Preventing the Activation or Cycling of the Rap1 GTPase Alters Adhesion and Cytoskeletal Dynamics and Blocks Metastatic Melanoma Cell Extravasation into the Lungs

Solid tumor metastasis is a multistep process wherein primary tumor cells become invasive, move through the stroma, penetrate blood or lymphatic vessels, enter the circulation, and traffic to distant sites. To establish secondary tumors at these sites, malignant cells must arrest within and exit from the microvasculature, invade the stroma of the new organ, and proliferate (1–5). Individual steps in this metastatic cascade can proceed by different mechanisms. For example, tumor cell invasion can occur via a “mesenchymal” mode or an “amoeboid” mode (3), distinct types of cell motility that are driven by mutually antagonistic signaling pathways (6). Thus, the design of antimetastatic therapies targeting a particular signaling pathway must take into account both tumor cell–specific and phenotype-specific contexts.

This is especially relevant for the Rap1 GTPase, a central regulator of cell adhesion and motility (7). Extracellular matrix (ECM) ligand binding by integrins leads to increased levels of the active GTP-bound form of Rap1, which, in turn, increases the affinity of integrins for their ligands and promotes the assembly of focal adhesion complexes (7, 8). Additionally, an initial asymmetrical distribution of activated Rap1 promotes cell polarity and migration by remodeling the actin cytoskeleton at the leading edge of the cell (9–13). Thus, Rap1 likely regulates multiple steps in the metastatic cascade. Indeed, altered expression of Rap1, the exchange factors that activate Rap1, or the GTPase-activating proteins (GAP) that convert Rap1 to the inactive GDP-bound state, has been observed in several tumor types, including melanoma (14–19). However, how altered Rap1 activation affects specific aspects of the metastatic cascade is cell type–dependent. For example, activated Rap1 promotes the metastatic invasion of breast, pancreatic, and prostate carcinoma cells but inhibits invasion by osteosarcoma and squamous cell carcinoma cells (14–17, 20, 21).

We now show that Rap1 may be a potential target for impeding the metastatic progression of tumors that exit the vasculature via transendothelial migration (TEM). Modeling the interactions between tumor cells and vascular endothelial cells in vitro showed that both blocking Rap1 activation and enforcing constitutive Rap1 activation altered tumor cell adhesion dynamics and decreased their ability to migrate across endothelial cell monolayers. Importantly, the ability of metastatic B16F1 murine melanoma cells (22, 23) to exit the lung microvasculature and form secondary lesions in vivo was greatly reduced when Rap1 activation was modulated in these ways.

B16F1, A375, MDA-MB-231 [American Type Culture Collection (ATCC)], K1735M1 (Robert Nabi, University of British Columbia, Vancouver, British Columbia, Canada), and bEND.3 cells (K. McNagny, University of British Columbia) were grown in DMEM with 8% FCS. Human umbilical vascular endothelial cells (HUVEC; ATCC) were grown in MCDB131 medium with 20% FCS, 16 units/mL heparin, and 20 μg/mL endothelial cell growth supplement (Sigma-Aldrich).

Antibodies used were FLAG (Sigma-Aldrich), Rap1 (Santa Cruz Biotechnology), HUTS4 (BD Pharmingen), paxillin (BD Biosciences), phospho-Y397 focal adhesion kinase (pFAK; Invitrogen), and AlexaFluor555–anti-PECAM-1 (Research Diagnostics). Human fibronectin, laminin, and tumor necrosis factor-α (TNF-α) were from Sigma-Aldrich. Fluorescently labeled phalloidin, AlexaFluor568-dextran, and cell-labeling dyes were from Invitrogen. The pIRM21-IRES-dsFP593 vector and derivatives encoding FLAG-Rap1V12 or FLAG-Rap1GAPII were from Michiyuki Matsuda (Kyoto University, Kyoto, Japan). pEGFP-C1–β₁ integrin was from Valerie Weaver (University of California, San Francisco, CA). pEGFP-C1–β-actin and pEGFP-C1–paxillin were from Robert Nabi. Cellular Lights talin–green fluorescent protein (GFP) was from Invitrogen.

To generate stable transfectants, expression vectors plus pMSCVpuro (BD Clontech) were transduced using calcium phosphate and cells were selected with 2 μg/mL puromycin, followed by fluorescence-activated cell sorting for dsFP593 expression. Transient transfections were performed using Lipofectamine 2000 (Invitrogen).

Rap1-GTP was precipitated from cell lysates using a glutathione S-transferase–RalGDS fusion protein and detected by immunoblotting with Rap1 antibodies (24).

Cells on ECM-coated substrata were either fixed and stained with antibodies or imaged live in a chamber maintained at 37°C and 5% CO₂. Images were acquired using the UPlan Apochromat 60×/1.35 numerical aperture (NA) objective on an Olympus FV1000 confocal microscope, captured with Olympus FluoView v1.6 software, and analyzed using ImagePro (Media Cybernetics).

Cells were plated on coverslips coated with fibronectin and a lawn of 1-μm fluorescent microspheres (FluoSpheres, Invitrogen). After 16 hours, the maximum distance from the edge of the cell to the edge of the cleared area was determined using FluoView v1.6 software.

For subcutaneous tumor growth, 10⁵ cells in 0.1 mL DMEM were injected into the hind flanks of C57BL/6 mice. Tumors were analyzed after 14 days. For metastasis assays, 2 × 10⁵ cells in 0.1 mL DMEM were injected into the tail vein. After 21 days, the lungs were removed. Visible colonies were photographed and counted.

For competitive arrest experiments, tumor cell populations were differentially labeled with CellTracker Green or CellTrace Far Red dyes and equal numbers (5 × 10⁵) were coinjected into the tail veins of C57BL/6 mice. For extravasation assays, labeled cells were coinjected with 5 mg/kg AlexaFluor568-dextran. At the indicated times, lungs were removed, fixed in paraformaldehyde, embedded in OCT, sectioned, stained with 4′,6-diamidino-2-phenylindole (DAPI), and imaged using the UPlan Fluorite 20×/0.7 NA objective of an Olympus FV1000 confocal microscope.

HUVECs were cultured as monolayers on flow chamber slides (IBIDI) and labeled with CellTrace Calcein Red. Tumor cells were labeled with CellTracker Green, mixed with mouse blood, and perfused across the HUVEC monolayer at 2 dynes/cm². After 10 minutes, the chamber was flushed and the cells were imaged.

bEND.3 murine microvascular endothelial cells (25) were cultured as monolayers on chamber slides, activated with 50 ng/mL TNF-α, and overlaid with CellTracker Green–labeled tumor cells. After 1 to 24 hours, cells were fixed and stained with rhodamine-phalloidin. Z-stacks were generated by confocal microscopy.

Collagen I (2 mg/mL in PBS; BD Biosciences) was polymerized on Transwell filters (Falcon). The lower chamber contained DMEM with 55% FCS and 10 ng/mL transforming growth factor-β (TGF-β; R&D Systems), and 5 × 10⁵ CMFDA-labeled B16F1 cells were added to the top chamber. After 20 hours, filters were fixed with paraformaldehyde, embedded in OCT, and sectioned vertically and Z-axis images were acquired.

Tumor cells were added to monolayers of TNF-α–activated bEND.3 cells. After fixation, cells were immunostained with the HUTS4 antibody and imaged by confocal microscopy using the 60×/1.35 NA objective. Fluorescence recovery after photobleaching (FRAP) was performed as described (26, 27) on β₁ integrin–GFP–expressing cells. A circular region of interest at the endothelial–tumor cell interface was photobleached using a 405-nm laser (100% intensity, 0.1 s). Fluorescence recovery was imaged until the intensity reached a plateau. The fluorescence signal was normalized to the prebleach intensity, and single exponential fit curves of the data were generated using Prism 4 software (GraphPad).

Mice were anesthetized and the cremaster muscle was exteriorized as described (28). AlexaFluor555-conjugated anti-PECAM-1 (0.2 mg/mL) was injected into the femoral artery along with 5 × 10⁵ CMFDA dye-labeled B16F1 cells. The cremaster muscle microvasculature was imaged in real time using an Olympus BX51 upright spinning disc confocal microscope with a ×20/0.95 XLUM Plan Fl objective.

P values were calculated using a two-tailed independent two-sample t test.

Tumor cell motility on two-dimensional substrates requires the activation of ECM-binding integrins and the subsequent formation and turnover of integrin-containing focal adhesion complexes. The Rap1 GTPase is a central regulator of integrin activation, and when murine B16F1 melanoma cells were plated on two-dimensional fibronectin or laminin, the level of active GTP-bound Rap1 increased (Fig. 1A). This led us to hypothesize that Rap1 activation may regulate focal adhesion formation in tumor cells.

To test this, we modulated Rap1 activation in B16F1 cells and other tumor cell lines by expressing either the constitutively active Rap1V12 mutant protein or Rap1GAPII, a Rap-specific GAP that blocks Rap1 activation (Supplementary Fig. S1; refs. 28–31). Imaging the cell-ECM interface such that only adhesive clusters of immunostained proteins were detected showed that Rap1-GTP levels correlated with focal adhesion formation. In both B16F1 (Fig. 1B) and K1735M1 (Fig. 1C) murine melanoma cells, the total area per cell occupied by clustered β₁ integrins was increased by Rap1V12 expression and decreased by Rap1GAPII expression. Similarly, using the HUTS4 antibody (32) to detect the active conformation of human β₁ integrins in A375 human melanoma cells and MDA-MB-231 human breast carcinoma cells showed that Rap1-GTP levels correlated with the amount of active β₁ integrin clustered at the cell-ECM interface (Fig. 1C; Supplementary Fig. S2). The same was true for paxillin and talin, cytosolic proteins that are recruited to focal adhesion complexes, and for the phosphorylated form of FAK (pTAK), an indicator of focal adhesion signaling (Fig. 1B). Consistent with these effects on focal adhesion formation, Rap1V12 expression increased, and Rap1GAPII expression decreased, the adhesiveness of B16F1 cells to rigid substrata (Supplementary Fig. S3). Although Rap1V12 and Rap1GAPII expression had opposite effects on the number of focal adhesions formed, both disrupted cell polarization and focal adhesion dynamics (see below).

Cell motility requires the dynamic turnover of focal adhesions and the establishment of a polarized morphology, where integrin signaling drives a dynamic reorganization of the actin cytoskeleton at the leading edge of the cell. Real-time imaging of paxillin-GFP clustering indicated that vector control B16F1 cells plated on a rigid fibronectin ECM formed new adhesions at their anterior ends and disassembled existing adhesions at their posterior ends (Fig. 2A). In contrast, Rap1V12-expressing cells formed stable adhesions that were localized radially along the cell periphery in a nonpolarized fashion. Rap1GAPII-expressing B16F1 cells formed few focal adhesions, but those that did form were also very stable, with little evidence of a polarized arrangement. Thus, focal adhesion turnover required both Rap1 activation and Rap1 cycling between the active GTP-bound and inactive GDP-bound states.

Modulating Rap1 activation or cycling also prevented B16F1 cells from establishing a polarized morphology with a dynamic actin cytoskeleton. Vector control cells generated a polarized distribution of F-actin (Supplementary Fig. S4A), and real-time imaging of actin-GFP indicated a highly dynamic turnover of actin-GFP filaments at the leading edge (Fig. 2B). In contrast, Rap1V12-expressing cells spread radially and generated nonpolarized parallel arrays of stable actin filaments (Fig. 2B; Supplementary Fig. S4A and B). Rap1GAPII-expressing B16F1 cells formed thin peripheral actin filaments that were also stable and not polarized (Fig. 2B; Supplementary Fig. S4A and B). These effects on actin correlated with a failure to establish a polarized distribution of PIP₃ (Supplementary Fig. S4B), a lipid that recruits cytoskeletal regulatory proteins. The establishment of cytoskeletal polarity required both Rap1 activation and cycling not only in B16F1 cells but also in K1735M1, A375, and MDA-MB-231 cells (Supplementary Fig. S4C).

The loss of anterior-posterior polarity caused by modulating Rap1 activation, as well as the decreased adhesion and cytoskeletal dynamics, resulted in a significant decrease in cell motility. As assessed using a bead-clearing assay, the movement of B16F1 cells on rigid two-dimensional fibronectin substrata was greatly reduced when Rap1V12 was expressed and when Rap1 activation was blocked by expressing either Rap1GAPII or the dominant-negative Rap1N17 protein (Fig. 2C).

Because altering Rap1 activation had dramatic effects on the in vitro adhesion, polarity, and migration of B16F1 cells, we asked if it interfered with tumorigenesis and metastasis in vivo. Modulating Rap1 activation did not prevent B16F1 cells from forming subcutaneous primary tumors in mice, although Rap1GAPII expression did decrease tumor growth in this context (Fig. 3A). In contrast, both Rap1GAPII and Rap1V12 expression dramatically reduced the ability of B16F1 cells to form metastatic pulmonary lesions after tail vein injection (Fig. 3B).

To identify specific aspects of the metastatic cascade that require Rap1 activation and cycling, vector control B16F1 cells were coinjected into the tail vein with differentially labeled Rap1V12- or Rap1GAPII-expressing B16F1 cells. After 15 minutes, similar numbers of each coinjected cell population were present in the lungs (Fig. 4A). However, after 2 hours, vector control cells vastly outnumbered the Rap1V12- and Rap1GAPII-expressing cells that remained in the lungs (Fig. 4A). Although Rap1GAPII-expressing cells exhibited impaired β₁ integrin–mediated adhesion to endothelial cells under flow conditions in vitro (Fig. 4B), a decrease in adhesion to the vascular wall cannot fully account for the dramatic decrease in post-arrest retention in the lungs, as Rap1V12-expressing cells adhered very efficiently to endothelial cells in vitro (Fig. 4B). This suggested that active extravasation might be essential for sustained retention of B16F1 cells in the lung.

To assay extravasation, we injected fluorescently labeled B16F1 cells i.v. and determined their localization in the lungs relative to the capillary microvasculature. Four hours after injection, 45% of the vector control cells in the lung tissue had extravasated and no longer colocalized with the red dextran that was coinjected to mark the microvasculature (Fig. 4C). In contrast, far fewer Rap1V12- and Rap1GAPII-expressing cells moved out of the microvasculature and into the surrounding lung tissue (Fig. 4C). Rap1GAPII-expressing B16F1 cells also exhibited a decreased ability to invade pliable three-dimensional collagen gels in vitro (Fig. 4D). However, Rap1V12-expressing B16F1 cells did invade collagen gels as well as control cells (Fig. 4D), indicating that movement across the endothelial wall is more dependent on Rap1 cycling than the subsequent invasion of the stromal ECM.

In closed capillary beds such as those of the lungs, tumor cells actively extend membrane processes across the endothelial wall as they extravasate (33, 34). To assess the effect of modulating Rap1 activation on this form of TEM, we used confocal microscopy to monitor process formation and the subsequent movement of B16F1 tumor cells across microvascular endothelial cell monolayers in vitro. After being in contact with endothelial cells for 1 hour, vector control B16F1 cells spread on the top of the endothelial cells and extended F-actin–containing processes along them (Fig. 5A, left). After 16 hours, the majority of the vector cells had undergone TEM by migrating across the endothelial monolayer and reaching its bottom side (Fig. 5A and B). In contrast, Rap1V12-expressing cells spread radially atop the endothelial cell monolayers, did not extend membrane processes, and exhibited a greatly reduced frequency of TEM compared with control cells (Fig. 5A and B). Rap1GAPII-expressing B16F1 cells spread poorly on the endothelial monolayers, did not extend processes, and also exhibited a significant reduction in TEM (Fig. 5A and B). This effect was not limited to B16F1 cells, as altering Rap1 activation and cycling significantly reduced TEM in K1735M1, A375, and MDA-MB-231 cells (Fig. 5B).

TEM by tumor cells requires the dynamic assembly and disassembly of adhesive contacts with the endothelial cells (5). Because modulating Rap1 activation altered the adhesion dynamics of tumor cells on rigid two-dimensional substrata, we determined if this was also the case when tumor cells interact with endothelial cell monolayers. To visualize activated β₁ integrins only in the tumor cells, we seeded A375 and MDA-MB-231 human tumor cells on murine endothelial monolayers and stained them with the HUTS4 antibody that recognizes active human, but not murine, β₁ integrins.

In vector control cells, clusters of active β₁ integrins at the tumor cell–endothelial cell interface were localized in a polarized fashion, predominantly on one side of the cell (Fig. 5C). Rap1GAPII expression significantly decreased the ability of the A375 and MDA-MB-231 human tumor cells to form such adhesive clusters. Moreover, FRAP analyses indicated that the mobility of β₁ integrin–GFP molecules within the adhesions that did form in the Rap1GAPII-expressing cells was increased (Fig. 5D; Supplementary Fig. S5), likely reflecting a reduction in integrin activation and ligand engagement when Rap1 activation is blocked.

Although Rap1V12-expressing tumor cells formed as many active β₁ integrin–containing adhesions as vector control cells, these adhesions were, for the most part, organized radially in a nonpolarized pattern (Fig. 5C). Additionally, FRAP analyses indicated that the mobility of β₁ integrin–GFP molecules within the adhesions of Rap1V12-expressing cells was significantly decreased (Fig. 5D; Supplementary Fig. S6). This decreased rate of integrin exchange may indicate a maturation of the adhesions into stable cytoskeleton-associated complexes. Indeed, disrupting the actin cytoskeleton with latrunculin A greatly increased β₁ integrin mobility within the adhesions of the Rap1V12-expressing tumor cells (Supplementary Fig. S6). Thus, the reduced ability of Rap1V12-expressing tumor cells to move along, and migrate through, the endothelial monolayer may be due in part to the increased maturity and stability of their adhesive complexes compared with control cells.

To better understand how Rap1 modulation inhibits TEM and extravasation in vivo, we directly imaged the dynamic behavior of B16F1 tumor cells within capillary beds. Because we could not image live cells within the lungs, we injected fluorescently labeled B16F1 cells into the femoral artery and used intravital microscopy to image them within the superficial capillaries of the cremaster muscle, which were marked with fluorescently labeled PECAM-1 antibodies.

We found that the initial arrest of B16F1 cells within the cremaster capillaries seemed to be independent of active cell adhesion. Trypsinizing vector control cells before injection to remove cell surface molecules did not prevent their arrest in the capillaries (Fig. 6A, right). Moreover, Rap1GAPII-expressing B16F1 cells, which adhered poorly to endothelial cells in vitro (Fig. 4B), arrested in the capillaries just as efficiently as nontrypsinized vector controls and the highly adherent Rap1V12-expressing cells (Fig. 6A).

Although it was not technically feasible to image cremaster capillaries long enough to detect completed tumor cell extravasation, real-time imaging of fluorescently labeled or actin-GFP–expressing cells showed that the majority of vector control B16F1 cells that arrested within capillaries formed dynamic, actin-rich pseudopodial protrusions that crossed the endothelial cell wall and entered the interstitium (Fig. 6A–C). The formation of these protrusive pseudopods was significantly reduced in Rap1V12-expressing B16F1 cells and in Rap1GAPII-expressing B16F1 cells (Fig. 6C). Thus, both Rap1 activation and cycling contribute to the formation of dynamic membrane protrusions, which likely facilitate tumor cell TEM and extravasation during metastasis (Fig. 6D).

Our results show that modulating Rap1 activation and cycling has distinct effects on different steps in the metastatic cascade. Blocking Rap1 activation in B16F1 mouse melanoma cells resulted in decreased adhesion and migration, as observed previously in human melanoma and human thyroid carcinoma cells (19, 35). This correlated with impaired invasion into three-dimensional collagen gels in vitro, decreased subcutaneous tumor growth, and greatly reduced pulmonary metastasis. Conversely, expressing the constitutively active Rap1V12 protein in B16F1 cells increased adhesion but, surprisingly, also blocked metastatic colonization of the lungs.

The reduced ability of both Rap1GAPII- and Rap1V12-expressing B16F1 cells to form lung metastases seems to be mediated, at least in part, by an impaired ability to exit the lung microvasculature. Previous studies showed that B16F1 cells extend protrusions through the capillary wall and into the subendothelial matrix before extravasation (34). We found that modulating Rap1 activation prevented B16F1 cells from forming such protrusions in vivo. Moreover, as summarized in Fig. 6D, our in vitro analyses showed that Rap1 activation and cycling were both important for tumor cells that had attached to endothelial cells to form dynamic adhesions and generate actin-rich protrusions that may promote TEM.

Blocking Rap1 activation and expressing constitutively activating Rap1 altered tumor cell–endothelial cell interactions in distinct ways. In Rap1GAPII-expressing tumor cells, decreased TEM was associated with a reduced ability of the cells to form adhesion complexes and spread on the apical surface of the endothelial cells. In contrast, Rap1V12-expressing cells formed robust, stable adhesions but they did not extend polarized protrusions toward the junctions between the endothelial cells and underwent TEM very inefficiently. Although Rap1V12 expression also altered adhesion stability and reduced tumor cell migration on rigid two-dimensional ECM substrata, as recently noted (8), this is the first report that this occurs when tumor cells interact with vascular endothelial cells. The highly stable focal adhesions that form in Rap1V12-expressing cells may limit the ability to generate protrusive leading edges that drive tumor cell migration, both on endothelial cells and on rigid substrata.

In contrast to our findings with melanoma cells, expressing constitutively active Rap1 increases the metastasis of prostate carcinoma cells to bone (20). This may reflect differences in the microvasculature of the two target organs. Unlike the closed capillaries of the lungs, the bone marrow sinusoids are mostly open. Thus, prostate carcinoma cells could reach the osteoblastic stroma without actively migrating along, or sending protrusions through, endothelial walls. The increased metastasis of Rap1V12-expressing prostate carcinoma cells may therefore reflect increased invasiveness into the tissue, as opposed to increased TEM. Consistent with this, we found that Rap1V12-expressing B16F1 cells efficiently invaded compliant three-dimensional collagen gels, although they exhibited reduced migration on rigid two-dimensional substrata and reduced TEM across endothelial cell monolayers.

The ability of Rap1V12 expression to inhibit two-dimensional motility but support three-dimensional invasion may relate to the biophysical properties of the substrata (36). Cell attachment to rigid two-dimensional substrata or to the surface of firmly attached endothelial cells may increase cytoskeletal tension within the tumor cell to the point where it augments the ability of Rap1-GTP to stabilize adhesions and impede cell movement; more compliant three-dimensional matrices may not do so. This could be tested directly by altering the stiffness of the collagen matrix, as has been done with breast carcinoma cells (37).

Although several studies concluded that extravasation is not a rate-limiting step in metastatic progression (38–40), our findings suggest that the Rap1-dependent formation of membrane protrusions that promote active extravasation may be a functionally important aspect of the process. Previous attempts to prevent extravasation by targeting matrix metalloproteinases (41, 42), integrins (43), or tetraspanins (5) have been unsuccessful. Our findings suggest that Rap1 could be a novel target for limiting the dissemination of metastatic cells that exit the vasculature via active TEM.

No potential conflicts of interest were disclosed.

We thank Michiyuki Matsuda and Valerie Weaver for critical reagents, Robert Nabi for reagents and advice, and Marcia Graves for technical assistance and advice.

Grant Support: Cancer Research Society, Canada (M.R. Gold and C.D. Roskelley).

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 with 18 U.S.C. Section 1734 solely to indicate this fact.