV (b and f), fibronectin (c and g), or vitronectin (d and h). 0-28 cells expanded as a contiguous front ofadvancing cells, whereas 0-112 demonstrated features of single cell outward migration.

was found that attachment to collagen was inhibited in a dosedependent manner (Fig. 2) but that only minimal effects on lamininadhesion were demonstrated. Effects on the attachment of the othercell lines to collagen and laminin by treatment with anti-@1 antibodieswas similar (data not shown). Attachment of SF767 to fibronectin andthe limited adhesion to vitronectin were also blocked by anti-@1antibodies; adhesion to laminin was not affected by either anti-@33oranti-f34 antibodies (data not shown).

RT-PCR Analysis of Integnin Expression. In orderto determinethe relative levels of expression of the integrin subunits by thedifferent astrocytoma cells, RT-PCR was done (Fig. 3). The valuesrepresent the number of PCR cycles needed to generate a product of0.1 absorbance unit/mm peak area. For each cell line, (3@dominatedthe integrin subunit expressed; however, all the cultured astrocyticcells also produced at least one other @3chain ((3@,(34, or both). Thelevels of a subunits expressed by the cells showed a wide variation.In 5 of 7 cell lines, a,,,was the most prevalent a subunit; a5 was neverthe most abundant. While this does not mean that the proteins wereexpressed at levels analogous to the mRNA or that the translatedintegnn proteins were functionally active, these results indicate that

_____________________ astrocytoma cells in culture exhibit varying profiles of integrin a and@3subunittranscription.

Migratory Response of Astrocytoma Cells to ECM Proteins.Astrocytoma cells deposited onto different purified ECM proteinsmanifest different behaviors relative to their lateral movement andpattern of dissemination. Generally, two different morphological patterns of cell dispersion were seen; these are presented using data from

100

50

0

150

100

50

0

100

50

0

In

4)x

0@

C

4)In0a)0C

(I)C

-C0

0::

Fig. 5. Migration behavior of human astrocytoma cells on different ECM proteins. Theincrease in area (measured in pixels) occupied by the migrating astrocytoma cell population is plotted as a function of time. t@.,laminin; 0, collagen type IV; 0, fibronectin; V,vitronectin. Points, means of triplicate determinations; bars, SD.

48

3900

•s,1 @,

•0@'@ @j@*dI@

_3 I@ “-@@1

-@ p @rt'@@

I @.‘

. t@

DETERMINANTS OF ASTROCYFOMA MIGRATION

0-28 and G-l 12 (Fig. 4). On all substrates, 0-28 cells expanded as a

contiguous front of advancing cells; however, on collagen 0-28 cellsshowed the fastest migration rate as well as a larger number of solitarymigrating cells (Fig. 4b). G-112 dispersed on laminin and collagen ina manner reminiscent of a starburst; these cells moved only minimallyon vitronectin, and the population appeared to retract onto itself whenplated on fibronectin.

The quantitative migratory behavior of the different cell linesshowed even greater differences. Migration was quantified as theincremental increase in area occupied by the cell population beyondthat area measured on the day the sedimentation cylinders wereremoved. This incremental expansion of the advancing rim of migrating cells follows a linear relationship over time on every substrate foreach cell line, regardless of the degree of permissiveness for migration(Fig. 5). 0-22 cells were essentially nonmigratory on collagen andexpanded on the other three ECM proteins, reaching the greatestdistance on laminin. G-28 was a relatively fast moving cell line on all

V

A V @P U

. . . .-,

V •• ••

. A

@ A

4

a 2

1 2 3

Attachment (Rank Order)

4

Fig. 6. Association between adhesion and migration of astrocytoma cells on differentECM proteins. Each ECM protein for each cell line was ranked from 1 to 4 according toits permissiveness for adhesion and migration (1 represents most adhesion or greatestmigration). A, laminin; •,collagen type N; U, fibronectin; V, vitronectin. , separation of data into 4 quadrants for t@analysis. Contingency was tested using y@analysis ofthedistributionof thedatapairs;P < 0.01.

if.i@

p -

I @i ¼

- ‘1

‘1

I,, .,*@,

, ..@ ‘-.4

Sb.@

C

,@ j't@ C

e _____________________Fig. 7. Polymerized actin cytoskeleton in human astrocytoma cells on different ECM proteins. 0-22 (a and b), 0-28 (c and d), and 0-112 (e and)) cells were seeded onto permissive

(left) or nonperinissive (right) ECM proteins. These were laminin (a and f), collagen (b and c), and fibronectin (d and e). Astrocytoma cells on permissive substrates containedpolymerized actin at the plasma membrane that modeled filopodia (a and e) and lamellipodia (c and e), whereas nonpermissive substrates induced heavy actin bundles that were largelyconfined to cytoplasmic structures (b and f) or immature lamellipodia (d).

3901

DETERMINANThOF ASTROCYFOMAMIGRATION

the substrates but showed the fastest rate of dispersion on collagen.0-1 12 cells displayed a wide range of migratory responses to thedifferent ECM proteins, with laminin being the most permissive andvitronectin being almost completely nonpermissive for movement;collagen was more permissive than fibronectin but both of these wereless effective than laminin.

The migratory behavior of astrocytoma cells on different ECMproteins demonstrated that these cells respond to their insoluble environment with consistent and stable rates of migration. The speed ofmigration on the different ECM proteins can be ranked according topermissiveness, ranging from 1 to 4, with a score of 1 indicating' ahighly migratory response. Similar scores were assigned for the adhesive behavior of the astrocytoma cells on the different ECM proteins. It was of interest to determine whether the pattern of adhesionto the different ECM proteins after 60 mm was related to the eventualmigratory behavior observed over 2 days. Paired adhesion and migration data from each cell line on each ECM protein were plotted (Fig.6). Therankeddatacouldbedividedinto4 quadrantsbasedonhigh/low attachment (ordinate) and high/low migration (abscissa).Statistically, this analysis indicates that migration of astrocytoma cellsis contingent on adhesion (P < 0.01,@ Substrates to which the cellsadhered poorly were relatively nonpermissive for migration; ECMproteins to which the cells showed a rapid attachment also wererelatively more supportive of migration.

Cytoskeletal Changes Associated with Migration. The influenceof different ECM proteins on the actin cytoskeleton was studied. Cellswere allowed to migrate in monolayer culture on surfaces coated withthe different ECM proteins and then stained for polymerized actin(Fig. 7). Cells on all substrates contained similar cytoplasmic actinbundles. Astrocytoma cells migrating on permissive substratesshowed intense staining of actin filaments that either condensedparallel to the lateral aspects of the cell membrane or projected intostructures reminiscent of broad lamellipodia and advancing filopodia.

Cells on substrates nonpermissive or less permissive for migrationshowed polymerized cytoplasmic actin bundles similar to cells onpermissive substrates; however, there were no fllopodia or maturelamellipodia. A frequent observation in nonmigrating cells was thestrong staining for actin at the plasma membrane that was interpretedto be immature lamellipodia (Fig. 7d). These structures appeared asclosely grouped, overlapping bubbles of actin-containing membrane,in contrast to more dispersed, maturing lamellipodia in migratingcells. The actin-containing plasma membrane in nonmigrating cellsappeared to be lamellipodia, which failed to develop into a “leadingedge―.Nonmigrating cells also showed less space between adjacentcells than migrating cells.

Astrocytoma-denived ECM: Adhesion and Migration. In orderto determine whether astrocytoma cells produce their own extracellular matrix conducive to migration, the influence of autologous ECMon cell migration was studied. Cells were allowed to grow at confluence for 10 days. Cells were then lysed and the residual ECM wasrinsed in PBS. Autologous cells were studied for adhesion in ECMcoated microtiter wells and for migration using sedimentation cylinders in 8-well chamber slides onto which autologous ECM had beendeposited. The rapid adherence of astrocytoma cells to autologousECM occurred to an extent similar to or greater than that to laminin(Fig. 8). However, autologous ECM was found to be less permissivefor migration than laminin (Fig. 9), with the exception of SF767 cells,which migrated more rapidly on their own ECM than on laminin.When other astrocytoma cells were tested on SF767 ECM, adhesion

and migration were also supported. These results suggest that themovement of astrocytoma cells in monolayer culture is almost exclusively determined by the nature of the surface features present on thematrix independent of modifications made by the tumor cells.

ci)0C0@00(I)@0

Lci)@0

E

z

ci)0

2.8

2.4

2.0

1.6

1.2

0.8

0.4

0.0

1.6

1.2

0.8

0.4

0.0

1.6

1.2

0.8

0.4

DISCUSSION

Our results indicate that human astrocytoma cells show preferentialadhesion and migration on specific extracellular matrix proteins andthat the preferred substrate varies according to cell line. Some celllines, such as U251-MG, 0-112, G-118, and QG-4, showed a singlepreference for high levels of rapid adhesion; others demonstratedmore promiscuous adhesion to several ECM proteins. Overall, laminm was the most effective substrate for the adhesion and migration ofastrocytoma cells.

Cell adhesion to collagen, fibronectin, and vitronectin was blockedpartially or fully by incubation of the cells with anti-@31antibodies,implicating p1-containing integrmns in adhesion. Since f3@integrindimerizes with a wide assortment of a subunits (1 1), leading toreceptors recognizing many ECM ligands, the key role of @31-containing integrins in astrocytoma adhesion is not unexpected.

It was of interest that adhesion of the astrocytoma cells to laminincould not be blocked by anti-f31 antibodies. Laminin is a complexmolecule relative to cell adhesion and has been described to containseveral domains that serve as ligands for integrins (31). The majorityof integrins which recognize different proteolytic cleavage fragmentsof laminin utilize the (3@chain combined with different a subunits(32);aj34 and @@j33havealsobeendescribedaslamininreceptors(Refs. 33 and 34, respectively). Because the different astrocytoma cell

0.0

:..::::i:::i:@@:i:

___ LamininECM

Fig. 8. Adhesion profile of human astrocytoma cells to autologous ECM. Cells werecultured 10 days past confluence and then lysed, leaving behind their ECM (see “Materialsand Methods―).The adhesion assay was done by standard procedures, using laminin andbovine serum albumin (BS4) as reference substrates for permissive and nonspecificsurfaces. Astrocytoma cells adhered as well, or better, to their own ECM as to laminin.Columns, mean of triplicate determinations; bars, SD.

3902

DETERMINANTSOF ASTROCYTOMAMIGRATION

protein activation are likely to be central features in the manifestationof functional integrmn dependent behavior of astrocytoma cells (1 1).Furthermore, the interactions of integrins with cytoskeletal elements(36) and other integrmn-mediated signal transduction mechanisms,such as phosphorylation (37—39),may override attempts at simplistic

@ correlations between levels of expression and function (40).(_) Migration of astrocytoma cells in monolayer culture on defined

Li@J ECM proteins follows a linear expansion over time (up to 3 days). All@ the cell lines migrated on at least one substrate, and several were

g@ nonmigratory on some substrates. This suggests that the cellular-@ determinants of migration are relatively unchanging and are not

-@ induced or modified over several days. There were some concerns that

4: astrocytoma cells customized their environment through the modifi

cation of an extracellular matrix that would impair our ability todetermine effects from individual ECM proteins deposited exogenously. The results in Figs. 8 and 9 demonstrate that with only oneexception, autologous cell-derived ECM is a poor substrate for astrocytoma migration, despite the finding that these same matrices wereeffective adhesive substrates for autologous cells. The one instance inwhich autologous ECM was a superior migratory substrate (SF767),also proved it to be a substrate on which other astrocytoma cells freely

@ moved. This diverges from the correlation between adhesion andLU migration seen on purified ECM proteins. It is likely that the concert

@ of ligands represented on cell-derived ECM activates a diversity of@ integrmns on adhering and migrating cells, which would be different

@2 thanthoseactivatedby matricesof single,purifiedproteins.These.@ results also suggest that integrins which mediate adhesion are not

@@4: necessarily the same integrins which mediate migration. The data do,

however, allow the conclusion that astrocytoma cells in culture areopportunistic for migration and do not generally synthesize their ownmatrix for migration.

Local environmental factors (i.e., the composition of the matrix orthe repertoir of cell surface ligands) determine the migratory behaviorof transplanted mouse astrocytes rather than the developmental stage

@ of the astrocytes (fetal versus adult) (41) or the anatomical regionuJ from which the transplanted cells are harvested (42). Astrocytes

@0 derived from different brain regions do not contain homing informa

.@ tion that targets their eventual migratory destination but rather move

@3 accordingto localenvironmentalcues(43).Ourresultssuggestthata localfactors(ECM,cell ligands,or localsolublefactors)in particular

@ regions of the brain may also dictate the patterns and routes ofr-@- astrocytomainvasion.u_ Theappearanceof thepolymerizedactincytoskeletonin cellson(1) permissive and nonpermissive substrates provides a lead for mecha

nisms by which the ECM modifies motility behavior of astrocytomacells. Actin remodeling is a critical mechanical determinant of cellmigration and is mediated by events at the cell membrane (44, 45).The actin cytoskeleton in the migrating astrocytoma cells in our studyshowed features of filopodia extension and maturation of lamellipodia. The latter structures could become rather large relative to theoverall cell area and yet this was highly indicative of cell migration.Nonmigrating cells were noticeably deficient in the generation offilopodia and mature lamellipodia. Frequently, the lamellipodia thatdid develop in nonmigrating cells resembled overlapping bubbles ofcondensed actin in the membrane. These we interpret to be immaturelamellipoida that were unable to advance over the substrate.

White matter tracts are the predominant route of migration fortransplanted normal astrocytes (46) and are the preferred avenues fordissemination of malignant astrocytes (47). There may be ligandswithin this anatomical structure, either as an accessible matrix forastrocytoma cells or as cell membrane epitopes, which are highlypermissive for migration and invasion. Our results strongly indicatethat migration is a constitutive behavior of human astrocytoma cells

C')

a)x

0@

C

a)(I)0a)

0C

U)

@00a:

150

100

50

0

150

100

50

0

150

100

50

0

100

50

0

200

150

100

50

0

Fig. 9. Migration behavior of human astrocytoma cells on autologous ECM (U),laminin (i@@),and bovine serum albumin (0). 0-112 and U-251M0 cells on autologousECM were nonmigratory but adopted a migratory behavior on ECM from SF767 cells.Bars, SD.

lines also contained mRNA for either@ or @34subunits, furtherstudies are in progress to determine whether integrmn heterodimerswith these subunits mediate adhesion to laminin (preliminary studiesindicate that they do not). Sobel (35) has described a nonintegrinreceptor for laminin. It is unknown whether human astrocytes orastrocytoma cells express this Mr 67,000 protein.

The inability to predict substrate adhesion from the levels of integrin subunit mRNA suggests that translational regulation and mature

0 24 48

Hours in Culture

3903

DETERMINANTSOF ASTROCYTOMAMIGRATION

which is dependent on, or modified by, the presence or absence ofpermissive ligands in the environment.

ACKNOWLEDGMENTS

Astrocytoma cell lines of the “NCE-G―series were supplied by Dr. ManfredWestphal at the University of Hamburg. Dr. Johannes Koeppen and Svenja

Dreilich (University of Hamburg) assisted in the design of the PCR studies and

Dr. Astrid Dangel assisted in the analysis. Dr. Caroline Damsky (University ofCalifornia at San Francisco) is acknowledged for supplying the anti-@31blocking antibodies. Julie Jones and John Iniguez provided technical assistance.

REFERENCES

1. Halperin, E. C., Burger, P. C., and Bullard, D. E. The fallacy of the localizedsupratentorial malignant glioma. mt. J. Radiat. Oncol. Biol. Phys., 15: 505—509,1988.

2. Sherbet, 0. V. The Metastatic Spread of Cancer. London: Macmillan Press, 1987.3. Berens, M. E., Ruth, J. T., and Rosenblum, M. L. Brain tumor epidemiology, growth,

and invasion. Neurosurg. am. North Am., 1: 1—18,1990.4. Liotta, L A., and Stetler-Stevenson, W. 0. Tumor invasion and metastasis: an

imbalance of positive and negative regulation. Cancer Res. (Suppl.), 51: 5054s—5O59s, 1991.

5. Sitrin, R. 0., Oyetko, M. R., Kole, K. L., McKeever, P., and Varani, J. Expression ofheterogeneous profiles of plasminogen activators and plasminogen activator inhibitors by human glioma lines. Cancer Res., 50: 4957—4961,1990.

6. Rao, J. S., Steck, P. A., Mohanam, S., Stetler-Stevenson, W. 0., Liotta, L. A., andSawaya, R. Elevated levels of Mr 92,000 type IV collagenase in human brain tumors.Cancer Res., 53: 2208-2211, 1993.

7. Rutka, J. T., Apodaca, 0., Stern, R., and Rosenblum, M. The extracellular matrix ofthe central and peripheral nervous system: structure and function. 3. Neurosurg., 69:155-170,1988.

8. Venstrom, K. A., and Reichard, L. F. Extracellular matrix 2: role of extracellularmatrix molecules and their receptors in the nervous system. FASEB J., 7: 996—1003,1993.

9. Hynes, R. 0. Integrmns:a family of cell surface receptors. Cell, 48: 549—554,1987.10. Palmer, E. L., Ruegg, C., Ferrando, R., Pytela, R., and Sheppard, D. Sequence and

tissue distribution of the integrin a,, subunit, a novel partner of f3, that is widelydistributed in epitheia and muscle. J. Cell Biol., 123: 1289—1297, 1993.

11. Hynes, R. 0. Integrmns:versatility, modulation, and signaling in cell adhesion. Cell,69: 11—25,1992.

12. Burride, K., Fath, K, Kelly, T., Nuckolls, 0., and Turner, C. Focal adhesions:transmembrane junctions between the extracellular matrix and the cytoskeleton.Annu. Rev. Cell Biol., 4: 487—525,1988.

13. Tawil, N., Wilson, P., and Carbonetto, S. Integrins in point contacts mediate cellspreading: factors that regulate integrin accumulation in point contacts versus focalcontacts.J. CellBiol.,120:261—271,1993.

14. Lana, E. 3. and Hitt, A. L. Cytoskeleton-plasma membrane interactions. Science(Washington DC), 258: 955—963,1992.

15. Kornberg, L 3., Earp, H. S., Turner, C. E., Prockop, C., and Juliano, R. L Signaltransduction by integrins: increased protein tyrosine phosphorylation caused byclustering of@ integrmns.Proc. NaIl. Mad. 5th. USA, 88: 8392-8396, 1991.

16. Schwartz, M. A., Ingber, D. E., Lawrence, M., Springer, T. A.. and Lechene. C.Multiple integrins share the ability to induce elevation of intracellular pH. Exp. CellRes., 195: 533—535,1991.

17. Hibbs, M. L., Xu, H., Stacker, S. A., and Springer, T. A. Regulation of adhesion toICAM-1 by the cytoplasmic domain of LFA-1 integrin@ subunit. Science (WashingtonDC),251:1611—1613,1991.

18. Chan, B. M. C., Kassner, P. D., Schiro, J. A., Byers, R., Kupper, T. S., and Hemler,M. E. Distinct cellular functions mediated by different VIA integrmna subunitcytoplasmic domains. Cell, 68: 1051—1060,1992.

19. Westphal, M., Hansel, M., Muller, D., Lass, R., Kunzmann, R., Rohde, E., Koenig,A., Hölzel,F., and Herrmann, H-D. Biological and karyotypic characterization of anew cell line derived from human gliosarcoma. Cancer Res., 48: 731—740,1988.

20. Westphal, M., Hansel, M., Hansel, W., Kunzmann, R., Hölzel,F., and Herrmann,H-D. Karyotype analyses of 20 human glioma cell lines. ArIa Neurochir., 126: 17—26,1993.

21. Bigner, D. D., Bigner, S. H., and Ponten, i. Heterogeneity of genotypic and phenotypic characteristics of 15 permanent cell lines derived from human gliomas. J.Neuropathol. Exp. Neurol., 40: 201—209,1981.

22. Berens, M. E., Weisman, A. S., Spencer, D. R., Dougherty, D. V., Elliger, S. S., and

Rosenblum, M. L. Growth properties and oncogene expression in 2 newly derivedglioma cell lines: assessment of growth determinants. Proc. Am. Assoc. Cancer Res.,31: 273, 1991.

23. Vlodavsky, I., Levi, A., Lax, I., Fuks, Z., and Schiessinger .1. Induction of cellattachment and morphological differentiation in a pheochromocytoma cell line andembryonal sensory cells by the extracellular matrix. Dcv. Biol., 93: 285-300, 1982.

24. Jones, P. A. Construction of an artificial blood vessel wall from cultured endothelialand smooth muscle cells. Proc. Nati. Acad. Sd. USA, 76: 1882-1886, 1986.

25. Kramer, R., Mc Donald, K. A., Crowley, E., Ramos, D., and Damsky, C. H.Integrin-related complexes mediate the adhesion of melanoma cells to basementmembrane. Cancer Res., 49: 393-402, 1989.

26. Hall, D. E., Reichert, L F., Crowley, E., Jolley, B., Moezzi, H., Sonnenberg, A., andDamsky, C. H. The a1f3, and a,j3, integrmnheterodimers mediate cell attachment todistinct sites on laminin. J. Cell Biol., 110: 2175—2184,1990.

27. Chomczynski, P., and Sacehi, N. Single step method of RNA isolation by acidguanidium thiocyanate-phenol-chloroform extraction. Anal. Biochem., 162: 156—159, 1987.

28. Bates, R. C., Rankin, L M., Lucas, C. M., Scott, J. L, Drissansen, 0. W., and Burns,0. F. Individualembryonicfibroblastsexpressmultiple@3chainsin associationwiththe a,, integrmnsubunit. J. Biol. Chem., 266: 18593—18599,1991.

29. Milam, S. B., Magnuson, V. L, Steffensen, B., Chen, D. and KIebe, R. J. IL-lB andprostaglandins regulate integrin mRNA expression. J. Cell. Physiol., 149: 173—183,1991.

30. Berens, M. E., Rief, M. D., Loo, M. A. and Oiese, A. A microliter-scale assay for invitro studies on the role of ECM, integrins, and stromal or parenchymal cells on tumorcell migration and proliferation. Clin. Exp. Metastasis, in press, 1994.

31. Timpl, R., Aumailley, M., Gerl, M., Mann, K., Nurcombe, V., Edgar, D., andDeutzmann, R. Structure and function of the laminin-nidogen complex. Ann. NYAcad. Sci., 580: 311—323,1990.

32. Sonnenberg, A., Linders, C. J. T., Moddermann, P. W., Damsky, C. H., Aumailley,M., and Timpl, R. Integrin recognition of different cell-binding fragments of laminin(P1, 83, E8) and evidence that a,.fi, but not@ functions as a major receptor forfragment ES. J. Cell Biol., 110: 2145—2155,1990.

33. Lee, E. C., I@otz,M. M., Steele, 0. D., and Mercurio, A. M. The integrin a@j34is alaminin receptor. J. Cell Biol., 117: 671—678, 1992.

34. Kramer, R. H., Cheng, Y. F., and Clyman, R. Human microvascular endotheial cellsuse $, and 133integrin receptor complexes to attach to laminin. J. Cell Biol., 111:1233—1243,1990.

35. Sobel ME. Differential expression of the 67 kDa laminin receptor in cancer. Semin.CancerBiol.,4: 311—317,1993.

36. Carpen, 0., Pallai, P., Staunton, D. E., and Springer, T. A. Association of intercellularadhesion molecule-i (ICAM-1) with actin-containing cytoskeleton and a-actinin.J. Cell BioL, 118: 1223—1234,1992.

37. Gum, J-L, and Shalloway, D. Regulation of focal adhesion-associated proteintyrosine kinase by both cellular adhesion and oncogenic transformation. Nature(Lond.), 358: 690—692,1992.

38. Huang, M-M., Lipfert, L, Cunningham, M., Brugge, J. S., Ginsberg, M. H., andShattil,S. J. Adhesiveligandbindingto integrmna1@,fi3stimulatestyrosinephosphorylation of novel protein substrates before phosphorylation of pp125'@'@. J. Cell Biol.,122: 473—483,1993.

39. Widmer, F., and Carom, P. Phosphorylation-site mutagenesis of he growth-associatedprotein OAP-43 modulates its effects on cell spreading and morphology. J. Cell Biol.,120: 503—512,1993.

40. Zachary, I., and Rozengurt, E. Focal adhesion kinase (pl25@'): a point of convergence in the action of neuropeptides, integrins, and oncogenes. Cell, 71: 891—894,1992.

41. Thou, H. F., and Lund, R. D. Migration of astrocytes transplanted to the midbrain ofneonatal rats. J. Comp. Neurol., 317: 145—155,1992.

42. Hatton, J. D., Garcia, R. and U, H. S. Migration ofgrafted rat astrocytes: dependenceon source/target organ. Ohs, 5: 251—258,1992.

43. Jacque, C., Suard, I., Collins, P., and Baumann, N. Migration patterns of donorastrocytes after reciprocal striatum-cerebellum transplantation into newborn hosts.J. Neurosci. Res., 29: 421—428,1991.

44. Tilney, L 0., Bonder, E. M., and DeRosier, D. J. Actin filaments elongate from theirmembrane-associated ends. J. Cell Biol., 90: 485—494,1981.

45. Cramer, L., and Mitchison, T. J. Moving and stationary actin filaments are involvedin spreading of postmitotic PtK2 cells. J. Cell BioL, 122: 833—843,1993.

46. Harvey, A. R., Fan, Y., Beilharz, M. W., and Grounds, M. D. Survival and migrationof transplanted male gus in adult female mouse brains monitored by a Y-chromosome-specificprobe.BrainRes.Mol.BrainRes.,12: 339-343,1992.

47. Schiffer, D. Neuropathology and imaging. The ways in which glioma spreads andvaries in its histological aspect. In: M. D. Walker and D. 0. T. Thomas (eds.), Biologyof Brain Tumour, pp. 163—172.Boston: Martinus Nijhoff, 1986.

3904