Differing modes of tumour cell invasion have distinct requirements for Rho/ROCK signalling and extracellular proteolysis

Table 1. Correlation between requirement for Rho or ROCK function and cell morphology Full Table

Table 2. Depth of invasion of different tumour cell lines Full Table

While analysing these experiments, we observed that the different cell lines had strikingly different morphologies when moving through the Matrigel. To investigate this further, we plated cells in serum-free media on top of a thick (500−1000-m) layer of Matrigel. After 24−36 h, the tumour cells that had migrated into Matrigel were imaged using confocal microscopy and their 3D morphology was reconstructed. BE and SW962 cells were elongated, with numerous small F-actin-rich spikes (Fig. 1a). In contrast, LS174T and A375m2 cells were very rounded, with localized patches rich in F-actin and small membrane blebs (Fig. 1a, arrows). Occasionally, we observed several cells invading the Matrigel together as large rounded aggregates (data not shown). The rounded morphology of A375m2 and LS174T cells was also observed in cells on top of the Matrigel, but contrasted sharply with their more elongated morphology on a rigid 2D substrate coated with a thin layer (10−20 m) of Matrigel (see Supplementary Information, Fig. S1a). A thin layer of Matrigel is much more rigid than a thick layer because it is closely associated with the inflexible plastic. These data suggest that a pliable extracellular matrix favours a rounded tumour cell morphology. The formation of membrane blebs in A375m2 cells was not affected by treatment with 50 M zVAD-fmk, which prevents apoptosis by blocking caspase activity, demonstrating that the membrane blebs were not an indication of apoptosis (data not shown). In contrast to the other cell lines we studied, cultures of WM266.4 exhibited a mixed morphology. Approximately 60% of WM266.4 cells had a rounded morphology with occasional membrane blebs, whereas the rest had an elongated morphology with F-actin-rich protrusions (Fig. 1a). Significantly, both elongated and rounded WM266.4 could invade Matrigel and moved similar distances. However, after blockade of Rho or ROCK function, only elongated cells were able to invade (data not shown and see Fig. 5). The mixed morphologies observed in WM266.4 cells were not a result of the cell line being a mixed population, because single-cell clones derived from the parental WM266.4 cells also exhibited both morphologies (see Supplementary Information, Fig. S1b). Similar morphologies were observed in cells invading collagen I matrices and growth-factor-reduced Matrigel, indicating that the morphology of the cells observed is primarily the result of the 3D environment and not growth factors contained in matrix (data not shown). Indeed, our attempts to switch the morphology of these cells in Matrigel by using different growth factors (for example, epidermal growth factor (EGF), platelet-derived growth factor (PDGF) and basic fibroblast growth factor (bFGF)) were unsuccessful (data not shown). These results demonstrate a direct correlation between the cell lines whose invasion is blocked by inhibition of Rho or ROCK signalling and cells that move with a rounded bleb-associated morphology (Table 1).

Figure 1. Different morphologies of tumour cells in a 3D matrix in vitro and in vivo. (a) Tumour cells that had invaded a layer of Matrigel were imaged by confocal microscopy. Panels show a 3D reconstruction generated using volocity software from at least 20 confocal sections. Phalloidin staining (F-actin) is in red and cortactin staining in green. Scale bars represent 20 m. (b) GFP-expressing tumour cells grown as a tumour xenograft were imaged ex vivo by capturing at least 20 confocal sections. These were subsequently reconstructed in three dimensions using volocity software (tumour cells shown in green). In addition, A375m2 cells were stained for active caspase 3 (red). Yellow arrowhead indicates an apoptotic cell. White arrows show elongated protrusions. Scale bars represent 20 m in all panels. Inset panel shows only active caspase staining. Full Figure and legend (60K)

Figure 5. Inhibition of extracellular proteases and ROCK cooperate to inhibit tumour cell invasion. Images show a single phalloidin-stained confocal section of WM266.4 cells invading Matrigel and treated with either a cocktail of protease inhibitors and/or 10 M Y27632. Scale bars represent 20 m. Full Figure and legend (24K)

Next, we investigated the effect of modulating Rho or ROCK signalling on the morphology of A375m2 and BE cells in a 3D environment. These cell lines were chosen because they have striking rounded and elongated morphologies, respectively, and are easily transfected. Treatment of A375m2 cells with TAT-C3 or Y27632 resulted in a loss of bleb-like structures and in some cases promoted the formation of membrane protrusions (Fig. 2a, middle), suggesting that the rounded morphology and blebs are dependent on Rho or ROCK function. Similar results were obtained in LS174T cells (data not shown), although treatment of BE cells with either TAT-C3 or Y27632 did not affect the formation of elongated actin-rich protrusions (Fig. 2a, bottom). To investigate the effect of activating Rho or ROCK signalling we used parental non-metastatic A375 cells (A375p)¹¹. These cells have a less rounded morphology with fewer blebs than the metastatic A375m2 cells (Fig. 2a, top), indicating that there is a correlation between rounded cell morphology and metastatic potential. Overexpression of RhoC, which promotes the invasion of these cells in vitro and their ability to metastasize in vivo¹¹, or constitutively active RhoA^G14V promoted a rounded morphology in A375p cells (Fig. 2a, top). In addition, this was prevented by Y27632 treatment (Fig. 2a, top), suggesting that ROCK functions downstream of Rho. Consistently, expression of an active ROCKI mutant (ROCK-3) resulted in a rounded cell morphology with large numbers of membrane blebs (Fig. 2a, top). These data demonstrate that Rho−ROCK signalling is necessary for a round morphology with bleb-like structures and that activation of Rho−ROCK signalling promotes this morphology and antagonizes the formation of elongated protrusions. Blockade of Rac1 function through expression of the constitutively GDP-bound Rac mutant, Rac1^T17N, had little effect on the rounded morphology of A375m2 cells but inhibited the formation of elongated cell protrusions in BE and WM266.4 cells (Fig. 2a and data not shown). Indeed, the invasion of all three cell lines was reduced by expression of Rac1^T17N (data not shown).

Figure 2. Different modes of motility have differential requirements for RhoA and ROCK signalling. (a) Cells were transfected with EF-Rac1T17N, EF-RhoAG14V, EF-RhoC or pCAGG-ROCK-3, and EGFP-N1 to enable identification of transfected cells, or treated with TAT-C3 or Y27632, before plating on a thick layer of Matrigel. Panels show a single phalloidin-stained (F-actin) confocal microscopy section. Arrowheads identify transfected cells. Scale bars represent 20 m. (b) Levels of total and active GTP-bound RhoA was determined from 70% confluent cells growing in 10% donor calf serum. One representative experiment out of two is shown. (c) Cells were transfected with EGFP-N1 (to enable identification of transfected cells) and EF-RhoAG14V or pCAGG-ROCK-3 before plating on a thick layer of Matrigel. Panels show a single phalloidin-stained (F-actin) confocal microscopy section of transfected or treated cells. Arrowheads identify transfected cells. Scale bars represent 20 m. (d) Change in Matrigel invasion of BE and WM266.4 cells expressing RhoAG14V and ROCK-3 relative to EGFP-N1-transfected cells is shown. Cells invading 8 m or more were scored as positive. An average of three or four experiments was performed. Error bars represent standard deviation. Approximately 150 transfected cells were analysed for each data point. Full Figure and legend (81K)

As elongated protrusive movement does not require Rho or ROCK signalling, whereas rounded blebbing movement requires this pathway, we investigated whether these different requirements for Rho−ROCK signalling were reflected in the levels of RhoA activation. Pull-down assays that capture only the active GTP-bound form of RhoA¹¹ were used to compare RhoA activity between the two melanoma and two colon carcinoma cell lines with different modes of motility. We found that LS174T and A375m2 cells, which have a rounded morphology, have high levels of RhoA activity when compared with BE and WM266.4 cells, which exhibit either a completely or partially elongated morphology (Fig. 2b). We then tested if increasing the level of Rho-ROCK signalling in cells with elongated morphology would change their morphology and allow them to move in a rounded blebbing fashion. Transfection of BE or WM266.4 cells with either RhoA^G14V or ROCK-3 resulted in loss of the elongated polarized morphology (Fig. 2c and data not shown). The effect of overexpressing RhoA^G14V was blocked by Y27632 treatment, indicating that ROCK is required downstream of Rho to promote a rounded cell morphology. A similar proportion of cells transfected with RhoA^G14V or ROCK-3 were invasive, compared with control cells (Fig. 2c, d), although the depth to which cells transfected with ROCK-3 invaded was slightly reduced (data not shown). In contrast, activation of Rho signalling in BE cells in a 2D environment completely inhibits their motility¹⁶. These results demonstrate that cell-intrinsic factors, such as the activity of RhoA or ROCK, influence the mechanism of cell motility through a 3D matrix.

To visualize cells migrating in a 3D matrix, we performed time-lapse confocal microscopy on A375m2 cells expressing GFP. A375m2 cells moved without elongated protrusions but had many bleb-like structures that were very rapidly remodelled (Fig. 3a and see Supplementary Information, Movie 1). Not all the cells were motile in the time-lapse studies: this is probably caused by the lack of chemo-attractant gradient once the cells have entered the Matrigel and the relatively short time period of analysis. Time-lapse analysis over a longer time demonstrated that even during translation of cell movement, A375m2 cells maintain a rounded morphology (Fig. 3b and see Supplementary Information, Movie 2). During translation we estimated that the cell speed was 5 m h^-1; however, over longer periods of time the net speed was 2 m h^-1. The motility of A375m2 cells in a 3D matrix contrasted sharply with their behaviour on a 2D substrate; furthermore, their motility on a 2D substrate was not blocked by Y27632 treatment (see Supplementary Information, Movies 3 and 4).

Figure 3. Time-lapse analysis of rounded cell motility. (a) A375m2 cells expressing GFP and moving within Matrigel were imaged by capturing at least 20 confocal sections at the indicated times. These were subsequently reconstructed in three dimensions using volocity software. GFP fluorescence is shown. Scale bar represents approximately 20 m. An animated sequence of these data is shown in Supplementary Information, Movie 1. (b) A375m2 cells expressing GFP and moving within Matrigel were imaged by capturing a single confocal section at the indicated times. GFP fluorescence is shown. Scale bar represents 20 m. An animated sequence of these data is shown in Supplementary Information, Movie 2. Full Figure and legend (51K)

Despite lacking any obvious polarization, the motility of A375m2 required a gradient of serum growth factors, indicating that rounded motility is responsive to chemoattractants and is not merely random cell motility (data not shown). Therefore, we investigated if particular molecules and cellular structures were orientated in the direction of cell movement. Previous studies have shown that cells moving in an elongated 'mesenchymal' manner adopt a polarized morphology towards the chemoattractant. In particular, the Golgi apparatus and MTOC are orientated towards the front of the cell² and phosphatidylinositol-3,4,5-trisphosphate (PtdInsP₃) is produced in lamellipodia at the leading edge of cell^{17, 18}. Preliminary analysis showed that A375m2 cells begin to enter Matrigel after 6−12 h, whereas BE cells take 12−24 h (data not shown). Therefore, we analysed the distribution of these components 8 h after plating A375m2 cells on Matrigel and 16 h after plating BE cells. To visualize the Golgi apparatus, we used an antibody to the Golgi protein -coatomer¹⁹. This revealed that the Golgi apparatus was orientated in the direction of movement in BE cells, but not in A375m2 cells (Fig. 4a). The pleckstrin-homology (PH) domain of protein kinase B (PKB)/Akt, which binds PtdInsP₃, was fused to GFP and used to localize PtdInsP₃ within the cell. In agreement with previous results, we found that the PH domain of PKB localized to the ends of cell protrusions and sites of cell−cell contact in BE cells (Fig. 4a)¹⁷. In A375m2cells, however, it failed to show any specific localization at the cell surface, suggesting that PtdInsP₃ is not localized to a specific region of the plasma membrane (Fig. 4a). These results suggest that cells moving with a rounded morphology are not polarized in the direction of movement in the same way as cells with an elongated morphology.

Figure 4. Analysis of rounded cell morphology. (a) A375m2 or BE cells were fixed and stained as they began to enter Matrigel, approximately 8 h after plating. Images were generated by 3D reconstruction using volocity software from at least 20 confocal sections. Red staining is F-actin visualized with phalloidin. Green and grey-scale staining is detailed in the relevant panels. White upwards arrow indicates direction of cell invasion into Matrigel. Yellow arrowhead indicates bleb-like structure. Scale bars represent approximately 20 m. (b) The change in Matrigel invasion of A375m2 or BE cells expressing ezrin1−310 relative to EGFP-N1-transfected cells is shown. Cells invading 8 m or more were scored as positive. An average of four experiments was performed. Error bars represent standard deviation. Approximately 150 transfected cells were analysed for each data point. Full Figure and legend (67K)

RhoA−ROCK signalling promotes the formation of large focal complexes containing integrins, paxillin and vinculin², and the influences the distribution of ezrin in cells cultured in a 2D environment⁷. Therefore, we investigated the localization of these proteins in cells entering Matrigel. Integrin ₁ was localized to patches all over the surface of A375m2 cells (Fig. 4a). In addition, it was localized to bleb-like structures orientated towards the Matrigel in approximately 30% of invading cells (Fig. 4a, yellow arrowhead; cells are invading Matrigel in an upwards direction). Inhibition of Rho or ROCK signalling did not affect the localization of integrin ₁ to the cell surface in A375m2 cells; however, it was not found in blebs, as these were no longer formed. Integrin ₁ was distributed throughout the cell surface in BE cells (Fig. 4a). Interestingly, in A375m2 cells, the focal adhesion protein paxillin was more uniformly distributed and not specifically localized with integrin ₁ or in membrane blebs (Fig. 4a). The distribution of ezrin in A375m2 cells was highly asymmetric, with the majority concentrated in one or two discrete patches in 52% of control cells (Fig. 4a, yellow arrowhead) but in less than 10% of TAT-C3- or Y27632-treated cells. These patches of localized ezrin were orientated towards the Matrigel into which the cells were invading (Fig. 4a). In contrast, ezrin was localized around the main body of BE cells and not in the protrusions orientated in the direction of movement. We next tested if interfering with ezrin function affected the ability of A375m2 cells to invade Matrigel. Cells were transfected with an ezrin truncation construct (residues 1−310) that has previously been shown to interfere with the function of endogenous ERM proteins²⁰. A375m2 cells transfected with ezrin¹⁻³¹⁰ had a reduced ability to invade Matrigel when compared with control cells (Fig. 4B); however transfection of BE cells with ezrin¹⁻³¹⁰ had no effect. Interference with integrin ₁ function inhibited the motility of all cell lines tested (data not shown). These results demonstrate that in rounded cell motility, ezrin is localized in the direction of cell movement in a Rho- and ROCK-dependent manner and is required for cell movement.

Table 3. Effect of combining ROCK and extracellular protease inhibition on tumour cell invasion Full Table

	Top

In this study, we distinguish modes of cell motility through 3D matrices by their requirement for Rho−ROCK signalling. Rounded blebbing movement requires Rho−ROCK signalling, whereas elongated protrusive movement does not. Furthermore, the morphologies of cells growing in vivo in a tumour are consistent with these different modes of motility. Rounded tumour cell motility is associated with small bleb-like protrusions; these blebs are similar to those observed during apoptosis, indeed apoptotic membrane blebs are also dependent upon ROCK function¹³. We hypothesize that contractile force generated by Rho−ROCK signalling generates hydrostatic pressure, generating bleb-like protrusions in all directions. In contrast to elongated motility, the Golgi apparatus is not orientated in the direction of motility during rounded cell movement. Directionality of movement arises because contacts with the extracellular matrix are made more efficiently in a particular direction, possibly through localized concentrations of chemoattractants. These contacts may function as anchors to enable the translation of cell movement in this direction. In support of this model, we find molecules involved in connecting the cytoskeleton to the extracellular matrix, such as integrin ₁ and ezrin, are concentrated in the direction of movement. Ezrin may link cell−matrix contacts made to the actin cytoskeleton. The absence of vinculin and paxillin from membrane blebs suggests that the kinds of adhesive complexes formed are distinct from focal adhesions observed on 2D substrates²¹.

	Top

BE, LS174T, SW962 and WM266.4 were obtained from the American Type Culture Collection. A375P and A375m2 cells were a gift from R. Hynes (Howard Hughes Medical Institute, MIT). Cells were routinely maintained in DMEM containing 10% donor calf serum. Transfections were performed using Fugene 6 (1 815 091; Roche, Basel, Switzerland). Briefly, cells were seeded at 50% confluency in 6-well plates 16 h before transfection. Immediately before transfection, cells were transferred into OptiMEM (31985-047; Invitrogen, Paisley, UK). DNA (1 g) was mixed with 100 l OptiMEM and 6 l Fugene 6 15 min before addition of the mix to the cells for 6 h. Y27632 (Tocris, Avonmouth, UK) was used at 10 M; TAT-C3 (described in ref. 13) was used at 500 nM; GM6001 (#364205; Calbiochem, San Diego, CA) was used at 20 M; calpeptin (#PI-101; Plymouth Meeting, Biomol) was used at 10 M; aprotinin (#A3428; Sigma, Poole, UK) was used at 10 g ml^-1; leupeptin (#L2884; Sigma) was used at 10 g ml^-1. Elongation factor 1 (EF) promoter-driven-RhoA^G14V and EF-Rac1^T17N are described³¹; EF-RhoC is described³²; pCAGG-ROCK-3 is described³³; EGFP-N1 is from Clontech (Basingstoke, UK); EGFP-N1−ezrin¹⁻³¹⁰ is gift from R. Lamb (Institute of Cancer Research, London, UK) and EGFP−Akt-PH domain is a gift from J. Downward (Cancer Research UK, London)¹⁷. Pull-down assays to analyse RhoA activity were performed as described³⁴.

For qualitative analysis, a layer of Matrigel containing growth factors (1 ng ml^-1 EGF, 15 ng ml^-1 IGF-1, 12 pg ml^-1 PDGF and 2.3 ng ml^-1 transforming growth factor (#356230; BD Biosciences, Oxford, UK) was prepared on a normal tissue-culture dish. Routinely, the thickness of the layer was 500−1000 m. Cells were then plated on top of the Matrigel in serum-free DMEM. After 6−8 h or 24−36 h, cells were fixed in 4% formaldehyde/PBS and the cells that had invaded the Matrigel were analysed by indirect immunofluorescence microscopy using a Bio-Rad MRC1024 confocal microscope (Bio-Rad, Hercules, CA). For quantitative assays, 100 l of Matrigel lacking growth factors was prepared in a Transwell (#3422; Costar) with 8-m pores. After 24 h of transfection or TAT-C3 treatment, or after 2 h of Y27632 treatment, approximately 30,000 cells were seeded on the opposite side of the Transwell from the Matrigel. The cells were allowed to adhere for 4 h before filling the lower chamber of the Transwell dish (which contain the cells) with serum-free DMEM and the upper chamber with 10% DCS/DMEM, except for LS174T cells when 1%DCS/DMEM was used. Tris-buffered saline was used as a vehicle control for TAT-C3 treatment and Y27632 was re-added after approximately 36 h. After 3 days, the cells were fixed in 4% formaldehyde/PBS and stained with propidium iodide to label nucleic acids. By taking optical sections with a Bio-Rad MRC1024 confocal microscope, the proportion of cells that had invaded specific distances into the Matrigel was determined. For time-lapse microscopy, the cells were prepared in a similar manner to that used for qualitative analysis, except that MatTek (Ashland, MA) glass-bottomed dishes were used. These were then inverted and placed on a Tokai Hit heated stage for time-lapse confocal microscopy.

After fixation, cells were permeabilized with 0.2% Triton X-100/PBS, washed three times with PBS and incubated with Texas-Red−Phalloidin (#T17471, used at 1:100; Molecular Probes, Eugene, OR) and ToPro3 (#T3605; Molecular Probes) for 1 h before three further washes PBS and mounting. For cells or tumour samples that were stained with antibodies, the procedure was essentially as above, except that a primary antibody was used in place of phalloidin and a further step was added to allow subsequent incubation with secondary antibodies. Primary antibodies were as follows: Active caspase 3 (#9661S; Cell Signalling, Beverly, MA), ezrin (#610603; BD Bioscience), paxillin (#P13520; Transduction Labs), integrin ₁ (#sc-13590; Santa Cruz Biotechnology, Santa Cruz, CA), cortactin (UBC7901-80; Europa Bioproducts). Secondary antibodies were from Jackson Stratech.

	Top

1. Fidler, I.J. Critical determinants of cancer metastasis: rationale for therapy. Cancer Chemother. Pharmacol. 43, S3–S10 (1999). | Article | PubMed  | ISI | ChemPort |
2. Ridley, A.J. Rho GTPases and cell migration. J. Cell Sci. 114, 2713–2722 (2001). | PubMed  | ISI | ChemPort |
3. Ridley, A.J., Paterson, H.F., Johnston, C.L., Diekmann, D. & Hall, A. The small GTP-binding protein Rac regulates growth factor-induced membrane ruffling. Cell 70, 401–410 (1992). | PubMed  | ISI | ChemPort |
4. Amano, M. et al. Formation of actin stress fibers and focal adhesions enhanced by Rho-kinase. Science 275, 1308–1311 (1997). | Article | PubMed  | ISI | ChemPort |
5. Kimura, K. et al. Regulation of myosin phosphatase by Rho and Rho-associated kinase (Rho-kinase) Science 273, 245–248 (1996). | PubMed  | ISI | ChemPort |
6. Bretscher, A., Edwards, K. & Fehon, R.G. ERM proteins and merlin: integrators at the cell cortex. Nature Rev. Mol. Cell Biol. 3, 586–599 (2002). | Article | PubMed  | ISI | ChemPort |
7. Yonemura, S., Matsui, T., Tsukita, S. & Tsukita, S. Rho-dependent and -independent activation mechanisms of ezrin/radixin/moesin proteins: an essential role for polyphosphoinositides in vivo. J. Cell Sci. 115, 2569–2580 (2002). | PubMed  | ISI | ChemPort |
8. Suwa, H. et al. Overexpression of the rhoC gene correlates with progression of ductal adenocarcinoma of the pancreas. Br. J. Cancer 77, 147–152 (1998). | PubMed  | ISI | ChemPort |
9. van Golen, K.L. et al. A novel putative low-affinity insulin-like growth factor-binding protein, LIBC (lost in inflammatory breast cancer), and RhoC GTPase correlate with the inflammatory breast cancer phenotype. Clin. Cancer Res. 5, 2511–2519 (1999). | PubMed  | ISI | ChemPort |
10. Fritz, G., Just, I. & Kaina, B. Rho GTPases are over-expressed in human tumors. Int. J. Cancer 81, 682–687 (1999). | Article | PubMed  | ISI | ChemPort |
11. Clark, E.A., Golub, T.R., Lander, E.S. & Hynes, R.O. Genomic analysis of metastasis reveals an essential role for RhoC. Nature 406, 532–535 (2000). | Article | PubMed  | ISI | ChemPort |
12. Kleinman, H.K. et al. Isolation and characterization of type IV procollagen, laminin, and heparan sulfate proteoglycan from the EHS sarcoma. Biochemistry 21, 6188–6193 (1982). | PubMed  | ISI | ChemPort |
13. Coleman, M.L. et al. Membrane blebbing during apoptosis results from caspase-mediated activation of ROCK I. Nature Cell Biol. 3, 339–345 (2001). | Article | PubMed  | ISI | ChemPort |
14. Uehata, M. et al. Calcium sensitization of smooth muscle mediated by a Rho-associated protein kinase in hypertension. Nature 389, 990–994 (1997). | Article | PubMed  | ISI | ChemPort |
15. Decaudin, D. et al. Peripheral benzodiazepine receptor ligands reverse apoptosis resistance of cancer cells in vitro and in vivo. Cancer Res. 62, 1388–1393 (2002). | PubMed  | ISI | ChemPort |
16. Vial, E., Sahai, E. & Marshall, C.J. MAPK/Fra-1 inactivates 1 integrin signalling to Rho-GTP in tumour cells permitting Rac-dependent cell motility. Cancer Cell (in the press).
17. Watton, S.R. & Downward, J. Akt/PKB localisation and 3' phosphoinositide generation at sites of epithelial cell–matrix and cell–cell interaction. Curr. Biol. 9, 433–436 (1999). | Article | PubMed  | ISI | ChemPort |
18. Insall, R.H. & Weiner, O.D. PIP₃, PIP₂, and cell movement — similar messages, different meanings? Dev. Cell 1, 743–747 (2001). | PubMed  | ISI | ChemPort |
19. Duden, R., Griffiths, G., Frank, R., Argos, P. & Kreis, T.E. -COP, a 110 kd protein associated with non-clathrin-coated vesicles and the Golgi complex, shows homology to -adaptin. Cell 64, 649–665 (1991). | PubMed  | ISI | ChemPort |
20. Crepaldi, T., Gautreau, A., Comoglio, P.M., Louvard, D. & Arpin, M. Ezrin is an effector of hepatocyte growth factor-mediated migration and morphogenesis in epithelial cells. J. Cell Biol. 138, 423–434 (1997). | Article | PubMed  | ISI | ChemPort |
21. Cukierman, E., Pankov, R. & Yamada, K.M. Cell interactions with three-dimensional matrices. Curr. Opin. Cell Biol. 14, 633–639 (2002). | Article | PubMed  | ISI | ChemPort |
22. Wolf, K. et al. Compensation mechanism in tumor cell migration: mesenchymal–amoeboid transition after blocking of pericellular proteolysis. J. Cell Biol. 160, 267–277 (2003). | Article | PubMed  | ISI | ChemPort |
23. Thiery, J.P. Epithelial–mesenchymal transitions in tumour progression. Nature Rev. Cancer 2, 442–454 (2002). | Article | PubMed  | ISI | ChemPort |
24. Hauck, C.R., Hsia, D.A., Ilic, D. & Schlaepfer, D.D. v-Src SH3-enhanced interaction with focal adhesion kinase at 1 integrin-containing invadopodia promotes cell invasion. J. Biol. Chem. 277, 12487–12490 (2002). | Article | PubMed  | ISI | ChemPort |
25. Wyckoff, J.B., Jones, J.G., Condeelis, J.S. & Segall, J.E. A critical step in metastasis: in vivo analysis of intravasation at the primary tumor. Cancer Res. 60, 2504–2511 (2000). | PubMed  | ISI | ChemPort |
26. Wang, W. et al. Single cell behavior in metastatic primary mammary tumors correlated with gene expression patterns revealed by molecular profiling. Cancer Res. 62, 6278–6288 (2002). | PubMed  | ISI | ChemPort |
27. Itoh, K. et al. An essential part for Rho-associated kinase in the transcellular invasion of tumor cells. Nature Med. 5, 221–225 (1999). | Article | PubMed  | ISI | ChemPort |
28. Takamura, M. et al. Inhibition of intrahepatic metastasis of human hepatocellular carcinoma by Rho-associated protein kinase inhibitor Y-27632. Hepatology 33, 577–581 (2001). | Article | PubMed  | ISI | ChemPort |
29. Hidalgo, M. & Eckhardt, S.G. Development of matrix metalloproteinase inhibitors in cancer therapy. J. Natl Cancer Inst. 93, 178–193 (2001). | Article | PubMed  | ChemPort |
30. Somlyo, A.V. et al. Rho kinase and matrix metalloproteinase inhibitors cooperate to inhibit angiogenesis and growth of human prostate cancer xenotransplants. FASEB J. 17, 223–234 (2003). | Article | PubMed  | ISI | ChemPort |
31. Hill, C.S., Wynne, J. & Treisman, R. The Rho family GTPases RhoA, Rac1, and CDC42Hs regulate transcriptional activation by SRF. Cell 81, 1159–1170 (1995). | PubMed  | ISI | ChemPort |
32. Sahai, E. & Marshall, C.J. ROCK and Dia have opposing effects on adherens junctions downstream of Rho. Nature Cell Biol. 4, 408–415 (2002). | Article | PubMed  | ISI | ChemPort |
33. Ishizaki, T. et al. p160^ROCK, a Rho-associated coiled-coil forming protein kinase, works downstream of Rho and induces focal adhesions. FEBS Lett. 404, 118–124 (1997). | Article | PubMed  | ISI | ChemPort |
34. Sahai, E., Olson, M.F. & Marshall, C.J. Cross-talk between Ras and Rho signalling pathways in transformation favours proliferation and increased motility. EMBO J. 20, 755–766 (2001). | Article | PubMed  | ISI | ChemPort |

	Top

Competing interests statement:  The authors declare that they have no competing financial interests.