cell migration strategies in 3-d extracellular matrix: differences in morphology, cell matrix...

Cell Migration Strategies in 3-D Extracellular Matrix:Differences in Morphology, Cell Matrix Interactions,and Integrin FunctionPETER FRIEDL,1* KURT S. ZANKER,2 AND EVA-B. BROCKER1

1Cell Migration Laboratory, Department of Dermatology, University of Wurzburg, Germany2Institute of Immunology, University of Witten/Herdecke, Germany

KEY WORDS collagen matrix; tumor cells; T lymphocytes; dendritic cells; matrix remodeling

ABSTRACT Cell migration in extracellular matrix is a complex process of adhesion anddeadhesion events combined with cellular strategies to overcome the biophysical resistance imposedby three-dimensionally interconnected matrix ligands. Using a 3-D collagen matrix migration modelin combination with computer-assisted cell tracking for reconstruction of migration paths andconfocal microscopy, we investigated molecular principles governing cell-matrix interactions andmigration of different cell types. Highly invasive MV3 melanoma cells and fibroblasts are large andhighly polarized cells migrating at low speed (0.1–0.5 µm/min) and at high directional persistence.MV3 melanoma cells utilize adhesive migration strategies as characterized by high b1 integrinsurface expression, b1 integrin clustering at interactions with matrix fibers, and b1 integrin-mediated adhesion for force generation and migration. In contrast, T lymphocytes and dendriticcells are highly mobile cells of lower b1 integrin expression migrating at 10- to 40-fold highervelocities, and directionally unpredictable path profiles. This migration occurs in the absence of focaladhesions and largely independent of b1 integrin-mediated adhesion. Whereas cell-matrix interac-tions of migrating tumor cells result in traction and reorientation of collagen fibers, partial matrixdegradation, and pore formation, leukocytes form transient and short-lived interactions with thecollagen lacking structural proteolysis and matrix remodeling. In conclusion, the 3-D extracellularmatrix provides a spatially complex and biomechanically demanding substrate for cell migration,thereby differing from cell migration across planar ligands. Highly adhesive and integrin-dependentmigration strategies detected in morphologically large and slowly migrating cells may result inreorganization of the extracellular matrix, whereas leukocytes favor largely integrin-independent,rapid, and flexible migration strategies lacking typical focal adhesions and structural matrixremodeling. Microsc. Res. Tech. 43:369–378, 1998. r 1998 Wiley-Liss, Inc.

INTRODUCTIONCell migration in different tissues occurs in a variety

of cell types either constitutively or after activation.Migration is a constitutive feature of leukocytes andfibroblasts throughout life. If required, enhanced migra-tion of these cells occurs upon inflammation and woundrepair. In the case of tumor metastasis, neoplastic cellspenetrate the tissue ultimately disseminating throughthe body via lymphatics and/or blood. Up to now themolecular mechanisms governing the migration of suchdiverse kinds of cells in different environments arebelieved to follow a common migratory program. Hapto-kinetic, adhesion receptor-dependent strategies of cellmigration across planar substrate are well establishedfor fibroblasts and some tumor cells (reviewed in Aki-yama et al., 1995; Huttenlocher et al., 1995; Lauffen-burger and Horwitz, 1996). However, it possible thatspecial migration strategies are present in different celltypes. In the case of migrating leukocytes, morphology,and cell shape, velocity and cytoskeletal organizationmay greatly differ from larger and potentially morecomplex cells (Lee et al., 1993a; Schor et al., 1983).Furthermore, migration across two-dimensional (2-D)surfaces lacking biophysical resistance towards the

advancing cell body should differ from migrationthrough three-dimensional (3-D) extracellular matrix(ECM) environments (Heino, 1996). In this study, weexplore cellular and molecular mechanisms of cellmigration in 3-D collagen matrices supporting essentialdifferences in migration strategies in different celltypes, i.e., T cells, dendritic cells, fibroblasts, and tumorcells.

IMPACT OF TISSUE ARCHITECTUREON CELL MIGRATION

For cell migration, at least two biomechanicallydifferent environments are encountered in differenttissues, i.e., migration in 3-D environments and migra-tion on surfaces.

*Correpondence to: Peter Friedl, M.D., Ph.D., Department of Dermatology,University of Wurzburg, Josef-Schneider-Str. 2, 97080 Wurzburg. E-mail: [email protected]

Received 19 June 1998; accepted in revised form 6 July 1998Contract grant sponsor: Deutsche Forschungsgemeinschaft; Contract grant

number: Fr 1155/2–1; Contract grant sponsor: Wilhelm-Sander Foundation;Contract grant number 96.030.1.

MICROSCOPY RESEARCH AND TECHNIQUE 43:369–378 (1998)

r 1998 WILEY-LISS, INC.

Migration in 3-D Extracellular MatrixThe three-dimensional extracellular matrix present

in interstitial tissues is the prototypic substrate for cellmigration in embryonic morphogenesis, immune de-fense, wound repair, and tumor invasion (Bischop,1997; Gumbiner, 1996; Turley et al., 1985). The extracel-lular matrix provides a 3-D adhesive substrate ofmultivalent interconnected ligands and also imposes aspatial/physical barrier to the cells (Heino, 1996). Formigration, the cell must not only interact with matrixligands for force generation but also develop strategiesto overcome biomechanical resistance imposed by thematrix network.

In general, cell-to-ligand interaction is thought tooccur via specific receptor-mediated adhesion to thesubstrate (Burridge and Chrzanowska-Wodnicka, 1996)accompanied by cytoskeletal action (Stossel, 1993).However, interactions with 3-D extracellular matrixmay not exclusively involve specific integrin-mediatedinteractions, but also low-affinity binding residuesand/or non-adhesive contacts resulting from shapechange (Friedl et al., 1998a; Mandeville et al., 1997;Schor et al., 1983). Hence, the total of simultaneouslyoccurring specific high-affinity as well as of low-affinityor biophysical interactions between cell surface andtissue environment contribute to the migratory process(Friedl et al., 1998a; Mandeville et al., 1997; Van derMerve and Barclay, 1994). In vitro, this migration typeis supported by 3-D extracellular matrix-based tissuemodels (Gershman, 1982) frequently using 3-D collagenlattices (Elsedale and Bard, 1972) in the presence orabsence of additional ECM components such as fibronec-tin and hyaluronan (Friedl et al., 1995a; Turley et al.,1985). The ultrastructure of these matrix models closelymimics the biochemical and biophysical architecture ofinterstitial tissues (Cidadao, 1989; Turley, 1985).

In the past, we have developed a 3-D collagen matrixin vitro model allowing quantitative and representativeanalysis of cell migration in space and time. Cells areeither suspended in collagen solution prior to polymer-ization of the lattice (Friedl et al., 1993) or allowed toinvade a preformed collagen lattice (Friedl et al., 1995b).Migration is monitored by time-lapse videomicroscopyand manual cell tracking (Friedl et al., 1993) or bycomputer-automated three-dimensional cell tracking(Friedl et al., 1994; Noble, 1987). The analysis ofmigratory parameters of randomly selected cell popula-tions includes migration velocity from step to step,stop-go patterns, the number of cells migrating at anytime point, the cumulative time individual cells aremigrating, angle distribution from step to step, and anoverall representation of the path (Friedl et al., 1993).Dynamic changes in migration, i.e., after migratoryinduction or inhibition, are obtained by continuous-time analysis for (sub)populations from step to step(time-dependent analysis) or assessed for (sub)groupsof individual cells during defined time intervals on acumulative basis (single cell analysis) (Friedl et al.,1993, 1995a).

Migration Across 2-D SurfacesMigration on 2-D surfaces composed of extracellular

matrix primarily occurs in epithelialization duringmorphogenesis or upon wound healing. Keratinocytes

migrating across a matrix defect are apparent at theedges of the wound synthesizing ECM while migrating(Lee et al., 1993b; see Clark et al. and Hudson et al.,this issue). In keratinocytes, deposition of ECM compo-nents and the remodeling of preexisting matrix cannotbe separated from the migration process per se (Pilcheret al., 1997). Cell migration across surfaces requires acertain degree of attachment to the underlying sub-strate, adhesion receptor coupling to the cytoskeleton,directional cytoskeletal action and coordinated detach-ment from the trailing edge (Huttenlocher et al., 1995;Palecek et al., 1997; Schmidt et al., 1993). In vitro, thismigration type is usually mimicked in haptokineticmigration models (for details see Lauffenburger et al.and Wells et al., this issue) favoring adhesion-depen-dent migration in the absence of biophysical matrixresistance or passive cell-matrix interactions (Lee etal., 1993a). Such graded adhesive interactions deter-mine not only the cell shape but also mediate substrate-guided formation and dynamic turn-over of focal con-tacts and, eventually, migration (Lauffenburger andHorwitz, 1996).

As a special form of migration on 2-D surfaces, cellscrawl across cell surfaces and through intercellularjunctions within multicellular complexes. Here, cellmigration represents a transient event in the process ofmigratory cell-cell interactions, such as observed intransendothelial leukocyte migration (Butcher, 1991),the vigorous crawling of T lymphocytes while scanningthe surface of antigen-presenting cells (Negulescu etal., 1996; Nikolai et al., 1998) or the migration ofdendritic cells inbetween keratinocytes emigrating fromthe epidermis following injury (Steinmann et al., 1995).The migration-driving force is provided by transientadhesive cell-cell junctions and, in principle, follows thebiomechanics of haptokinetic migration across surfaceswith the difference that cell-cell adhesion receptorsincluding aLb2 (LFA-1) and b7 integrins rather thanECM-binding adhesion molecules are involved (Dustinet al., 1992; Del Pozo et al., 1996).

CELL BIOLOGY AND ADHESION MOLECULESGOVERNING CELL MIGRATION IN 3-D

COLLAGEN LATTICESMultiple cell characteristics and functions may con-

tribute to migration in 3-D connective tissues. The totalbinding strength may be a function of expression andactivation level of integrins, also determining the qual-ity of the off-rates for detachment and, consequently,the migration speed (Palecek et al., 1997). Biophysicalmigration properties are determined by morphologicalfeatures including size, shape, and flexibility of the cellbody relative to the spatial properties of the ECM.Furthermore, the capacity to reorganize the ECM, i.e.,proteolysis (Birkedal-Hansen, 1995) and/or the forma-tion of paths of minor resistance may favor tissuepenetration (Brooks et al., 1996) and also to facilitatemigration of following cells (Friedl et al, 1997a; Thomasand Yamada, 1992). In certain cases, such as angiogen-esis and tumor invasion (Friedl et al., 1995b; Gumbiner,1996), the maintenance of cell-cell junctions resultingin the migration of multicellular complexes (see below)may require special migratory action and compart-mentalization, i.e., the directed mechanical and

370 P. FRIEDL ET AL.

proteolytic action at the leading edge for force genera-tion and matrix remodeling (Davis and Camarillo,1995; Tusch et al., 1998).

Cell Morphology and PolarityFor most cells, migration is assumed to follow a

single cell pattern and, consequently, most of the experi-mental data is based on individually migrating cells(Friedl and Brocker, 1997). Cell morphology is deter-mined by an inherent program characteristic for everycell type, but is also reactive to the nature and spatialproperties of the extracellular environment (Davis andCamarillo, 1995; Dickinson et al., 1994; Ezzel et al.,1993; Gumbiner, 1996). In 2-D migration models depend-ing on adhesiveness, an almost complete lack of bio-physical resistance may result in greatly varying cellmorphologies ranging from ameboid and simple-shaped(Dictyostelium, leukocytes) (Del Pozo et al., 1996;Niewohner et al., 1997), spread-out (keratinocytes,endothelial cells) (Lee et al., 1993b), to bi- to multi-polar (fibroblasts, myoblasts) (Abercrombie et al., 1970;Hynes and Lander, 1992). After transfer from a culturedish into 3-D matrices, a previously spread-out morphol-ogy rapidly changes towards a spindle-like elongatedshape as was shown for embryonic fibroblasts, endothe-lial and epithelial cells, and many tumor cell lines(Friedl and Brocker, 1997; Heath and Reachey, 1989;

Niggemann et al., 1997; Noble and Levine, 1986; Noble,1987; Schor et al., 1985). The transition from initiallyround (non-motile) to polarized and spindle-shapedimplicates the development of a tensile force betweenthe cell edges (Shyy and Chien, 1997). These shapechanges require cytoskeletal action (Guilford et al.,1995; Heidmann and Buxbaum, 1998) and are accompa-nied by the redistribution of cell surface receptorsincluding integrins, signal induction, cell spreadingand/or contraction, and the induction of gene expres-sion manifested by both degradation and synthesis ofextracellular macromolecules (Aggeler et al., 1984;Akiyama et al., 1995; Heino, 1996; Langholz et al.,1995).

Cell Size, Migration Paths, and SpeedIf cells of different ontogeny are compared, cell size,

and morphology are a hallmark of the cellular differen-tiation and the activation state, but are also modulatedby the environment encountered (Ezzel et al., 1993;Gumbiner, 1996). In 3-D collagen lattices, migratingfibroblasts and tumor cells (Fig. 1) frequently develop alength axis of up to 100 µm and the mean maximaldiameter of the cell body may surpass 20 µm (Heathand Reachey, 1989; Niggemann et al., 1997; Noble andLevine, 1986). The leading edge consisting of one orseveral pseudopods protrudes for the creation of new

Fig. 1. Differences in cell size and morphology in T lymphocytes,dendritic cells, and MV3 melanoma cells migrating in 3-D collagenmatrices. Images were obtained from time-lapse video recordings. The1:1 ratio of cell size relative to the MV3 melanoma cell is indicated in

insets. MV3 melanoma cells are large bi- to tetrapolar cells, whereasleukocytes are 5- to 10-fold smaller. White arrows indicate thedirection of migration.

371CELL MIGRATION IN 3-D MATRIX

attachments followed by forward traction of the cellbody, putatively provided by myosin motors (Heidmannand Buxbaum, 1998). Consequently, the migration pathsof MV3 melanoma cells in the absence of chemotacticgradients are directionally persistent and rarely inter-rupted by changes in direction (Friedl and Brocker,1997). Here, the dimension of the length axis andputative high morphological rigidity coincide with highdirectional persistence (Niggemann et al., 1997).

In contrast, migrating T cells are very flexible cellsdeveloping a length axis of 10 to 15 µm and consisting ofa leading edge and a trailing uropod (Fig. 1). Theleading edge exhibits dynamic membrane oscillationsforming new contacts with collagen fibers, as obtainedfrom confocal reflection time-series (Friedl et al., 1998a).The uropod is rich in adhesion molecules and is as-sumed to anchor the cell to the substrate therebystabilizing the 3-D orientation of the cell body (Del Pozoet al., 1996; Friedl et al., 1998a).

In dendritic cells, size varies from 15 to 40 µm inlength axis (Fig. 1), generating multiple highly dy-namic spider-like pseudopods at the leading edge forforce generation. The cell body is pulled forward byattachments at the leading edge and considerabledeformation of the cell body relative to fiber strands isfrequently observed (Gunzer et al., 1997).

Importantly, in 3-D collagen lattices, morphologicallylarge cells, as determined by the longitudinal axis afterpolarization and by confocal cross-sectional area of thecell body (Friedl et al., 1997), migrate at low speed. Theaverage T cell velocity in periods of active migration, asmeasured by computer-assisted cell tracking of indi-vidual paths, is 7 µm/min (Fig. 2) and peak velocities of25 µm/min are not uncommon (Friedl et al., 1994).Dendritic cells locomote at 3-fold lower velocities andmaximal rates of 5 µm/min are observed (Fig. 2). Incontrast, migration of MV3 melanoma cells ranges at amean velocity of 0.2 µm/min and the peak speed doesnot exceed 1 µm/min. This difference in migrationvelocity is highly representative for most invasive andmetastatic tumor lines investigated in 3-D collagen gels(Niggemann et al., 1997) and also corresponds to themigration speed of fibroblasts (0.2– 1 µm/min) (Maaserand Friedl, unpublished data). In conclusion, for migra-tion through a 3-D collagen matrix, cell size may beinversely correlated with migration speed potentiallyreflecting both adhesion receptor function and biophys-ics of cell-matrix interactions.

Expression Levels of Integrinson Different Cell Types

In direct comparison, expression levels of b1 inte-grins vary considerably between different cell types.Based on flow cytometry, b1 integrin expression in MV3melanoma cells is 35-fold higher as compared to den-dritic cells, and approximately 250-fold higher than themean b1 expression on freshly isolated human periph-eral T lymphocytes (Fig. 3). In fibroblasts, b1 integrinexpression ranges approximately 3-fold below MV3melanoma cells (Maaser et al., unpublished data) tobecome considerably upregulated upon contact withfibrillar collagen (Klein et al., 1991). This implicatesthat in 3-D collagen lattices, high b1 integrin expres-sion and adhesive strength are inversely correlated

with migration speed, as recently described for haptoki-netic migration models (Palecek et al., 1997). Thismodel is further supported by the finding that addi-tional antibody-induced conformational change en-abling a high-affinity interaction of b1 integrins withtheir ligands leads to adhesion-dependent immobiliza-tion of both MV3 cells and T lymphocytes, resulting inmarked reduction in migration velocity (Friedl et al.,1998a; Maaser et al., 1998).

Redistribution of Integrins Relative to CellMatrix Interactions and Cytoskeletal

OrganizationIn fibroblasts (Fig. 4A–C) and MV3 melanoma cells

(Fig. 4D–F) migrating within 3-D collagen lattices, b1integrins (Friedl et al., 1997) are clustered at manyphysical interactions with collagen fibers whereasirrelevant adhesion receptors such as CD44 shownon-clustered even distribution (Friedl et al., 1997).These interactions are prominent at the leading edge ofthe cell and characterized by radiary fiber alignmenttowards the cell body (Fig. 4C) and fiber bundling tosolid strands (Fig. 4F, arrowhead) (Friedl et al., 1997).After addition of blocking anti-b1 integrin antibody,integrin clustering and fiber traction as well as cellpolarization are lost (Maaser et al., 1998), indicatingthat ligand binding is required for integrin clusteringand force generation (Choquet et al., 1997). Thesefindings suggest that melanoma cell and fibroblast

Fig. 2. Significant differences in migration velocities: T lympho-cytes and dendritic cells are 10- to 30-fold faster than MV3 melanomacells. Cells were incorporated in 3-D collagen lattices and the sponta-neous migration velocity was assessed by time-lapse videomicroscopyand computer-assisted cell tracking (Friedl et al., 1993). The timeframe of cell tracking was 3 hours for T cells and dendritic cells, and 18hours for MV3 cells. One representative of n . 10 independentexperiments is shown (30 individually locomoting cells; mean values 6SD are indicated).


migration involve integrin engagement and clustering,and, putatively, the formation of focal contacts, althoughF-actin and vinculin distribution is not well correlated withb1 staining and although stress fibers are absent in thesecells (Maaser et al., unpublished observations).

In contrast, migrating T lymphocytes developingmultiple interactions with collagen fibers are free ofclustered b1 integrins (Fig. 4H, I; arrowheads) (Friedlet al., 1998a) although some clustered phosphotyrosineresidues are observed at lateral interaction sites withcollagen fibers (Entschladen et al., 1997). The leadingedge is weakly stained or b1-negative whereas promi-nent b1 staining is present in the uropod (Fig. 4H). Thearchitecture of the cytoskeleton in migrating T cellsalso differs from that of other cells. Upon spontaneousonset of migration, focal adhesion kinase becomestyrosine phosphorylated and is redistributed towardsthe leading edge (Entschladen et al., 1997), whereasF-actin forms a uniform subcortical staining patternexcluding the leading edge (Friedl et al., 1998a). Pro-tein kinase C is redistributed into the uropod(Entschladen et al., 1997) in close proximity to b1integrins and other adhesion moleules, includingICAM-1 and -3, CD44, and CD43 (Del Pozo et al., 1996).This unique compartmentalization characterizes T cellsas ‘‘bipolar sensors’’ optimized for cell-cell and cell-matrix interactions (Del Pozo et al., 1996; Negulescu etal., 1996), which, however, does not follow the focaladhesion model of cell-matrix interactions (Burridgeand Chrzanowska-Wodnicka, 1996). Because T lympho-cytes migrate so rapidly, interactions with multivalentligands are brief (contact time to individual fibersestimated to 1–4 minutes, depending on the actualvelocity) and of low integrin engagement, preventingthe development of typical focal adhesions (Burridgeand Chrzanowska-Wodnicka, 1996; Friedl et al., 1998a;Lee et al., 1993a).

Adhesion-Dependence of MigrationCoinciding with the clustering state of b1 integrins,

blocking anti-b1 antibodies result in an immediate andalmost complete abrogation of migration in MV3 mela-noma cells (Fig. 5), and to a slightly lesser extent infibroblasts (Maaser, unpublished data), as quantifiedby time-lapse videomicroscopy and computer-assistedcell tracking of the total cell paths (Friedl et al., 1993).Both, migration and migration-dependent cell polariza-tion are blocked by antibodies against b1 and a2integrins but not against a3, a5, av, and CD44 (Maaser

et al., 1998), indicating that engagement of a2b1 inte-grin is a prerequisite for the development of polarizedmorphology and migration, as established for 2-D hap-tokinetic migration models (Felsenfeld et al., 1996).

In spontaneously locomoting leukocytes, neither cellpolarization, the number of migrating cells nor theirvelocity are affected in the presence of a panel ofadhesion-blocking anti-b1 antibodies. This holds true ofhuman T cells (Friedl et al., 1998a), dendritic cells (Fig.5), and monocytic U937 cells (Friedl, unpublished data).In T cells, simultaneous blocking of b1, b2, b3, and avintegrins has no effect on cell polarity, interactiondynamics with collagen fibers and migration excludingpotential collaborative action among these integrins(Friedl et al., 1998a). This is in marked contrast topreviously described b1 integrin-dependent migrationof activated T cells across fibronectin-coated surfacesand/or filters (Arencibia and Sundquist, 1989; Davis etal., 1990; Hauzenberger et al., 1995). Although thefunction of adhesion receptors other than integrins andCD44 cannot be fully excluded at present (Friedl et al.,1995a), we propose that T cell migration in a 3-Dcollagen matrix involves highly transient and largelyintegrin-independent cell-matrix interactions providedby low-adhesive (‘‘sticky’’) cell surface residues (van derMerve and Braclay, 1994) or biophysical interactions.Such biomechanical interactions include shape change,pushing via lateral protrusions of the cell body (‘‘foot-holds’’), and squeezing through matrix pores and pathsof least resistance (Friedl et al., 1998a; Haston et al.,1982; Mandeville et al., 1997; Schor et al., 1983). Thesefindings support the concept that, in principle, a mini-mal tensile binding strength to the surrounding colla-gen fiber network may be sufficient for T cell migrationthrough 3-D ECM.

Upon activation of T cells by environmental factors,i.e., transendothelial migration, cross-linking of a4 oraL integrins, and a multitude of cytokines/chemokines,integrin function is assumed to rapidly shift towards ahigh affinity binding state and/or clustering for in-creased avidity (Friedl et al., 1995a; Gilat et al., 1994;Hauzenberger et al., 1995, 1997; Hershkoviz et al.,1992; Jiang et al., 1994; Romanic et al., 1997; Stewartand Hogg, 1996). This notion is consistent with thefinding that chemokine-triggered T cell migration dif-fers from spontaneous T cell motility in its dependenceon protein kinase C (Entschladen et al., 1997) as well asseveral integrin a chains (Friedl, 1995a), although cellmorphology, short-term integrin expression and distri-

Fig. 3. b1 integrin expression of different celltypes prior to the incorporation into collagen lat-tices. Cells were stained by anti-b1 integrin mAb4B4 (black profiles) or isotypic control antibody(open profiles) and secondary FITC-conjugatedFab-fragments and assessed by flow cytometry.MV3 cells were detached from the culture dishusing EDTA. mFl, mean fluorescence.


Fig. 4. b1 integrin clustering and matrix remodeling in migratingfibroblasts (A–C), MV3 melanoma cells (D–F) but not in T lympho-cytes (G–I). Fibroblasts and T cells were incorporated into collagenlattices for . 1 hour, fixed, stained for b1 integrins (B, H, red channel),and assessed by confocal microscopy. Because of low integrin expres-sion on T cells, five different anti-b1 mAbs were used simultaneouslyfor signal enhancement. MV3 cells were stained by non-blockinganti-a2 mAb Gi19 (E, red channel) prior to incorporation into thecollagen lattice. Collagen fibers (A, D, G, and C, F, I, green channel)were detected by confocal reflection contrast, as described (Friedl et

al., 1997). Yellow color represents the superimposition of high reflec-tion (green) and fluorescence (red). The putative direction of migrationcorresponding to the migratory morphology of these cells (as detectedby time-lapse videomicroscopy of unfixed cells) is indicated by blackarrows. Clustered integrins in areas of fiber traction (B, F) andselected non-reorganizing cell-fiber interactions (G, I) are indicated bywhite arrowheads. E, attachment (‘‘A’’) and detachment zones (‘‘D’’) ofthe MV3 cell are indicated. Bars 5 15 (C), 18 (F), and 6 µm (I). Images(D–I) are modified from Friedl and Brocker (1997) and Friedl et al.(1998a).

bution, and migration velocity do not differ from sponta-neously migrating counterparts (Friedl et al., 1998a).In conclusion, while T cells are able to migrate sponta-neously in an integrin-independent manner, activationsignals induce migratory T cell subsets of differentadhesion and signaling requirements. However, themolecular regulation of integrin function, i.e., the rela-tive contribution of affinity vs. avidity regulation aswell as integrin-mediated adhesive vs. signaling func-tion are unresolved at present (Hauzenberger et al.,1995; Lee et al., 1993a; Stewart and Hogg, 1996).

Migration-Associated Matrix ReorganizationIn fibroblasts and osteoblasts, contact with 3-D colla-

gen, not, however, attachment to a 2-D collagen-coatedsurface, leads to the induction of matrix-metalloprotein-ases (Langholz et al., 1995; Rijkonen et al., 1995). It isbelieved that proteolytic tissue remodeling is crucial tomigratory penetration of a complex tissue matrix(Brooks et al., 1996). In addition to fiber traction andalignment towards attachment sites (Fig. 4), migratingfibroblasts and tumor cells create profound changes inmatrix architecture, ultimately leading to contractionof the lattice (Klein et al., 1991). In 3-D collagenmatrices at a collagen concentration . 1 mg/ml, themean pore size between fiber strands is smaller thanthe maximal diameter of the cell body of fibroblasts andmany tumor cells, as quantified by confocal microscopy(Friedl et al., 1997). Consequently, the migratory pro-cess is accompanied by the formation of detachment-associated circumscribed pores and tube-like matrixdefects (Fig. 4D and F) (Friedl and Broker, 1997; Friedlet al., 1997). Additionally, vast amounts of patched cellsurface determinants including a2b1 integrin and CD44are deposited into these remodeled matrix areas (Fig. 4E, ‘‘D’’) (Friedl and Brocker,1997). Also, the fact thatF-actin-containing cytoskeletal components are re-leased with these deposits suggests that cell detach-ment involves not only internalization of integrinsand/or the shedding of individual molecules (Hutten-locher et al., 1995), but also detachment-associateddisruption of entire cell portions including cytoplasm,

which is reinforced by continuous shape change andblebbing of these particles directly after release (asdetected from time-lapse videorecordings; published asCD-ROM in Friedl et al., 1998b).

In contrast to melanoma cells, migrating T cells (Fig.4D) and dendritic cells morphologically adapt to preex-isting matrix structures following collagen fibers ratherthan structurally changing the matrix architecture(Friedl et al., 1998a; Gunzer et al., 1997; Schor et al.,1983). Although the function of matrix proteases wasimplicated in the migratory action of leukocytes (Lep-pert et al., 1995), the 3-D collagen matrix model provid-ing large enough matrix gaps and pores does notrequire lysis or reorientation of collagen fibers irrespec-tive of whether resting or activated T lymphocytes areused (Friedl et al., 1998a). In contrast to leukocytes,large cells of strong integrin engagement require me-chanical and putatively cell-dependent proteolytic strat-egies to create migration pathways of least resistance(Brooks et al., 1996; Davis and Camarillo, 1995; Yebraet al., 1996).

CELLULAR AND MOLECULAR DIFFERENCESIN CELL MIGRATION WITHIN 3-D

EXTRACELLULAR MATRIX: A MODELBased on cell size and morphological flexibility, cell

migration may on the one hand follow principles ofadhesive strength and slow yet steady and directionallypersistent migration, such as observed in MV3 mela-noma cells (Fig. 6). This migratory prototype includeshigh integrin expression levels, integrin clustering atinteraction sites, and structural reorganization of theECM to overcome biophysical matrix resistance. On theother hand, small and morphologically more flexiblecells expressing low integrin levels may utilize short-lived and relatively weak interactions with the extracel-lular matrix that lack focal adhesions and cannot beblocked by anti-integrin antibodies. This ‘‘ameboid’’migration type of low adhesivity is characterized byhigh migration speed, flexibility in stop-go pattern, andextensive directional oscillations following preexisting

Fig. 5. Integrin function in MV3 cells, T lymphocytes, and den-dritic cells migrating in 3-D collagen lattices. Cells were pretreatedwith blocking anti-b1 mAb 4B4 (10 µg/ml) or PBS prior to theincorporation into the collagen, and additional antibody was added tothe supernatant after matrix polymerization. Data were obtained bytime-lapse videomicroscopy and computer-assisted cell tracking and

represent population data based on steady state percentage of migrat-ing cells and migration velocity of locomoting cells from step to stepwithin the first 3 hours (T cells, dendritic cells) or 9 hours (MV3 cells)(n . 120 randomly selected cells from at least 3 independent experi-ments). Significance levels, ** P , 0.01, (*) P 5 0.09 (marginallysignificant, Student8s t-test for independent means).


collagen fibers rather than leading to structural matrixremodeling.

As a special event in cell migration, large multicellu-lar aggregates, sheets, and clusters of cells may developin the process of morphogenesis, angiogenesis, or tumorinvasion migrating in 3-D tissues (Friedl et al., 1995b,1997; Senger et al., 1997; reviewed in Friedl et al.,1998b). The migration of cell clusters is directionallypersistent and requires highly coordinated cell-celladhesion and communication for maintenance of clus-ter polarity as well as integrin-mediated interactions tocollagen fibers and associated matrix remodeling (Friedland Brocker, 1997; Tusch et al., 1998). Referring to theabove model (Fig. 6), migration of clustered cells shouldfollow integrin-dependent migration principles ofstrength and steadiness (Senger et al., 1997).

In conclusion, based on the proportions of morphologi-cal flexibility, adhesive strength, and biophysical ma-trix resistance, cell migration in 3-D collagen matricesresults from either slow integrin-dependent and reorga-nizing migration principles or fast and dynamic migra-tory action lacking focal adhesions and matrix remodel-ing (Fig. 6). Transitions from one strategy to the othercould result from changes in ligand characteristics anddensity (e.g., low to high fibronectin concentration inthe tissue upon inflammation), differential integrinengagement (i.e., by activating cytokines or anti-adhesion molecules), cellular maturation and differen-tiation (e.g., migratory monocytes differentiating intosessile macrophages), or the acquisition or loss ofintercellular junctions as present in migrating tumorcell clusters upon mAb treatment. The use of a certainmigration strategy may be dynamically regulated atthe cellular level as well as by changes in tissue

architecture, implicating varying migration strategiesin development, repair, and pathologic conditions.

ACKNOWLEDGMENTSThe authors are very grateful to Peter B. Noble, now

retired Head of the Department of Oral Biology, McGillUniversity of Montreal, Canada, whose support, encour-agement, and controversial yet highly stringent ideasstimulated these studies. We thank many collegues andgraduate students, in particular Frank Entschladen,Matthias Gunzer, Miriam Tusch, Eckhart Kampgen, C.Eberhard Klein, Kerstin Maaser, Bernd Niggemann,Gerd Nikolai, and Yael Hegerfeldt who contributed tothis project.

REFERENCESAbercrombie, M., Heaysman, J. E. M., and Pegrum, S. M. (1970) The

locomotion of fibroblasts in culture. I. Movements of the leadingedge. Exp. Cell Res., 59:393–398.

Aggeler, J., Frisch, S. M., and Werb, Z. (1984) Changes in cell shapecorrelate with collagenase gene expression in rabbit synovial fibro-blasts. J. Cell Biol., 98:1662–1671.

Akiyama, S. K., Olden, K., and Yamada, K. M. (1995) Fibronectin andintegrins in invasion and metastasis. Cancer Metast. Rev., 14:173–189.

Arencibia, I., and Sundqvist, K.-G. (1989) Collagen receptor on Tlymphocytes and the control of lymphocyte motility. Eur. J. Immu-nol., 19:929–934.

Birkedal-Hansen, H. (1995) Proteolytic remoddeling of extracellularmatrix. Curr. Opin. Cell Biol., 7:728–735.

Bischop, J. (1997) Cell adhesion and angiogenesis. J. Clin. Invest.,99:373–376.

Brooks, P. C., Stromblad, S., Sanders, L. C., von Schalscha, T. L.,Aimes, R. T., Stetler-Stevenson, W. G., Quigley, J. P. und Cheresh, D.A. (1996) Localization of matrix metalloproteinase MMP-2 to thesurface of invasive cells by induction with integrin avb3. Cell85:683–693.

Fig. 6. A model for different migration strategies in 3-D collagen lattices. Locomotor prototypes, asdescribed by a set of parameters, may undergo transition from adhesive to non-adhesive mechanisms andoverlapping characteristics may exist.


Burridge, K., and Chrzanowska-Wodnicka, M. (1996) Focal adhesions,contractility, and signaling. Annu. Rev. Cell Dev. Biol., 12:463–518.

Butcher, E. C. (1991) Leukocyte-endothelial cell recognition: Three (ormore) steps to specificity and diversity. Cell 67:1033–1036.

Choquet, D., Felsenfeldt, D. P., and Sheetz, M. P. (1997) Extracellularmatrix rigidity causes stregnthening of integrin-cytoskeleton link-ages. Cell, 88:39–48.

Cidadao,A. J. (1989) Interactions between fibronectin, glycosaminogly-cans and native collagen fibrils: an EM study in artificial three-dimensional extracellular matrices. Eur. J. Cell Biol., 48:303–312.

Cidadao, A. J., and David-Ferreira, J. F. (1985) A method for TEMvisualization of the extracellular matrix three-dimensional organiza-tion in tissues. J. Microsc. 142:49–62.

Davis, G. E., and Camarillo, C. W. (1995) Regulation of endothelial cellmorphogenesis by integrins, mechanical forces, and matrix guid-ance pathways. Exp. Cell Res. 216:113–123.

Davis, J. M., St. John, J., and Cheung, H. T. (1990) Haptotactic activityof fibronectin on lymphocyte migration in vitro. Cell. Immunol.129:67–79.

Del Pozo, M. A., Sanchez-Mateos, P., and Sanchez-Madrid, F. (1996)Cellular polarization induced by chemokines: a mechanism forleukocyte recruitment? Immunol. Today 17:127–131.

Dickinson, R. B., Guido, S., and Tranquillo, R. T. (1994) Biased cellmigration of fibroblasts exhibiting contact guidance in orientedcollagen gels. Annals Biomed. Eng., 22:342–356.

Dustin, M. L., Carpen, O., and Springer, T. A. (1992) Regulation oflocomotion and cell-cell contact area by the LFA-1 and ICAM-1adhesion receptors. J. Immunol. 148:2654–2663.

Elsedale, T. and J. Bard. 1972. Collagen substrate for the study of cellbehaviour. J. Cell Biol. 41:298–311.

Entschladen, F., Niggemann, B., Zanker, K. S., and Friedl, P. (1997)Differential requirement of protein tyrosine kinases and proteinkinase C in the regulation of T cell locomotion in three-dimensionalcollagen lattices. J. Immunol. 159:3203–3210.

Ezzell, R. M., Toner, K., Hendricks, J. C., Dunn, R., Tompkins, G.,andYarmush, M. L. (1993) Effect of collagen gel configuration on thecytoskeleton in cultured rat hepatocytes. Exp. Cell Res., 208:442–452.

Felsenfeld, D. P.,Choquet, D., and Sheetz, M. (1996) Ligand bindingregulates the directed movement of b1 integrins on fibroblasts.Nature, 383:438–440.

Friedl, P., and Brocker, E.-B. (1997) Cancer cell interactions with theextracellular matrix involved in tissue invasion and cancer cellmigration: Motility mechanisms beyond the single cell paradigm. In:Extracellular matrix and the groundregulation system. H. Heineand M. Rimpler, eds. Gustav-Fischer Verlag, Leipzig, Heidelberg,pp. 7–18.

Friedl, P, Noble, P. B., and Zanker, K. S. (1993) Lymphocyte locomotionin three-dimensional collagen gels: comparison of three quantitativemethods for analysing cell trajectories. J. Immunol. Methods,165:157–165.

Friedl, P., Noble, P. B., Shields, E. D., and Zanker, K. S. (1994)Locomotor phenotypes of unstimulated CD45RAhigh and CD45ROhigh

CD41 and CD81 lymphocytes in three-dimensional collagen lattices.Immunology, 82:617–624.

Friedl, P., Noble, P. B., and Zanker, K. S. (1995a) Lymphocytelocomotion in a three-dimensional collagen matrix: Expression andfunction of cell adhesion molecules. J. Immunol., 154:4973–4985.

Friedl, P., Noble, P. B., Walton, P. A., Laird, D. W., Chauvin, P. J.,Tabah, R. J., Black, M., and Zanker, K. S. (1995b) Migration ofcoordinated cell clusters in mesenchymal and epithelial cancerexplants in vitro. Cancer Res., 55: 4557–4560.

Friedl, P., Maaser, K., Klein, C. E., Niggemann, B., Krohne, G., andZanker, K. S. (1997) Migration of highly aggressive MV3 melanomacells in 3-D collagen lattices results in local matrix reorganizationand shedding of a2 and b1 integrins and CD44. Cancer Res.57:2061–2070.

Friedl, P., Entschladen, F., Conrad, C., Niggemann, B., and Zanker,K. S. (1998a) T lymphocytes migrating in 3-D collagen lattices lackfocal adhesions and utilize b1 integrin-independent strategies forpolarization, interaction with collagen fibers, and migration. Eur. J.Immunol. 28:2331–2343.

Friedl, P., Brocker, E.-B., and Zanker, K.S. (1998b) Integrins, cellmatrix interactions and cell migration strategies: Fundamentaldifferences in leukocytes and tumor cells. Cell Adhesion Commun.(in press).

Gershman, H. (1982) Three-dimensional models for the study ofinvasion and metastasis.In: Tumor Invasion and Metastasis. I.A.Liotta and I.R. Hart, eds. Martins Nijhoff Publ., Hague, pp. 231–250.

Gilat, D., Hershkoviz, R., Mekori, Y. A., Vlodavsky, I., and Lider, O.(1994) Regulation of adhesion of CD41 T lymphocytes to intact orheparinase-treated subendothelial extracellular matrix by diffus-ible or anchored RANTES and MIP-1b. J. Immunol., 153:4899–4906.

Guilford, W. H., Lantz, R. C. and Gore, R. W. (1995) Locomotive forcesproduced by single leukocytes in vivo and in vitro. Am. J. Physiol.,268:C1308–1312.

Gumbiner, B. M. (1996) Cell adhesion: The molecular basis of tissuearchitecture and morphogenesis. Cell, 84:345–357.

Gunzer, M., Kampgen, E., Brocker, E.-B., Zanker, K. S., and Friedl , P.(1997) Migration of dendritic cells in 3-D collagen lattices: Confocalreflection analysis of dynamic interactions with collagen fibers andreceptor distribution. Adv. Exp. Med. Biol., 417:97–104.

Haston, W. S., Shields, J. M., and Wilkinson, P. C. (1982) Lymphocytelocomotion and attachment on two-dimensional surfaces and three-dimensional matrices. J. Cell Biol., 92:747–752.

Hauzenberger, D., Klominek, J., Bergstrom, S. E., and Sundqvist,K. G. (1995) T lymphocyte migration: the influence of interactionsvia adhesion molecules, the T cell receptor, and cytokines. Crit. Rev.Immunol., 15:285–316.

Heath, J. P. and Reachey, L. D. (1989) Morphology of fibroblasts incollagen gels: a study using 400 keV electron microscopy andcomputer graphics. Cell Motil. Cytoskeleton, 14:382–392.

Heidmann, S. R. and Buxbaum, R. E. (1998). Cell crawling: First themotor, now the transmission. J. Cell Biol., 141:1–4.

Heino, J. (1996) Biology of tumor cell invasion: Interplay of celladhesion and matrix degradation. Int. J. Cancer, 65:717–722.

Hershkoviz, R., Miron, S., Cohen, I. R., Miller, A., and Lider, O. (1992)T lymphocyte adhesion to the fibronectin and laminin components ofthe extracellular matrix is regulated by the CD4 molecule. Eur. J.Immunol., 22:7–13.

Huttenlocher, A., Sandborg, R. R., and Horwitz, A. F. (1995) Adhesionin cell migration. Curr. Opin. Cell Biol., 7:697–706.

Hynes, R. O., and Lander, A. D. (1992) Contact and adhesive specifici-ties in the associations, migrations, and targeting of cells and axons.Cell, 68:303–322.

Jiang, Y., Zhu, J. F., Luscinskas, F. W., and Graves, D. T. (1994)MCP-1-stimulated monocyte attachment to laminin is mediated bybeta 2-integrins. Am. J. Physiol., 267:C1112–C1118.

Klein, C. E., D. Dressel, D., Steinmacher, T., Mauch, C., Eckes, B.,Krieg, T., Bankert, R., and Weber, L. (1991) Integrin a2b1 isupregulated in fibroblasts and highly aggressive melanoma cells inthree-dimensional collagen lattices and mediates the reorganizationof collagen I fibrils. J. Cell Biol., 115:1427–1436.

Langholz, O., Rockel, D., Mauch, C., Kozlowska, E., Bank, I., Krieg T.,and Eckes, B. (1995) Collagen and collagenase gene expression inthree-dimensional collagen lattices are differentially regulated bya1b1 and a2b1 integrins. J. Cell Biol.,131:1903–1915.

Lauffenburger, D. A. and Horwitz, A. F. (1996) Cell migration: Aphysically integrated progress. Cell, 84:359–369.

Lee, J., Ishihara, A., and Jacobson, K. (1993a) How do cells move alongsurfaces? TIBS 3:366–370.

Lee, J., Ishihara, A., Theriot, J. A., and Jacobson, K. (1993b) Principlesof locomotion for simple-shaped cells. Nature, 362:167–171.

Leppert, D., Waubant, E., Galardy, R., Bunnett, N. W., and Hauser,S. L. (1995) T cell gelatinases mediate basement membrane transmi-gration in vitro. J. Immunol., 154:4379–4389.

Lloyd, A. R., Oppenheim, J. J., Kelvin, D. J., and Taub, D. D. (1996)Chemokines regulate T cell adherence to recombinant adhesionmolecules and extracellular matrix proteins. J. Immunol., 156:932–938.

Maaser, K., Klein, C. E., Niggemann, B., Zanker, K. S., and Friedl, P.(1998) Functional hierarchy of simultaneously expressed adhesionreceptors: Integrin a2b1 but not CD44 mediates MV3 melanoma cellmigration and matrix reorganization within three-dimensional hyal-uronan containing collagen matrices (submitted).

Mandeville, J. T., Lawson, M. A., and Maxfield, F. R. (1997) Dynamicimaging of neutrophil migration in three dimensions: mechanicalinteractions between cells and matrix. J. Leuk. Biol. 61:188–200.

Negulescu, P. A., Krasieva, T. B., Khan, A., Kerschbaum, H. H., andCahalanM. D. (1996) Polarity of T cell shape, motility, and sensitiv-ity to antigen. Immunity, 4:421–430.

Niewohner, J.; Weber, I.; Maniak, M.; Muller-Taubenberger, A. andGerisch, G. (1997) Talin-null cells of Dictyostelium are stronglydefective in adhesion to particle and substrate surfaces and slightlyimpaired in cytokinesis. J. Cell Biol., 138:349–361.

Niggemann, B., Maaser, K., Lu, H., Kroczek, R., Zanker, K. S., andFriedl, P. (1997) Locomotory phenotypes of human tumor cell linesand T lymphocytes in a three-dimensional collagen lattice. CancerLett., 118:173–180.


Nikolai, G., Niggemann, B., Werner, W, Zanker, K. S., and Friedl, P.(1998) Cross-linking of CD2 and CD3 in human CD41 T lympho-cytes results in rapid cytoskeletal engagement and the induction ofmigration. Immunology 95:62–69.

Noble, P. B. (1987) Extracellular matrix and cell migration: locomotorycharacteristics of MOS-11 cells within a three-dimensional collagenlattice. J. Cell Sci., 81:241–248.

Noble, P. B., and Levine, M. D. (1986) Computer-assisted analyses ofcell locomotion and chemotaxis. CRC Press, Boca Raton.

Palecek, S. P., Loftus, J. C., Gisberg, M. H., Lauffenburger, D. A., andHorwitz, A. F. (1997) Integrin-ligand binding properties govern cellmigration speed through cell-substratum adhesiveness. Nature,385:537–540.

Pilcher, B. K., Dumin, J. A., Subeck, B. D., Krane, S. M., Welgus, H. G.,and Parks, W. C. (1997) The activity of collagenase I is required forkeratinocyte migration on a type I collagen matrix. J. Cell Biol.,137:1445–1457.

Rijkonen, T., Westermarck, J., Koivisto, L., Broberg, A., Kahari, V. M.,and Heino, J. (1995) Integrin a2b1 is a positive regulator ofcollagenase (MMP-1) and collagen a1(I) gene expression. J. Biol.Chem., 270:13548–13552.

Romanic, A. M., Graesser, D., Baron, J. L., Visintin, I., Janeway, C. A.,Jr., and Madri, J. A. (1997) T cell adhesion to endothelial cells andextracellular matrix is modulated upon transendothelial cell migra-tion. Lab. Invest., 76:11–23.

Schmidt, C. E., Horwitz, A. F., Lauffenburger, D. A., and Sheetz, M. P.(1993) Integrin-cytoskeletal interactions in migrating fibroblastsare dynamic, asymmetric, and regulated. J. Cell Biol., 123:977–991.

Schor, S. L., Allen, T, D., and Winn, B. 1983. Lymphocyte migrationinto three-dimensional collagen matrices: A quantitative study. J.Cell Biol., 96:1089–1096.

Schor, S. L., Schor, A. M., Allan, T. D., and Winn, B. 1985. Theinteraction of melanoma cells with fibroblasts and endothelial cells

in three-dimensional macromolecular matrices: A model for tumorcell invasion. Int. J. Cancer, 36:93–102.

Senger, D. R., Claffey, K. P., Benes, J. E., Perruzzi, C. A., Sergiou, A. P.,and Detmar, M. (1997) Angiogenesis promoted by vascular endothe-lial growth factor: regulation through alpha1beta1 and alpha2beta1integrins. Proc. Natl. Acad. Sci. U.S.A., 94:3612–13617.

Shyy, J. Y., and Chien, S. (1997) Role of integrins in cellular responsesto mechanical stress and adhesion. Curr. Opin. Cell Biol., 9:707–713.

Steinmann, R. M., Hoffman, L., and Pope, M. 1995. Maturation andmigration of cutaneous dendritic cells. J. Invest. Dermatol., 105:2S–7S.

Stewart, M. and Hogg, N. (1996) Regulation of leukocyte integrinfunction: Affinity vs. avidity. J. Cell. Biochem., 61:554–561.

Stossel, T. P. (1993) On the crawling of animal cells. Science, 260:1086–1094.

Thomas, L. A., and Yamada, K. M. (1992) Contact stimulation of cellmigration. J. Cell Sci., 103:1211–1214.

Turley, E. A., Erickson, C. A., and Tucker, R. P. 1985. The retention andultrastructural appearances of various extracellular matrix mol-ecules incorporated into three-dimensional hydrated collagen lat-tices. Dev. Biol., 109:347–369.

Tusch, M., Hegerfeldt, Y., Muradali, S., Brocker, E.-B., and Friedl, P.(1998) Invasion and migration of clustered cells from primary tumorexplants: differential expression and function of b1 integrins. J.Invest. Dermatol., 110:504 (Abstract Commun.).

Van der Merve, P. A., and Barclay, N. (1994) Transient intercellularadhesion: the importance of weak protein-protein interactions.TIBS, 19:354–358.

Yebra, M., Parry, G. C. N., Stromblad, S., Mackman, N., Rosenberg, S.,Mueller, B. M., and Cheresh, D. A. (1996) Requirement of receptor-bound urokinase-type plasminogen activator for integrin avb5-directed cell migration. J. Biol. Chem., 271:29393–29399.


cell migration strategies in 3-d extracellular matrix: differences in morphology, cell matrix...

Documents