Tracking retrograde flow in keratocytes: news from the front

Actin assembly at the leading edge of the cell is believed to drive protrusion, whereas membrane resistance and contractile forces result in retrograde flow of the assembled actin network away from the edge. Thus, cell motion and shape changes are expected to depend on the balance of actin assembly and retrograde flow. This idea, however, has been undermined by the reported absence of flow in one of the most spectacular models of cell locomotion, fish epidermal keratocytes. Here, we use enhanced phase contrast and fluorescent speckle microscopy and particle tracking to analyze the motion of the actin network in keratocyte lamellipodia. We have detected retrograde flow throughout the lamellipodium at velocities of 1-3 microm/min and analyzed its organization and relation to the cell motion during both unobstructed, persistent migration and events of cell collision. Freely moving cells exhibited a graded flow velocity increasing toward the sides of the lamellipodium. In colliding cells, the velocity decreased markedly at the site of collision, with striking alteration of flow in other lamellipodium regions. Our findings support the universality of the flow phenomenon and indicate that the maintenance of keratocyte shape during locomotion depends on the regulation of both retrograde flow and actin polymerization.     

Coupling of cytoskeleton functions for fibroblast locomotion

Using a chick cell phenotype specialised for locomotion with morphometric measurements made possible by modern instrumentation technology, we have reinvestigated motile functions in fibroblast locomotion. Quantitative analysis of rapid fluctuations in cell form and organelle distribution during locomotion showed many significant correlations between different parts of the cell despite much irregularity in individual displacements over the time scale of the order of one second. These broke down when external perturbations caused changes in shape or direction. Partial energy deprivation caused the cells to lose control of shape and organelle distribution even though forward protrusion continued unaffected. Cytoplasmic displacements shown by marker mitochondria correlated with adjacent fluctuations at the leading edge, and drug treatments which increased the amplitude of mitochondrial movements caused visible protrusions in projected positions at the leading edge. We conclude that fibroblast locomotion may be driven coordinately by a common set of motility mechanisms and that this coordination may be lost as a result of physical or pharmacological disturbance. Taking our evidence with results from other Laboratories, we propose the following cytoskeleton functions. (i) Protrusive activity, probably based on solation--gelation cycles of the actin based cytoskeleton and membrane recycling which provides cellular and membrane components for streaming through the cell body to the leading edge; this is Ca++ sensitive but relatively energy insensitive. (ii) Constraining activity on the cell membrane and on certain organelles to maintain shape and so facilitate directionality and the drawing along of the trailing body; this is Ca++ insensitive but relatively energy sensitive. (iii) Channeling function of microtubules to direct the flow towards multiple foci on the leading edge, and so determine cell polarity. Such a mechanism of locomotion for fibroblasts has many features consistent with evidence for other cell types, especially amoebae and leukocytes.     

Ultrastructural evidence for contacts between migrating L5222 rat leukemia cells and extracellular matrix components of the rat mesentery

The nature of interactions between cells migrating through tissues and their structural surroundings are largely unknown. We have therefore examined the ultrastructural relationship between L5222 rat leukemia cells, moving through the loose connective tissue of the mesentery, and components of the extracellular matrix (ECM). Ultrathin tissue sections, fixed in the presence of ruthenium hexammine trichloride (RHT), revealed the following: Constitutents of fibrillar and nonfibrillar elements of the ECM are in contact with the plasma membrane of L5222 cells. Linear nonfibrillar ECM elements contact the plasma membrane at point-like sites, often associated with root-like structures present within the submembraneous microfilament mesh. Aggregates of ECM material are connected to patch-like cell membrane sites, associated with a condensed, plate-like part of the microfilament mesh. Point-like and patch-like contacts are more numerous at the anterior part of polarized migrating L5222 cells than on the posterior end. In round resting leukemia cells they are evenly distributed around the cell periphery. We suggest that the ECM-cell membrane contacts represent tissue adhesion sites. We therefore hypothesize that in migrating cells a coordinate interaction occurs between the contact sites and the continuous microfilament meshwork which results in a simultaneous backward movement of ECM-membrane contacts on the cell body and in a net forward movement of the whole cell. Since Dembo et al. (1981) present a similar mechanism for in vitro locomotion of granulocytes, we assume that blood cell locomotion in vivo and in vitro depends on similar molecular mechanisms: force generation by the cell, transmembraneous linkage between cytoskeletal and ECM elements, and membrane fluidity. The major difference in blood cell locomotion through a three-dimensional tissue or on a plane substratum would then be given by the distribution of contact sites, occurring around the cell periphery or limited to the ventral cell surface, respectively.     

Electric field-directed fibroblast locomotion involves cell surface molecular reorganization and is calcium independent

Directional cellular locomotion is thought to involve localized intracellular calcium changes and the lateral transport of cell surface molecules. We have examined the roles of both calcium and cell surface glycoprotein redistribution in the directional migration of two murine fibroblastic cell lines, NIH 3T3 and SV101. These cell types exhibit persistent, cathode directed motility when exposed to direct current electric fields. Using time lapse phase contrast microscopy and image analysis, we have determined that electric field-directed locomotion in each cell type is a calcium independent process. Both exhibit cathode directed motility in the absence of extracellular calcium, and electric fields cause no detectable elevations or gradients of cytosolic free calcium. We find evidence suggesting that galvanotaxis in these cells involves the lateral redistribution of plasma membrane glycoproteins. Electric fields cause the lateral migration of plasma membrane concanavalin A receptors toward the cathode in both NIH 3T3 and SV101 fibroblasts. Exposure of directionally migrating cells to Con A inhibits the normal change of cell direction following a reversal of electric field polarity. Additionally, when cells are plated on Con A- coated substrata so that Con A receptors mediate cell-substratum adhesion, cathode-directed locomotion and a cathodal accumulation of Con A receptors are observed. Immunofluorescent labeling of the fibronectin receptor in NIH 3T3 fibroblasts suggests the recruitment of integrins from large clusters to form a more diffuse distribution toward the cathode in field-treated cells. Our results indicate that the mechanism of electric field directed locomotion in NIH 3T3 and SV101 fibroblasts involves the lateral redistribution of plasma membrane glycoproteins involved in cell-substratum adhesion.     

Protein kinase D-mediated anterograde membrane trafficking is required for fibroblast motility

BACKGROUND: Locomoting cells exhibit a constant retrograde flow of plasma membrane (PM) proteins from the leading edge lamellipodium backward, which when coupled to substrate adhesion, may drive forward cell movement. However, the intracellular source of these PM components and whether their continuous retrograde flow is required for cell motility is unknown.RESULTS: To test the hypothesis that the anterograde secretion pathway supplies PM components for retrograde flow that are required for lamellipodial activity and cell motility, we specifically inhibited transport of cargo from the trans-Golgi network (TGN) to the PM in Swiss 3T3 fibroblasts and monitored cell motility using time-lapse microscopy. TGN-to-PM trafficking was inhibited with a dominant-negative, kinase-dead (kd) mutant of protein kinase D1 (PKD) that specifically blocks budding of secretory vesicles from the TGN and does not affect other transport pathways. Inhibition of PKD on the TGN inhibited directed cell motility and retrograde flow of surface markers and filamentous actin, while inhibition of PKD elsewhere in the cell neither blocked anterograde membrane transport nor cell motile functions. Exogenous activation of Rac1 in PKD-kd-expressing cells restored lamellipodial dynamics independent of membrane traffic. However, lamellipodial activity was delocalized from a single leading edge, and directed cell motility was not fully recovered.CONCLUSIONS: These results indicate that PKD-mediated anterograde membrane traffic from the TGN to the PM is required for fibroblast locomotion and localized Rac1-dependent leading edge activity. We suggest that polarized secretion transmits cargo that directs localized signaling for persistent leading edge activity necessary for directional migration.     

Computational Estimates of Membrane Flow and Tension Gradient in Motile Cells

All parts of motile cells, including the plasma membrane, have to translocate in the direction of locomotion. Both directed intracellular membrane transport coupled with polarized endo- and exocytosis and fluid flow in the plane of the plasma membrane can contribute to this overall plasma membrane translocation. It remains unclear how strong a force is required to generate this flow. We numerically solve Stokes equations for the viscous membrane flow across a flat plasma membrane surface in the presence of transmembrane proteins attached to the cytoskeleton and find the membrane tension gradient associated with this flow. This gradient is sensitive to the size and density of the transmembrane proteins attached to the cytoskeleton and can become significant enough to slow down cell movement. We estimate the influence of intracellular membrane transport and actin growth and contraction on the tension gradient, and discuss possible ‘tank tread’ flow at ventral and dorsal surfaces.     

Sequential expression of cell surface C3bi receptors during neutrophil locomotion

Fluorescence microscopy has been used to study the cell surface distribution of the complement receptor for C3bi (CR3) on human neutrophils during locomotion. CR3 is an integral membrane protein that participates in cell attachment phenomena including chemotaxis. Fluorescein- and rhodamine-conjugated monoclonal IgG or Fab fragments were used to label CR3. We have previously shown that CR3 is uniformly distributed on unstimulated cells. During cell locomotion the fluorescent labels redistribute to the uropod and retraction fibers. To better understand the role of CR3 in chemotaxis, we have performed sequential two-color labeling experiments in conjunction with fluorescence microscopy. Double-labeling experiments were conducted by labeling adherent neutrophils with fluorescein-conjugated anti-CR3 followed by chemotaxis in a gradient of FMLP (10(-7) M). The cells were then labeled again with rhodamine-conjugated anti-CR3. The uropod and distal training filopodia were labeled with fluorescein, whereas the cell body and occasionally proximal filopodia near the uropod were labeled with rhodamine. When neutrophils were fixed and permeabilized prior to the second CR3 labeling, the second fluorescent label was localized to a granule-like compartment(s), often near the lamellipodium. The results suggest a flow of CR3 from intracellular granules----lamellipodia and cell body----uropod----trailing filopodia during chemotaxis.     

Relationship between plasma membrane mobility and substrate attachment in the crawling movement of spermatozoa from Caenorhabditis elegans

Caenorhabditis elegans sperm are nonflagellated cells that lack actin and myosin yet can form pseudopods to propel themselves over solid substrates. Surface-attached probes such as latex beads, lectins, and antimembrane protein monoclonal antibodies move rearward over the dorsal pseudopod surface of sessile cells. Using monoclonal antibodies against membrane proteins of C. elegans sperm to examine the role of localized membrane assembly and rearward flow in crawling movement, we determined that substrates prepared by coating glass with antimembrane protein antibodies, but not naked glass or other nonmembrane-binding proteins, promote sperm motility. Sperm locomotion is inhibited in a concentration-dependent fashion when cells are bathed with soluble antimembrane protein monoclonal antibodies but not with antimouse Ig antibodies or a monoclonal antibody against a sperm cytoplasmic protein. Our results suggest that C. elegans sperm crawl by gaining traction with substrate-attached ligands via their surface proteins and by using the motor that moves those proteins rearward on unattached cells to pull the entire cell forward. Continuous insertion of new proteins at the front of the cell and their subsequent adhesion to the substrate allows this process to continue.     

Dynamic modes of the cortical actomyosin gel during cell locomotion and division

Tight regulation of the contractility of the actomyosin cortex is essential for proper cell locomotion and division. Enhanced contractility leads, for example, to aberrations in the positioning of the mitotic spindle or to anomalous migration modes that allow tumor cells to escape anti-dissemination treatments. Spherical membrane protrusions called blebs occasionally appear during cell migration, cell division or apoptosis. We have shown that the cortex ruptures at sites where actomyosin cortical contractility is increased, leading to the formation of blebs. Here, we propose that bleb formation, which releases cortical tension, can be used as a reporter of cortical contractility. We go on to analyze the implications of spontaneous cortical contractile behaviors on cell locomotion and division and we particularly emphasize that variations in actomyosin contractility can account for a variety of migration modes.     

Motility of cultured fish epidermal cells in the presence and absence of direct current electric fields

The motile behavior and cytoskeletal structures of fish epidermal cells (keratocytes) in the presence and absence of direct current (DC) electric fields were examined. These cells spontaneously show highly directional locomotion in culture, migrating at rates of up to 1 micron/s. When DC electric fields between 0.5 and 15 V/cm are applied, single epidermal cells as well as cell clusters and cell sheets migrate towards the cathode. Cell clusters and sheets break apart into single migratory cells in the upper range of these field strengths. Cell shape and morphology are unaltered when the keratocytes are guided by an electric field. Neither the spontaneous locomotion nor the electrically guided motility were found to be microtubule dependent. 1 mM La3+, 10 mM Co2+, 50 microM verapamil, and 50 microM nitrendipine (calcium channel antagonists) reversibly inhibited lamellipod formation and cell locomotion in both spontaneously migrating and electrically guided cells. Ciba-Geigy Product 28392, which stimulates the opening of calcium channels, and is a competitive inhibitor of nitrendipine, has no effect on the locomotion of keratocytes. Cell motility was also unaffected by hyperpolarizing and depolarizing (low and high K+) media. It is argued that while a tissue cell may accommodate changes in resting membrane potential without becoming more or less motile, the cell may not be able to counterbalance the effects of depolarization and hyperpolarization simultaneously. In this context, a gradient of membrane potential, which is induced by an external DC electric field, will serve as a persistent stimulus for cell locomotion.     

Dynamic deformation of migratory efferent lymph-derived cells "trapped" in the inflammatory microcirculation

The cellular immune response depends on the delivery of lymphocytes from the lymph node to the peripheral site of antigenic challenge. During their passage through the inflammatory microcirculaton, the migratory cells can become transiently immobilized or "trapped" in small caliber vessels. In this report, we used intravital microscopy and temporal area mapping to define the dynamic deformation of efferent lymph-derived mononuclear cells trapped in the systemic inflammatory microcirculation. Mononuclear cells obtained from the efferent lymph draining the oxazolone-stimulated microcirculation were labeled with fluorescent dye and reinjected into the feeding arterial circulation. Intravital video microscopy observed thousands of cells passing through the microcirculation; 35 cells were "trapped" in the oxazolone-stimulated microcirculation. Temporal area maps of the trapped cells demonstrated dramatic slowing and deformation. The cells were trapped in the microcirculation for a median of 8.90 sec (range 5-17 sec) prior to returning to the flow stream. During this period, the cells showed sustained movement associated with both antegrade locomotion (mean cell velocity = 7.92 microm/sec; range 1.16-14.23 microm/sec) and dynamic elongation (median cell length = 73.8 microm; range 58-144 microm). In contrast, efferent lymph-derived cells passing unimpeded through the microcirculation demonstrated rapid velocity (median velocity = 216 microm/sec) and spherical geometry (median diameter = 14.6 microm). Further, the membrane surface area of the "trapped" cells, calculated based on digital image morphometry and corrosion cast scanning electron microscopy, suggested that the fractional excess membrane of the cells in the efferent lymph was significantly greater than previous estimates of membrane excess. These data indicate that transient immobilization of efferent lymph-derived mononuclear cells in the systemic inflammatory microcirculation is rare. When "trapping" does occur, the shape changes and sustained cell movement facilitated by excess cell membrane may contribute to the return of the "trapped cells" into the flow stream.     

Quantitative morphodynamics of endothelial cells within confluent cultures in response to fluid shear stress

To evaluate shear stress-induced effects on cultured cells we have extended the mechanical setup of a multichannel in vitro rheological system and developed software allowing entire processing control and image data analysis. The values of cell motility, degree of orientation (alignment), and cell elongation were correlated as a function of time (morphodynamics). Collective and individual endothelial cells within confluent cultures displayed a shear stress-dependent characteristic phase behavior of the following time course: resting conditions (phase I), change of motility (phase II), onset of alignment (phase III), and finally cell elongation (phase IV). Especially cell motility was characterized by a randomized zigzag movement around mean trajectories (fluctuations) together with mean cell locomotion. Onset of shear stress caused a down-regulation of fluctuations of 30% within <10 min and simultaneously increased locomotion velocities preferring the flow direction (phase II). After a lag period of 10 to 20 min cells orientated in the direction of flow (phase III) without significant cell elongation, which finally occurs within hours (phase IV). These data provide first evidence that cells within confluent endothelial monolayers respond to shear stress with a characteristic phase behavior.     

MICROFILAMENTS AND CELL LOCOMOTION

The role of microfilaments in generating cell locomotion has been investigated in glial cells migrating in vitro. Such cells are found to contain two types of microfilament systems: First, a sheath of 50–70-A in diameter filaments is present in the cytoplasm at the base of the cells, just inside the plasma membrane, and in cell processes. Second, a network of 50-A in diameter filaments is found just beneath the plasma membrane at the leading edge (undulating membrane locomotory organelle) and along the sides of the cell. The drug, cytochalasin B, causes a rapid cessation of migration and a disruption of the microfilament network. Other organelles, including the microfilament sheath and microtubules, are unaltered by the drug, and protein synthesis is not inhibited. Removal of cytochalasin results in complete recovery of migratory capabilities, even in the absence of virtually all protein synthesis. Colchicine, at levels sufficient to disrupt all microtubules, has no effect on undulating membrane activity, on net cell movement, or on microfilament integrity. The microfilament network is, therefore, indispensable for locomotion.     

Fluorescence microscopy study of polymorphonuclear leukocyte substrate attached materials

We describe a technique to visualize substrate-attached materials (SAM) of polymorphonuclear leukocytes (PMN) using the fluorescent lipid analog 1,1'-dioctadecyl-3,3,3',3',-tetramethylindocarbocyanine-perchlorate (DiC18Icc). DiC18Icc was incorporated into the membranes of living cells or SAMs. Since cell preparation does not require fixation, SAMs can be rapidly visualized by fluorescence microscopy. SAMs are generated by subjecting attached cells to a shearing force by rinsing with phosphate-buffered saline (PBS). The SAM-labeling protocol identified a membrane compartment as shown by detergent extraction. The SAMs of PMN leukocytes observed with this technique display complex patterns of interconnecting filaments, foci with radiating filaments, and smooth membranous areas with interconnecting filaments. The sensitivity and nondestructive nature of the DiC18Icc-labeling procedure have allowed us to observe filopodia of motile cells. The results are consistent with the hypothesis that locomotion involves a series of attachment and detachment steps. After 60 minutes of locomotion, these trailing filopodia have been measured at lengths up to 100 micron. The amount of membrane associated with these filopodia accounts for roughly 10% of the total membrane area of resting cells. These data set limits for models of membrane flow during chemotaxis.     

Velocity distribution of the anterograde and retrograde transport of extracellular particles byAmoeba proteus

Summary The transverse velocity profiles of the anterograde flow of particles on the cell surface and around it are approximately parabolic. The peak velocity is recorded close to the membrane and the descendent arm of the profile is viscosity-dependent. It indicates that the extracellular forward flow is probably generated by a forward movement of the fluid fraction of the membrane itself. The retrograde component of extracellular movements is manifested by particles kept on the cell surface by adhesion, which behave exactly as the ectoplasmic layer on the opposite side of the membrane,i.e., they probably reflect the movement of that fraction of the surface material which is attached to the cortical microfilaments. In the longitudinal profile, the velocity of anterograde flow rises from the tail to the front of amoeba, but is generally related to the effective cell locomotion rate and not to the movements of any intracellular layer. Around the cells deprived of any attachment to the substratum, which cannot locomote but manifest vigorous intracellular movements, the anterograde flow ceases at least along 2/3 of their lenght. It persists, however, around the frontal fountain zone, where other particles still move backwards together with the retracted ectoplasmic layer. This indicates that the role of the forward flow of and on the cell surface is to compensate for: (1) the increase of the surface area in the frontal regions due to locomotion, (2) the withdrawal of a part of material which is hauled back by the retracting cortical layer. A comprehensive scheme of the velocity distribution within the different layers of a moving amoeba and around it has been constructed on the basis of present and earlier data.Study supported by the Research Project CPBP 04.01 of the Polish Academy of Science.I dedicate this paper to Professor K. E. Wohlfarth-Bottermann with the best wishes for his 65th birthday.     

Biophysical integration of effects of epidermal growth factor and fibronectin on fibroblast migration

Cell migration is regulated simultaneously by growth factors and extracellular matrix molecules. Although information is continually increasing regarding the relevant signaling pathways, there exists little understanding concerning how these pathways integrate to produce the biophysical processes that govern locomotion. Herein, we report the effects of epidermal growth factor (EGF) and fibronectin (Fn) on multiple facets of fibroblast motility: locomotion speed, membrane extension and retraction activity, and adhesion. A surprising finding is that EGF can either decrease or increase locomotion speed depending on the surface Fn concentration, despite EGF diminishing global cell adhesion at all Fn concentrations. At the same time, the effect of EGF on membrane activity varies from negative to positive to no-effect as Fn concentration and adhesion range from low to high. Taking these effects together, we find that EGF and Fn regulate fibroblast migration speed through integration of the processes of membrane extension, attachment, and detachment, with each of these processes being rate-limiting for locomotion in sequential regimes of increasing adhesivity. Thus, distinct biophysical processes are shown to integrate for overall cell migration responses to growth factor and extracellular matrix stimuli.     

Analysis of lateral redistribution of a plasma membrane glycoprotein- monoclonal antibody complex [corrected] [published erratum appears in J Cell Biol 1988 Apr;106(4):following 1403]

The lateral redistribution of a major murine glycoprotein, GP80, was studied on locomoting fibroblasts, using rhodamine-conjugated mAbs and ultralow light level digitized fluorescence microscopy. Confirming an earlier study (Jacobson, K., D. O'Dell, B. Holifield, T.L. Murphy, and J. T. August. 1984. J. Cell Biol. 99:1613-1623), the distribution of GP80 was coupled with cell locomotion; motile cells exhibited a gradated distribution of the GP80-mAb complex over the cell surface, increasing from the front to the rear, whereas stationary cells exhibited a nearly uniform GP80 distribution. By monitoring locomoting single cells, we found the gradated fluorescence distribution to be maintained as an approximate steady state. Newly extended leading edges were almost devoid of the fluorescence labeling. This was strikingly demonstrated in prechilled cells in which the extension of fluorescence-free leading edges caused a pronounced boundary between fluorescent and nonfluorescent zones. Subsequently this boundary eroded gradually in a manner consistent with diffusional relaxation. Evidence indicated that the GP80 redistribution was primarily caused by the lateral motion of GP80 in the plasma membrane and not via intracellular membrane traffic. Two cell locomotion models which, in principle, could account for the GP80 redistribution were tested: the retrograde lipid flow (RLF) model (Bretscher, M. S., 1984. Science (Wash. DC). 224:681-686) and an alternative hypothesis, the retraction-induced spreading (RIS) model. The predictions of these models were stimulated by computer and compared with experiment to assess which model was more appropriate. Whereas both models predicted steady-state gradients similar to the experimental result, only the RIS model predicted the lack of retrograde movement of the fluorescent boundary.     

Direct evidence that cancer cell locomotion contributes importantly to invasion

This study was undertaken to clarify whether active locomotion of cancer cells is important for their ability to invade. The most rapidly moving cells were isolated from a cultured murine parent fibrosarcoma by successive cycles of migration through a micropore membrane. Cells were isolated by unstimulated locomotion and by haptotaxis to laminin, and the selected cells did indeed constitute rapidly locomoting subpopulations. These cells invaded biological tissues more efficiently than did the unselected parent cells. The cells selected by haptotaxis to laminin invaded most rapidly through amnion with basement membranes (containing laminin). Cancer cell haptotaxis to laminin in basement membranes thus promotes penetration of these tissue barriers. These results show in a direct manner that cancer cell locomotion is in fact important in invasion of biological tissues.     

CELL SORTING IN THE PRESENCE OF CYTOCHALASIN B

The ability of cytochalasin B to inhibit ruffled membrane activity and cellular locomotion of vertebrate cells in monolayer culture prompted its use to study the necessity for this kind of active cellular locomotion in cell sorting in heterotypic cell aggregates. Cell sorting was inhibited in chick embryo heart-pigmented retina aggregates but a remarkable degree of sorting did occur in neural retina-pigmented retina aggregates. In these experiments, the levels of cytochalasin B employed (5 or 10 µg/ml) are sufficient to inhibit completely locomotion of these cell types in monolayer culture. It is proposed that the degree of cell movement achieved during sorting in neural retina-pigmented retina aggregates in the presence of cytochalasin B is the result of changes in cell contact resulting from adhesive interaction of cells. The effect of cytochalasin B on the initial aggregation of dissociated cells was also tested. With the cell types used in this study (chick embryo neural retina and limb bud), aggregation was not affected for a period of several hours.     

Competition amongst Eph receptors regulates contact inhibition of locomotion and invasiveness in prostate cancer cells

Metastatic cancer cells typically fail to halt migration on contact with non-cancer cells. This invasiveness is in contrast to normal mesenchymal cells that retract on contact with another cell. Why cancer cells are defective in contact inhibition of locomotion is not understood. Here, we analyse the dynamics of prostate cancer cell lines co-cultured with fibroblasts, and demonstrate that a combinatorial code of Eph receptor activation dictates whether cell migration will be contact inhibited. The unimpeded migration of metastatic PC-3 cells towards fibroblasts is dependent on activation of EphB3 and EphB4 by ephrin-B2, which we show activates Cdc42 and cell migration. Knockdown of EphB3 and EphB4 restores contact inhibition of locomotion to PC-3 cells. Conversely, homotypic collisions between two cancer cells results in contact inhibition of locomotion, mediated by EphA-Rho-Rho kinase (ROCK) signalling. Thus, the migration of cancer cells can switch from restrained to invasive, depending on the Eph-receptor profile of the cancer cell and the reciprocal ephrin ligands expressed by neighbouring cells.