Introduction

Extracellular matrix (ECM) degradation by proteases such as matrix metalloproteinases (MMPs) contributes to tumour invasion and metastasis (Fearon and Vogelstein, 1990; Mignatti and Rifkin, 1993; Chambers and Matrisian, 1997; Goss et al., 1998; Sternlicht and Werb, 2001; Zucker and Vacirca, 2004). However, the precise role that MMPs play in the highly invasive subpopulations of tumour cells is currently under debate (Wagenaar-Miller et al., 2004; Zucker and Vacirca, 2004). Although pharmacological inhibitors of MMPs proved effective at suppressing progression of a number of tumours in animal models (Lozonschi et al., 1999; Brand et al., 2000; Adachi et al., 2001), clinical trials of MMP inhibitors in humans with gastric, pancreatic, prostate or lung cancer were disappointing (Shepherd et al., 2002; Pavlaki and Zucker, 2003; Wagenaar-Miller et al., 2004; Zucker and Vacirca, 2004).

A number of reasons have been put forward that may account for the lack of success with MMP inhibitors in preventing tumour invasion, including a possible lack of MMP inhibitor specificity. However, another contributory explanation may lie in recent studies that identified an MMP-independent mode of tumour invasion through three-dimensional matrices (Sahai and Marshall, 2003; Wolf et al., 2003). Sahai and Marshall (2003) showed that BE colon carcinoma cells and SW962 squamous carcinoma cells display an elongated morphology when cultured in the three-dimensional matrix, Matrigel, and this invasive mode is sensitive to MMP inhibitors. In contrast, A375m2 melanoma cells and LS174T colon carcinoma cells display a rounded morphology in Matrigel and invasion of these cells is independent of extracellular proteases, including MMPs (Sahai and Marshall, 2003). Wolf et al. (2003) also showed that tumour cells can exhibit plasticity, switching between such distinct modes of invasion. For example, HT1080 fibrosarcoma cells invade through fibrillar collagen gels with an elongated morphology characteristic of mesenchymal cells, and this type of invasion is associated with polarized clustering of 1 integrins and MMP-mediated breakdown of collagen fibres. This type of invasion, which is similar to that displayed by BE colon carcinoma cells and SW962 squamous carcinoma cells, is termed 'mesenchymal invasion'. However, when exposed to inhibitors of extracellular proteases, including MMP inhibitors, HT1080 cells adopt a rounded morphology and utilize an 'amoeboid invasion' strategy that is linked to loss of clustered integrins and associated ECM proteolysis. Under these conditions, the cells invade in a manner that is similar to A375m2 melanoma and LS174T colon carcinoma cells. Together, these studies have led to the proposal that some tumour cells can undergo a so-called 'mesenchymal to amoeboid' transition when treated with extracellular protease inhibitors (PIs), thus allowing MMP-independent invasion (Friedl and Wolf, 2003; Friedl, 2004).

Recent developments in live cell imaging techniques have confirmed that dynamic regulation of cell–substrate adhesion, remodelling of the actin cytoskeleton and cell–substrate de-adhesion all contribute to cell migration across planar two-dimensional substrates in vitro (Huttenlocher et al., 1995; Zamir and Geiger, 2001; Carragher and Frame, 2004; Webb et al., 2004). Cell adhesion to ECM substrates is primarily mediated by heterodimeric integrin receptors composed of and subunits (Hynes, 1992). The turnover of integrin-mediated adhesions, in particular release of cell–substrate contacts at the rear of cells, is a rate-limiting step during two-dimensional planar migration. Tyrosine kinase-mediated phosphorylation and proteolysis of focal adhesion components by the calpain family of intracellular proteases are key events contributing to adhesion turnover (Carragher et al., 1999; Krylyshkina et al., 2002; Carragher and Frame, 2004; Webb et al., 2004). Indeed, calpain-mediated cleavage of the focal adhesion component talin has recently been implicated in controlling the turnover of integrin-linked focal adhesions (Franco et al., 2004), whereas calpain 2 acts downstream of the epidermal growth factor receptor and Src kinase to promote calpain-mediated turnover of integrin-linked adhesions and cell motility (Glading et al., 2000, 2004; Carragher et al., 2003; Cuevas et al., 2003; Carragher and Frame, 2004). However, although now well recognized as an important component of two-dimensional planar migration, the role of integrin adhesion turnover has not been studied in tumour cell invasion through three-dimensional matrices.

Here, we have studied the calpain 2 dependency of mesenchymal and amoeboid phenotypes of a number of tumour cell lines, including HT1080 cells, which can be induced to switch between mesenchymal and amoeboid invasive phenotypes. We found that the mesenchymal, but not the amoeboid, invasion strategy displays calpain 2 dependency, reflecting a likely requirement for integrin adhesion turnover. This was clearly linked to integrin function in HT1080 cells, with lack of calpain 2 requirement correlating with weakened adhesion to collagen, consistently reduced cell surface 21 integrin expression and impaired focal adhesion kinase (FAK) autophosphorylation. Treatment of amoeboid-like cells with a ROCK inhibitor restored calpain 2 dependency, as well as normal surface 21 integrin expression and function that is more typical of mesenchymal-like cells. These data provide the first evidence that calpain 2 dependency, which is linked to integrin status, defines two distinct modes of tumour cell invasion. Consistent with a distinct requirement for mediators of integrin adhesion turnover, we found that the mesenchymal, but not the amoeboid, mode of invasion was sensitive to inhibitors of the Src family kinases, which normally cooperate with calpain to mediate focal adhesion turnover in cells migrating in two dimensions (Carragher et al., 2003; Carragher and Frame, 2004).

Results

Transition of HT1080 cells on fibrillar collagen to a rounded amoeboid-like phenotype after treatment with protease inhibitors

Initial studies describing the distinct mesenchymal and amoeboid modes of tumour invasion showed that HT1080 fibrosarcoma cells expressing MT1-MMP acquire an amoeboid-like invasive phenotype following treatment with a cocktail of inhibitors against extracellular proteases (Wolf et al., 2003). We examined HT1080 cells either within, or on top of, a fibrillar collagen gel, in the absence or presence of PIs, including inhibitors that target a wide range of extracellular protease classes including the broad MMP inhibitor GM6001, E64 (targets cysteine proteases such as cathepsins B, L, H and K), pepstatin A (inhibits aspartic proteases including cathepsin D), leupeptin (a broad inhibitor of cysteine proteases and cathepsin D) and aprotinin (an inhibitor of serine proteases, such as urokinase and tissue plasminogen activator). As shown by gelatin zymography, this PI mix suppresses the activity of endogenous MMP2 secreted by HT1080 cells cultured on fibrillar collagen (Figure 1a). Treatment of HT1080 cells cultured on top of (Figure 1b) or within (results not shown) a fibrillar collagen gel with the PI mix results in a more rounded morphology, similar to the amoeboid phenotype previously described (Wolf et al., 2003). We found that all of the inhibitors used in the cocktail (with the exception of pepstatin A) are required for amoeboid conversion, confirming that inhibition of multiple classes of extracellular proteolytic enzymes secreted by these HT1080 cells caused the so-called mesenchymal to amoeboid transition.

Figure 1.

Transition of HT1080 cells on fibrillar collagen to a rounded amoeboid phenotype following treatment with extracellular protease inhibitors (PIs). (a) Cells were cultured on top of fibrillar collagen in the absence or presence of the PI mix for 3 days. Conditioned media were collected and gelatin zymography used to determine the activity of secreted MMP2. Pro- and active forms of MMP2 from cell supernatants are indicated. (b) Cell morphology of HT1080 cells cultured on fibrillar collagen in the absence and presence of PI mix for 3 days was evaluated by phase contrast microscopy ( 400 magnification). (c) An in vitro assay was used to measure three-dimensional invasion through fibrillar collagen matrix. HT1080 cells cultured on fibrillar collagen in the absence or presence of PI mix were stained with calcein (which also acts as a marker of cell viability). After 3 days, a z-series of images from the top of the fibrillar collagen gel down 55 m into the collagen was captured using a confocal microscope. Invasion assays were performed in triplicate and cell invasion was quantified and presented as the percentage of total number of cells invading beyond 20 m through to 55 m. Invasion data represent mean values from three separate experiments standard deviation.

Full figure and legend (209K)

To determine whether HT1080 cells treated with the PI mix can invade through fibrillar collagen by the extracellular protease-independent amoeboid-like strategy, we quantitated HT1080 cell invasion through collagen. HT1080 cells were seeded on top of a fibrillar collagen gel in a transwell, in the absence or presence of the PI mix. Cells were seeded in medium containing 0.2% fetal calf serum (FCS), and medium containing 10% FCS was placed in the bottom well of the transwell as the chemotactic source. After 3 days, cells were stained with the fluorescent dye calcein AM, as previously described (Scott et al., 2004), and non-invading cells on top of the collagen gel, or cells that had invaded into the collagen gel at sequential 5 m distances, were imaged using confocal microscopy (Figure 1c). Fluorescence intensity values relating to the number of cells on top and at each 5 m layer within the collagen gel were quantified by computer software as described in Materials and methods. Data bars represent the total number of cells that have invaded beyond 20 m, and up to 55 m, into the fibrillar collagen (expressed as a percentage of the total number of cells adherent to the top and invaded within the collagen) (Figure 1c). This showed that treatment of HT1080 cells with the PI mix suppressed their invasion through fibrillar collagen (Figure 1c), although a significant number of cells still invaded, presumably by the protease-independent amoeboid-like mechanism (Figure 1c). The apparent decrease in invasion caused by the PI mix suggests that the amoeboid invasion of HT1080 cells in our experiments is less efficient, or less rapid, than mesenchymal invasion. This is confirmed by cell tracking experiments, in which HT1080 cells in the presence of PI mix move more slowly through collagen than control cells (see Figure 4).

Figure 4.

Computer-assisted tracking of three-dimensional invasion. HT1080 (a and b), LS174T (c) and BE cells (d) were prestained in suspension with the lipophillic fluorescent dye Cell-Tracker green CMFDA (Molecular Probe, Europe). Time-lapse microscopy of Cell-Tracker green-stained cells embedded within a three-dimensional fibrillar collagen gel (HT1080 cells) or Matrigel (LS174T and BE cells) was performed as described in Materials and methods. Generation of three-dimensional reconstructed time-lapse movies and subsequent tracking of total distance of cellular invasion was performed using Image ProPlus software (Media Cybernetics, UK). Tracking data from representative invading cells are shown. The Y axis represents accumulated distance cells have invaded (m). The X axis represents sequential time frames collected every 30 min. Each line represents an individual cell tracked within the three-dimensional matrix.

Full figure and legend (538K)

Distinct requirement for calpain 2 during mesenchymal and amoeboid tumour invasion

As we, and others, had previously established a need for the intracellular protease calpain in integrin adhesion turnover associated with two-dimensional planar cell migration (Carragher and Frame, 2004; Franco et al., 2004), we addressed whether calpain was required for the mesenchymal or amoeboid invasion type. Addition of the cell-permeable calpain inhibitors ALLN and PD150606 suppressed mesenchymal invasion of HT1080 cells through fibrillar collagen, but not amoeboid invasion (Figure 2a). To confirm the calpain requirement, and to specifically examine the role of the calpain 2 isoform that has been previously implicated in focal adhesion turnover, cell motility and tumour invasion (Carragher et al., 2001; Mamoune et al., 2003; Glading et al., 2004), we depleted calpain 2 in HT1080 cells by sequential rounds of small interfering RNA (siRNA) transfection (Figure 2b). Calpain 2 siRNA (C2siRNA) treatment impaired proteolysis of the focal adhesion component talin, as judged by prevention of appearance of the well-characterized 190 kDa N-terminal calpain cleavage fragment and a parallel increase in the level of native talin (Figure 2b), when compared to treatment with a similar siRNA sequence containing two missense mutations (C2mut) that does not deplete calpain 2 (Figure 2b). Consistent with the calpain inhibitor data, siRNA-mediated depletion of calpain 2 inhibited mesenchymal invasion of HT1080 cells through fibrillar collagen, but did not suppress, and actually promoted, amoeboid invasion in the presence of the PI mix (Figure 2c). The effect of C2siRNA on invasion was compared with cells expressing mutant siRNA (C2mut) and the results were expressed as fold increase, or decrease, in invasion. These results were reproduced using a second siRNA calpain 2 sequence (oligo 2). These findings directly implicate calpain activity in the mesenchymal, but not the amoeboid, invasion strategies utilized by HT1080 fibrosarcoma cells. We confirmed that residual cell migration in the presence of calpain inhibitors was by the amoeboid-like mechanism (not shown). Confocal microscopic images of mesenchymal and amoeboid (+PI mix) invasion of HT1080 cells both in the absence and presence of calpain inhibitor or siRNA suppression of calpain 2 are represented (Figure 2d).

Figure 2.

Distinct requirements for calpain 2 during mesenchymal- and amoeboid-like tumour invasion. (a) HT1080 cell invasion through fibrillar collagen by the mesenchymal and amoeboid tumour invasion mechanism (in the absence and presence of protease inhibitor (PI) mix, respectively) was monitored by confocal analysis in response to treatment with 10 M of the cell-permeable calpain inhibitors (ALLN) or 50 M PD150606. The effect on mesenchymal and amoeboid invasion of HT1080 cells was quantified essentially as described in Figure 1. Mean values are shown from three separate experiments standard deviation, and results are presented as fold increase over corresponding mesenchymal and ameoboid control invasion. (b) Calpain 2 was depleted from HT1080 cells by three sequential rounds of transient transfection of calpain 2 siRNA (C2siRNA) oligonucleotides. SDS–polyacrylamide gel electrophoresis and immnoblotting were used to monitor calpain 2, talin and actin protein levels in HT1080 cells transfected with two separate C2siRNA, or corresponding missense mutated siRNA (C2mut). (c) HT1080 cells subjected to sequential transfection of C2siRNA, or corresponding mutant RNAi, were placed into invasion assays under mesenchymal and amoeboid conditions (the latter by addition of PI mix) and analysed by confocal microscopy. Results are expressed as mean values from three separate experiments standard deviation of the mean values. Data are presented as fold change over control cells transfected with missense mutated siRNA oligonucleotide. (d) Montage of confocal z-series images representing mesenchymal and amoeboid invasion in the presence of calpain inhibitor (ALLN) and C2siRNA (as quantified in (a) and (c)) is shown.

Full figure and legend (223K)

Next, we investigated invasion of other cancer cell lines, such as the amoeboid-like LS174T cells, mesenchymal-like BE cells (Sahai and Marshall, 2003) and H1299 cells. The proportions of cells that are invasive for each cell type are shown (Figure 3a). We further demonstrated that invasion of amoeboid-like LS174T was relatively insensitive to the calpain inhibitors ALLN (as used previously by Carragher et al., 2001, 2003) and PD150606 (Wang et al., 1996; Cooke and Patel, 2005) when invading through Matrigel (Figure 3b). To show that this could be a more general phenomenon, we also identified other cancer cells that displayed calpain-insensitive invasion, for example, H1299 non-small-cell lung cancer cells (Figure 3c), which display a rounded amoeboid-like morphology while migrating (not shown). Suppression of the calpain-induced 190 kDa talin cleavage product by ALLN was confirmed in these cells at the concentration used (Figure 3c). In contrast, invasion of the mesenchymal-like BE cells (Sahai and Marshall, 2003) was substantially impaired by both ALLN and PD150606 (Figure 3d). These findings imply that the requirement for calpain was more generally associated with the mesenchymal type of invasion strategy, and not restricted to the plastic HT1080 cells.

Figure 3.

Other cancer cell lines display different sensitivity to calpain inhibition. (a) The proportion of LS174T, BE and H1299 cells invading beyond 40 M through Matrigel (LS174T and BE) or collagen (H1299) was monitored as described for Figure 1. (b) LS174T cell invasion through Matrigel (these cells do not invade through fibrillar collagen), (c) H1299 cell invasion through fibrillar collagen and (d) BE cell invasion through Matrigel were further monitored in response to the effects of 10 M ALLN on LS174T and BE cells, and 1 M ALLN on H1299 cells (as 10 M ALLN proved to be growth inhibitory to these) and 50 M PD150606 (solid bars)). The effect of a reduced ALLN concentration on talin proteolysis in H1299 cells is shown (c, top right panel). An actin blot (as loading control) is shown (c, bottom right panel).

Full figure and legend (128K)

To demonstrate that ALLN was not killing individual cells resulting in reduced cell numbers within the three-dimensional matrix, it was important to carry out computer tracking of individual viable cells invading through fibrillar collagen matrix. ALLN suppressed the rate of cell movement of individual HT1080 cells through collagen (Figure 4a), but not when the PI mix was present (Figure 4b). In addition, ALLN did not obviously affect invasion of amoeboid-like LS174T cells (Figure 4c), but did generally suppress the rate of mesenchymal BE cell movement through Matrigel (Figure 4d). These findings are consistent with a greater requirement for calpain during the mesenchymal type of invasion, and demonstrated that calpain inhibition differentially affects the rate of individual cell migration through three-dimensional matrices.

Conversion of HT1080 cells from mesenchymal to amoeboid invasion is accompanied by modulation of the 21 integrin collagen receptor

Current evidence suggests that suppression of integrin-mediated adhesion of tumour cells, induced by a variety of experimental approaches, may promote an amoeboid mode of invasion (Friedl, 2004). Therefore, we examined the dependence of amoeboid (in the presence of PI mix) and mesenchymal (in the absence of PI mix) invading HT1080 cells on the 21 integrin collagen receptor. We quantified amoeboid and mesenchymal invasion of HT1080 cells following treatment with a cyclic RGD integrin blocking peptide and an anti-21 integrin blocking antibody. Relative to cells that were preincubated with a control cyclic RAD peptide, the cyclic RGD peptide, which blocks the adhesive interactions of 51, V1, V3 and V5 integrins, had no substantial effect on either mesenchymal or amoeboid invasion through fibrillar collagen (Figure 5a). However, preincubation with a functional blocking anti-21 antibody more effectively suppressed mesenchymal-like invasion than amoeboid-like invasion, although that too displayed some limited dependence on 21 integrin (Figure 5a). In addition, flow cytometric analysis using a specific anti-21 antibody demonstrated reduced expression of the 21 heterodimer on the surface of amoeboid HT1080 cells (when compared to mesenchymal cells) (Figure 5b). Although not a large effect, the observed reduction in the presence of PI mix was consistent (see also Figure 6b; note the log scale on the x axis). In addition, SDS–polyacrylamide gel electrophoresis (SDS–PAGE) analysis demonstrates no substantial change in the total cellular levels of 2 and 1 integrin subunits. However, a higher molecular weight form of both 2 and 1 appears in amoeboid cells, suggesting the possibility that post-translational modification of these integrin subunits takes place during transition to the amoeboid phenotype (Figure 5b).

Figure 5.

21 integrin function is altered in amoeboid invading cells. (a) Mesenchymal (control) and amoeboid (+protease inhibitor (PI) mix) invading HT1080 cell invasion through 20–55 m fibrillar collagen was quantified by confocal analysis following pretreatment with integrin blocking cyclic RGD peptide or a specific 21 integrin blocking antibody. The influence that cyclic RGD peptide has on mesenchymal and amoeboid invasion is presented as fold increase over cells treated with a non-blocking cyclic RAD peptide control. The influence of the 21 integrin blocking antibody is presented as fold increase over cells treated with a non-blocking 21 antibody control. Results are expressed as mean values from three separate experiments standard deviation of the means. (b) Flow cytometry analysis was employed in conjunction with an anti-21 antibody to examine cell surface expression of the 21 receptor in HT1080 cells that had been cultured for 3 days in fibrillar collagen under mesenchymal (blue) or amoeboid conditions (red). SDS–polyacrylamide gel electrophoresis and immnoblotting were also used to monitor total protein levels of 2 and 1 integrin subunits. (c) HT1080 cells cultured on fibrillar collagen for 3 days under mesenchymal or amoeboid conditions were recovered from the fibrillar collagen by collagenase digestion, stained with a fluorescent dye (Cell-Tracker green) and plated on fibronectin- or monomeric collagen-coated dishes. After 30 min, cell adhesion to collagen- or fibronectin-coated substrates was quantified using a fluorescence plate reader. Results are expressed as the percentage of cells seeded that are adherent after 30 min. Data represent mean values from three separate experiments standard deviation. (d) The autophosphorylation status of focal adhesion kinase (FAK) in mesenchymal and amoeboid HT1080 cells (- and + PI mix, respectively) was monitored by immunoprecipitating FAK from protein lysates and immunoblotting with an antibody specific for phospho-FAK-Tyr-397. Focal adhesion kinase protein levels are also shown (lower panel).

Full figure and legend (122K)

Figure 6.

Rho/ROCK signalling contributes to the amoeboid phenotype by integrin modulation. (a) The morphology of HT1080 cells cultured on fibrillar collagen in the presence of the protease inhibitor (PI) mix minus and plus (+) Y27632 was monitored by phase contrast microscopy ( 400 magnification). (b) Flow cytometry analysis was used in conjunction with an anti-21 antibody to monitor cell surface expression in HT1080 cells that had been cultured for 3 days in fibrillar collagen under mesenchymal (blue) and amoeboid (red) conditions, and under amoeboid conditions in the presence of Y27632 (green). (c) Autophosphorylation status of focal adhesion kinase (FAK) in mesenchymal and amoeboid cells, and amoeboid cells treated with Y27632, was monitored by immunoprecipitating FAK from protein extracts and immunoblotting with an anti-phospho-FAK-Tyr-397 antibody. Total FAK levels are also shown (lower panel).

Full figure and legend (159K)

Together, these data suggest that amoeboid invasion is less dependent upon 21 integrin-mediated adhesive interactions and can proceed at lower levels of functional 21 integrin activity when compared to mesenchymal invasion. To assess integrin functionality, mesenchymal and amoeboid cells were recovered from fibrillar collagen following collagenase digestion and subjected to adhesion assays on collagen- and fibronectin-coated substrates. HT1080 cells that were recovered after culture on fibrillar collagen for 3 days in the presence of the PI mix (largely amoeboid cells) displayed reduced capacity to adhere to collagen, and to fibronectin, when compared to untreated mesenchymal-like cells (Figure 5c). These data imply that integrin modulation, as judged by reduced cell surface expression, impairs adhesion. As FAK autophosphorylation controls assembly of, and signalling from, integrin-linked focal adhesions (Reiske et al., 1999; Sieg et al., 1999), we examined phosphorylation of FAK-Tyr-397, the integrin-induced site of autophosphorylation, in both amoeboid- and mesenchymal-like invading cells. To ensure that cells maintained their mesenchymal or amoeboid phenotypes at the time of protein extraction, cells were lysed while still in the three-dimensional collagen matrix by using a low detergent lysis buffer that solubilizes cellular proteins, but not the fibrillar collagen (see Materials and methods). Although there was no difference in FAK protein expression, we consistently found reduced FAK-Tyr-397 phosphorylation when cells were treated with the PI mix (Figure 5d). This implies that post-translational modulation of the 21 integrin collagen receptor, resulting in impaired formation of functional integrin receptor 21 complexes at the cell surface, leads to both weakened adhesion and impaired signalling downstream. Moreover, such integrin modulation is associated with the mesenchymal to amoeboid transition of invading HT1080 cells.

Rho/ROCK signalling contributes to integrin modulation and amoeboid phenotype

Recent studies have implied that signalling through the RhoA/ROCK pathway may maintain the amoeboid invasive phenotype, at least in part, through modulation of the actin cytoskeleton (Sahai and Marshall, 2003). We next addressed whether RhoA/ROCK signalling promotes the amoeboid invasive phenotype by contributing to integrin modulation in amoeboid-like invading HT1080 cells. Treatment of the rounded amoeboid invasive cells on fibrillar collagen with the ROCK inhibitor Y27632 restored predominant cell morphology to a more elongated morphology more typical of mesenchymal HT1080 cells (Figure 6a, right panel). Inhibition of ROCK signalling by Y27632 prevented the electrophoretic mobility shift of 2 and 1 subunits (data not shown) and increased the surface expression of the 21 integrin heterodimer in amoeboid cells to levels that were indistinguishable from those in mesenchymal cells (Figure 6b). Furthermore, addition of Y27632 increased FAK-Tyr-397 phosphorylation in the presence of the PI mix, implying that signalling downstream of the integrin was at least partially restored by the ROCK inhibitor (Figure 6c). These studies demonstrate that RhoA/ROCK signalling is likely to contribute to the amoeboid transition of HT1080 cells, at least in part, by modulation of the 21 collagen receptor.

Inhibition of RhoA/ROCK signalling suppresses amoeboid invasion and switches their invasive phenotype to a calpain-dependent mechanism

RhoA/ROCK signalling is also required for amoeboid-like invasion of other cancer cells, but not by cells invading by the mesenchymal type of mechanism (Sahai and Marshall, 2003). Consistent with this, we found that the ROCK inhibitor Y27632 suppressed amoeboid-like, but not mesenchymal-like, invasion of HT1080 cells (Figure 7a). Interestingly, we also found that, in the presence of Y27632, treatment with calpain inhibitor (ALLN) (Figure 7b) and C2siRNA (Figure 7c) caused further suppression of HT1080 cell invasion, implying that the ROCK inhibitor promotes transition from a calpain 2-independent amoeboid-like strategy to a calpain 2-sensitive invasion mechanism more typical of mesenchymal invading HT1080 cells. Similarly, the Y27632 ROCK inhibitor causes LS174T cells, which normally use an amoeboid invasion strategy, to display invasion through Matrigel that is sensitive to the calpain inhibitor ALLN (Figure 7d). This implies that inhibition of amoeboid-like invasion by Y27632 generally restores calpain dependence, and this is linked to restoration of mesenchymal morphology and normal 21 integrin function in HT1080 cells that can switch between these two invasion types.

Figure 7.

Inhibition of RhoA/ROCK signalling suppresses amoeboid invasion and switches cells to a calpain-dependent invasion mechanism. (a) HT1080 cell invasion under mesenchymal and amoeboid conditions (in the absence and presence of protease inhibitor (PI) mix, respectively) was quantified by confocal analysis following treatment with the Rho kinase inhibitor (Y27632). Results are expressed as fold change over control mesenchymal and amoeboid invasion, and presented as mean values standard deviation. (b) The effect that small interfering (siRNA)-mediated depletion of calpain 2 has upon amoeboid invasion of HT1080 cells, in the absence and presence of Y27632, is shown. Results are expressed as mean values from three separate experimentsstandard deviation. Data are presented as fold change in invasion over control cells transfected with missense mutated siRNA oligonucleotide. (c) The effect of adding Y27632 to LS174T cells invading through Matrigel in the presence of 10 M ALLN is shown. Data points are significantly different (^***P<0.005) as determined by Student's t-test.

Full figure and legend (38K)

Inhibiton of Src kinases suppresses mesenchymal, but not amoeboid, invasion

If it is true that mesenchymal, but not amoeboid, invading cells require mediators of dynamic integrin adhesion turnover like calpain 2, then inhibitors of the Src family kinases, which act upstream of calpain 2 to promote focal adhesion turnover, should also differentially impair the two modes of invasion. To test this, we therefore made use of two pharmacological Src inhibitors, PP2 (Hanke et al., 1996) and AP23464, a novel and recently characterized selective Src inhibitor (Brunton et al., 2005). Like ALLN, 10 M PP2 suppressed mesenchymal invading HT1080 cells (in the absence of PI mix), but not the amoeboid invading HT1080 cells (in the presence of PI mix) (Figure 8a). In fact, like ALLN, PP2 seemed to cause some enhancement of invasion of amoeboid cells (Figure 8a). Similarly, invasion of BE mesenchymal cells, but not LS174T amoeboid cells, was inhibited by PP2 (Figure 8b). Finally, we used AP23464 to demonstrate inhibition of mesenchymal, but not amoeboid, invading HT1080 cells at a concentration that effectively blocks Src autophosphorylation on Tyr-416 (Figure 8c). These data further imply that the weakened integrin function associated with amoeboid invasion confers an invasion mechanism that does not require the normal regulators of integrin adhesion turnover.

Figure 8.

Src inhibitors differentially affect mesenchymal and amoeboid invasion. (a) HT1080 cell invasion under mesenchymal and amoeboid conditions (in the absence and presence of protease inhibitor (PI) mix, respectively) was quantified by confocal analysis following treatment with the Src inhibitor PP2 (10 M). For comparison, the effects of 10 M ALLN are included. Results are expressed as fold change over control mesenchymal and amoeboid invasion, and presented as mean values standard deviation. (b) The effect of PP2 on invasion of BE and LS174T cells through Matrigel is reflected in changes of invading cells as a percentage of the total cells. (c) The effect of 1 M AP23464 Src inhibitor on mesenchymal and amoeboid invading HT1080 cells is shown. The effect of 1 M AP23464 on Src autophosporylation (as judged by reactivity with an anti-phospho-Src-Tyr-416 antibody) is shown (upper right panel). Total Src is shown by the immunoblot with an anti-Src antibody (lower right panel). Data points are significantly different (^***P<0.005; ^**P<0.01) as determined by Student's t-test.

Full figure and legend (105K)

Discussion

Recent studies identifying a distinct mode of tumour invasion, which is not dependent upon extracellular protease activity, challenge traditional views that tumour invasion requires degradation of ECM barriers (Sahai and Marshall, 2003; Wolf et al., 2003). Understanding such distinct modes of tumour invasion may lead to the development of more effective anti-invasive therapies.

The activity of kinases such as FAK, Src, and extracellular regulated kinase/mitogen-activated protein kinase, as well as calpain 2 proteolytic activity, is generally required for efficient turnover of integrin-linked focal adhesions and cell migration across two-dimensional substrates (Palecek et al., 1998; Carragher et al., 2001; Dourdin et al., 2001; Webb et al., 2004). However, in the context of cancer cell invasion, the requirement for mediators of focal adhesion turnover in three-dimensional matrices is not known. Although integrin-linked adhesion complexes clearly also form in cells cultured within a three-dimensional matrix, these are known to be different in composition from those in two dimensions. In particular, these are probably structurally distinct from the large mature focal adhesions formed by cells cultured on a two-dimensional substrate (Cukierman et al., 2002; Yamada et al., 2003). Recent studies showed that integrin-linked adhesions form contacts with three-dimensional fibrillar collagen and also provide an intracellular link with actin microfilament bundles that generates traction required for cell movement in that context (Petroll and Ma, 2003). Here, we demonstrate that optimal mesenchymal tumour invasion through fibrillar collagen or Matrigel depends on the activity of calpain 2 and the Src family kinases, which normally act upstream of calpain 2 to cause integrin adhesion turnover. In contrast, amoeboid tumour invasion proceeds by mechanisms that are relatively insensitive to inhibition of calpain 2 or Src.

In support of a reduced requirement for integrin-mediated adhesion during amoeboid invasion, we found that blocking antibodies against 21 inhibits mesenchymal invasion, but only partially suppresses amoeboid invasion. The distinct requirement for integrin function during mesenchymal and amoeboid invasion of HT1080 cells is specific for the collagen receptor 21 integrin, as incubation with a blocking RGD peptide had little effect on either mesenchymal or amoeboid invasion through fibrillar collagen. Indeed, experimental disruption of integrin function, including selecting for low endogenous 1 expression, blocking integrin antibodies and knockout strategies all appear to promote an amoeboid invasion mechanism (Friedl, 2004). In our experiments, conversion of the plastic HT1080 cells to an amoeboid mode of invasion is associated with consistently reduced cell surface expression of 21 integrin and weakened adhesiveness to ECM components. Although the reason for the reduction in cell surface collagen receptor is unknown, it could be a result of post-translational modification and/or altered integrin recycling in three dimensions, as total levels of 21 are not altered.

Signalling downstream of 21, as judged by reduced autophosphorylation of FAK at Tyr-397, is also suppressed in amoeboid invading HT1080 cells. However, it is not clear whether reduced autophosphorylation of FAK is simply a consequence of suppressed integrin activity, or whether it may contribute to the impaired integrin-mediated adhesion of amoeboid invading cells. Nevertheless, we can conclude that reduced integrin cell surface expression, impaired adhesion and suppressed signalling downstream are visibly linked to a switch to the amoeboid type of invasion. This provides an explanation for why there is a reduced requirement for calpain 2 and Src family kinases, as the weaker 21 integrin adhesion and function associated with amoeboid invasion reduces the need for dynamic integrin adhesion turnover. In fact, we found that reduced calpain 2 expression or activity could be associated with increased invasion of amoeboid cells over and above untreated control cells (for example, see Figure 2). This type of migration-promoting effect of calpain inhibition has been seen before during neutrophil chemotaxis. In this case, calpain inhibition enhanced neutrophil spreading and migration speed, at least in part by activation of Cdc42 and Rac1 activity (Lokuta et al., 2003).

Our findings are consistent with previous suggestions that integrin-mediated adhesion plays a reduced role during amoeboid invasion. For example, polarized clustering of integrins is lost during the mesenchymal to amoeboid transition of HT1080 cells (Wolf et al., 2003), whereas invasion of T-cell lymphomas and small cell lung carcinomas in vivo (which exhibit many of the characteristics of amoeboid invasive tumour cells in vitro) express low levels of 1 and 3 integrin subunits (Falcioni et al., 1994; Jaspars et al., 1996; Rintoul and Sethi, 2002; Friedl and Wolf, 2003). Also, the motility of the amoeba dictyostelium, from which the term amoeboid invasion was coined, is characterized by low-affinity cell–matrix interactions that are independent of integrin function, as dictyostelium does not express integrins (Friedl and Wolf, 2003). Lymphocytes and neutrophils also migrate in two dimensions by an amoeboid type movement that is driven by cortical filamentous actin and weak cell–matrix interactions, at least as judged by a lack of mature integrin-linked focal adhesions (Friedl et al., 2001; Friedl, 2004). As mentioned previously, cell migration in two dimensions is generally suppressed by calpain inhibitors (Huttenlocher et al., 1997; Glading et al., 2000, 2004; Carragher et al., 2001), but calpain inhibitors actually promote the amoeboid-like motility of neutrophils (Lokuta et al., 2003). Thus, there is an inverse correlation between the requirement for calpain in cell migration/invasion and cell surface integrin activity.

RhoA and ROCK are associated with the pathogenesis and progression of several human tumours (reviewed by Malliri and Collard, 2003). In our experiments, there is mutually exclusive sensitivity to inhibitors of ROCK and calpain 2, suggesting that the combination of inhibiting acto-myosin-based contractility and integrin adhesion turnover may provide an effective strategy to inhibit cancer cell invasion, minimizing the likelihood of escape by cells that can switch between distinct invasion modes. In keeping with this, calpain inhibition promoted HT1080 cell invasion that was sensitive to a ROCK inhibitor, and treatment with the ROCK inhibitor restored cell surface integrin expression and FAK autophosphorylation and calpain dependence. This suggests that Rho/ROCK signalling contributes to amoeboid invasion by altering integrin function, promoting a relatively integrin-, calpain-, and extracellular protease-independent mode of tumour cell invasion. It is likely that this is via the influence of Rho/ROCK signalling on remodelling the actin cytoskeleton (Etienne-Manneville and Hall, 2002), which, in turn, can modulate the function of integrin-associated adhesion structures (Bobak et al., 1997; Brakebusch and Fassler, 2003; Zhai et al., 2003). Moreover, Rho/ROCK signalling may positively influence localized acto-myosin-based contractility that drives membrane blebbing and amoeboid-like migration through three-dimensional matrix.

Although the HT1080 fibrosarcoma cells are useful because they display plasticity, and can readily switch between mesenchymal and amoeboid invasion strategies, it was important to use other cancer cell lines to confirm the generality of our findings. The mesenchymal invading cells (including BE colon cancer cells) generally require calpain 2 and Src kinase activity, whereas amoeboid cells (including LS174T colon cancer cells and H1299 non-small-cell lung cancer cells) do not. In contrast, the amoeboid mode of invasion is generally dependent upon RhoA/ROCK signalling, whereas mesenchymal invasion is independent of RhoA/ROCK (Sahai and Marshall, 2003). Our data show that weakened integrin function, and a consequent reduction in requirement for the proteins that normally regulate dynamic integrin adhesion turnover, including calpain 2 and Src family kinases, is more generally associated with the amoeboid mode of cancer cell invasion (summarized in Figure 9).

Figure 9.

Characterization of distinct mesenchymal and amoeboid tumour invasion mechanisms. The mesenchymal and amoeboid modes of tumour invasion (which are dependent and independent of extracellular proteases, respectively) can be distinguished by their requirement for intracellular calpain 2 proteolytic activity and Src kinase activity, which normally regulate integrin adhesion turnover in cells migrating in two dimensions. The lack of requirement for regulators of focal adhesion turnover is linked to weakened 21 integrin function and signalling downstream. RhoA/ROCK signalling contributes to the amoeboid invasion phenotype by modulating integrin function and this can be monitored by the sensitivity to inhibitors of dynamic integrin adhesion turnover. For cells displaying plasticity, which can readily switch between mesenchymal and amoeboid invasion strategies, a combination of inhibitors would therefore be required to ensure optimal suppression of invasion.

Full figure and legend (147K)

As cancer cell invasion is one of the defining hallmarks of malignancy, inhibiting this remains a potentially exciting target for developing therapeutic strategies. Such strategies are most likely to be effective in the context of suppressing the development of invasive tumours or in preventing further spread of residual metastatic disease after surgery. Detailed molecular characterization of distinct modes of tumour cell invasion is therefore important, and may have predictive value in guiding the use of potentially anti-invasive or anti-metastatic therapies. Our studies imply that evaluation of integrin ECM-receptor status, and activities of proteins like calpain 2 and Src kinases, as well as ROCK, will be potentially useful clinical indicators of invasion mode, and may be relevant in ultimately tailoring the use of appropriate therapeutic combinations for individual patients. However, there remain major practical obstacles to designing rapid and informative clinical trials to test anti-invasion strategies. In the case of Src kinase inhibitors, some of which are currently in Phase l clinical trials, there is evidence from in vitro work that these may be more effective as anti-invasion agents than as growth inhibitory anti-tumour agents. It would therefore be exciting to test such agents in preclinical models of cancer invasion and metastasis, and to monitor the effects of inhibitory agents on individual tumour cells invading by mesenchymal or amoeboid strategies by real-time tumour cell imaging in vivo.

Materials and methods

Cell culture and extracellular matrix preparation

HT1080 human fibrosarcoma cells and H1299 non-small lung carcinoma cells were subcultured at 37°C in 1 Dulbecco's modified Eagle's medium (DMEM), supplemented with 10% FCS and 2 mM L-glutamine. HT1080 and H1299 cells were seeded on top of or incorporated within three-dimensional fibrillar collagen gels. Fibrillar collagen type I gels (1 mg/ml final concentration) were prepared by neutralizing a solution of collagen type I (Vitrogen 100, Cohesion, Paulo Alto, CA, USA) with 1/10 volume 10 DMEM concentrate, diluted to a final concentration of 1 with distilled H₂O, to which 1/11 volume 0.1 N NaOH was added. BE and LS174T colon carcinoma cells were subcultured with DMEM, supplemented with 10% FCS and 2 mM L-glutamine. BE and LS174T cells were seeded on top or within Matrigel™ Basement Membrane Matrix (BD Biosciences, Oxford, UK).

Monomer collagen-coated dishes were prepared by incubating with 0.1 mg/ml of collagen solution in 0.1 M acetic acid for 24 h at 37°C. Monomer collagen-coated dishes were washed twice with 1 DMEM before cell seeding. Fibronectin-coated dishes (2 g/ml) were prepared by incubation with human plasma fibronectin (BD Biosciences) for 2 h at room temperature. Dishes were rinsed with phosphate-buffered saline (PBS) and air dried before cell seeding.

Invasion assays

HT1080 or H1299 cells (1 10⁴) in 1 DMEM, supplemented with 0.2% FCS and 2 mM L-glutamine were seeded on top of a fibrillar collagen gel (60 l volume), set within the upper chamber of a transwell, above an 8 m pore-size polycarbonate filter. 1 DMEM, supplemented with 10% FCS and 2 mM L-glutamine (750 l total volume), was placed in the lower chamber to provide a chemotactic gradient, and the assay mixture was left for 3 days. Invasion was visualized by staining the cells directly with 4 M calcein acetoxymethyl ester (Molecular Probes Europe), as previously described (Scott et al., 2004). Briefly, cells were incubated with 4 M calcein in serum-free DMEM for 1 h at 37°C followed by confocal microscopic analysis using a Bio-Rad MRC 600 confocal illumination unit attached to a Nikon Diaphot inverted microscope. Use of this dye ensures that only viable cells were scored. Quantification of cell invasion was performed with slight modifications to the previously described protocol (Hennigan et al., 1994). Briefly, using a 10 objective, optical sections were scanned at 5 m intervals from the top of the collagen gel down. The number of cells per optical section is proportional to the number of positive pixels per optical section. To quantify invasion, the accumulated sum of positive pixels present in optical sections between 20 and 55 m below top of collagen gels was expressed as a percentage of the number of positive pixels on top and within the collagen gel. This value therefore represents the number of cells invading beyond 20 m through to 55 m in distance, expressed as a relative percentage of total cells in the assay. BE and LS174T invasion through Matrigel was analysed as above with the exception that images were collected at sequential 10 m section through Matrigel. Invasion was quantified as the number of cells invading beyond 40 m through to 100 m in distance.

Computer-assisted time-lapse reconstruction

HT1080, LS174T and BE cells (1 10⁵) were prestained in suspension with the lipophillic fluorescent dye Cell-Tracker green CMFDA (Molecular Probe Europe) as per the manufacturer's instructions. Briefly, cells were incubated in normal culture media with Cell-Tracker green (10 M) for 45 min at 37°C. Cells were washed once in normal culture media and incubated for a further 30 min. Stained cells were then mixed with 1.5 ml of freshly made fibrillar collagen or Matrigel solution and incubated in a 60 mm glass bottom dish (Iwaki). The cell-containing gels were allowed to polymerize for 5 h at 37°C. Normal culture media (2 ml) were subsequently overlaid and fibrillar collagen- or Matrigel-containing stained cells were incubated for 24 h. Time-lapse microscopy of Cell-Tracker green-stained cells embedded within the three-dimensional matrices was performed with a Leica SP2 confocal imaging unit combined with an inverted microscope. Generation of three-dimensional reconstructed time-lapse movies and subsequent tracking of total distance of cellular invasion was performed using Image ProPlus software (Media Cybernetics UK).

Antibodies and reagents

Antibodies for Western blot detection and immunocytochemistry included ^354-534N pp125^FAK (BD Transduction), calpain 2 (Research Diagnostics Inc., Concord, MA, USA), 397Y-FAK (Biosource, UK), talin (Sigma, Poole, UK), actin (Sigma), 327-Src (Oncogene) and 416Y-Src (New England Biolabs, Hitchin, UK). 2 (Chemicon, Temicula, CA, USA), 1 (Chemicon). Anti-mouse and anti-rabbit peroxidase-conjugated secondary antibodies were purchased from New England Biolabs Inc. Inhibition of extracellular proteolytic activity was achieved by using the following PI cockatil (PI mix): GM6001 (50 M; Calbiochem-Novabiochem Corp., Nottingham, UK), E64 (250 M; Sigma), pepstatin A (100 M; Sigma), leupeptin (2 M; Sigma) and aprotinin (2.2 M; Sigma).

Calpain inhibitor studies were performed using calpain inhibitor 1 (ALLN-1/10 M; Calbiochem-Novabiochem Corp., Nottingham, UK) and PD150606 (50 M; Calbiochem-Novabiochem Corp.). Rho kinase was inhibited using Y27632 (10 M; Calbiochem-Novabiochem Corp.). To block 21 integrin function, HT1080 cells were preincubated at 37°C for 30 min with non-blocking (PIH6) and blocking (BHA2.1) anti-21 antibodies (Chemicon). RGD binding integrins were blocked by preincubation with 100 M cyclic RGD peptide (Peptides International, Louisville, KY, USA) at 37°C for 30 min. Cells were also incubated with non-blocking cyclic RAD peptide (Peptides International) as control. Src inhibitor studies were performed using PP2 (10 M; Sigma) and AP23464 (1 M; Ariad Pharmaceuticals, Cambridge, MA, USA). Both drugs were made up as stocks of 10 mM and diluted to final concentrations in culture medium.

Protein immunoblotting

Cells cultured on fibrillar collagen gels were washed twice with PBS and transferred to a centrifuge tube. Cells within fibrillar collagen gels were subjected to several washes in PBS to remove serum from collagen gel. Cells and collagen were then incubated in lysis buffer (10 mM Tris pH 7.4, 150 mM NaCl, 0.5% NP-40, 1 mM ethylenediamine tetraacetic acid (EDTA), 1 mM ethylene glycol bis (-aminoethylether)-N,N,N',N'-tetraacetic acid (EGTA), 1 mM dithiothreitol (DTT), 0.5 mM NaF, 10 mM -glycerophosphate, 10 mM Na₄P₂O₇, 100 M NaVO₄, with PIs, 1 mM PMSF, 10 g/ml leupeptin and 10 g/ml aprotinin) on ice for 30 min. Lysates were clarified by high-speed centrifugation at 4°C, and supernatants containing solubilized cellular protein extracts were collected, supplemented with SDS-sample buffer and separated by 10% SDS–PAGE and immunoblotted with specific antibodies. All blots shown are representatives of several repeated blots.

Protein immunoprecipitation

Cells were washed twice with PBS and lysed in modified RIPA buffer (10 mM Tris pH 7.4, 150 mM NaCl, 0.5% NP-40, 1 mM EDTA, 1 mM EGTA, 1 mM DTT, 0.5 mM NaF, 10 mM -glycerophosphate, 10 mM Na₄P₂O₇, 100 M NaVO₄, with PIs, 1 mM PMSF, 10 g/ml leupeptin and 10 g/ml aprotinin) on ice for 30 min. Lysates were clarified by high-speed centrifugation at 4°C and incubated overnight with 5 g of anti-FAK (BD Transduction) at 4°C with constant mixing. Anti-mouse IgG conjugated to agarose beads was added to the reactions for 1 h at 4°C with constant mixing before washing twice with modified RIPA buffer. Immunoprecipitation reactions were resuspended in SDS-sample buffer and separated by SDS–PAGE, transferred to membrane and immunoblotted with an antibody specific for the phosphorylated tyrosine 397 site on FAK (Biosource, UK).

Gelatin zymography

Serum-free conditioned media were collected from HT1080 cells cultured on fibrillar collagen in the absence or presence of PI mix. A 5 ml portion of conditioned media was concentrated using Amicon ultrafilters (Millipore) to 500 l volume. Samples were supplemented with non-denaturing loading buffer and separated on a 10% SDS–polyacrylamide gel impregnated with 1 mg/ml gelatin. Following electrophoresis, gels were washed twice in 2.5% Triton X-100 for 30 min, briefly rinsed with water and incubated for 24 h at 37°C in collagenase buffer (50 mM Tris-HCl buffer, pH 7.5, 200 mM NaCl and 10 mM CaCl₂). Gels were subsequently fixed and stained in Coomassie blue fixative solution (50% methanol and 10% acetic acid containing 0.25% Coomassie blue R250) for 2 h at room temperature. Finally, gells were destained by several rinses in dH₂O.

siRNA-mediated depletion of calpain 2

siRNA duplexes were synthesized and purified by Proligo (Paris, France). The siRNA targeted sequence specific for calpain 2 (C2siRNA) used in these studies was Oligo 1 (5'-TGAAGAAATCCTGGCTCGATT-3'); as a control, a missense mutated form (C2mut) of the previous sequence was used (5'-TGATGAAATGCTGGCTCGATT-3'). siRNA results were reproduced using a second siRNA (Oligo 2) that targeted a distinct sequence of calpain 2 mRNA (5'-GACTTCACCGGAGGCATTGTT-3'); as a control, a missense mutated form (C2mut) of the previous sequence was used. HT1080 cells were subjected to three sequential rounds of transfection with the above siRNA duplexes. siRNA was transfected into cells using oligofectamine (Invitrogen, Paisley, UK) based on manufacturer's instructions plus modification. Briefly, 8 l of oligofectamine was added to 2 l of Optimem (Invitrogen) and incubated for 5 min at room temperature. This mixture was added to a second mixture containing 20 l of siRNA (20 M) and 20 l of Optimem. The combined transfection mixture was then incubated for 20 min at room temperature before addition to cells. HT1080 cells to be transfected were cultured in 60 mm dishes until 70–80% confluent, then cells were washed once with Optimem and incubated with 1.6 ml Optimem. The combined transfection mixture was then added to cell cultures, which were incubated for 4 h at 37°C. Transfected cultures were then supplemented with a further 1 ml of culture media supplemented with 3 (30%) FCS, to give 10% FCS final concentration in culture media. Transfected cells were incubated for 24 h, trypsinized and seeded into new culture dishes and incubated for a further 24 h to reach 70–80% confluence before re-transfection with siRNA as described above. This process was repeated one further time. Following the third and final transfection with siRNA, cells were cultured for a further 24 h before protein extraction or use in invasion assays.

Flow cytometric analysis of integrin expression

HT1080 cells were maintained on fibrillar collagen gels for 3 days under untreated conditions as described above or in the presence of PI mix, or PI mix+Y27632 (10 M). Cells were recovered from fibrillar collagen by incubation with 2.5 mg/ml type I collagenase (Worthington Biochemical Corp., Lakewood, NJ, USA) for 30 min at 37°C. Cells were subsequently washed in 1 DMEM containing 10% FCS, PBS and fluorescence activated cell sorting buffer (PBS, 0.2% BSA, 0.1% Na azide). 1 10⁶ cells were transferred to separate tubes and resuspended in 100 l of FACS buffer to which 5 g of anti-21 (Chemicon) antibody was added and incubated at 4°C for 1 h with constant mixing. Cells were then washed twice with FACS buffer and incubated with 1:50 dilution of fluorescein isothiocyanate conjugated anti-mouse antibody (Jackson Immunoresearch Laboratories, West Grove, PA, USA) for 30 min at 4°C in the dark with constant mixing. Cells were subsequently washed twice with FACS buffer, fixed with 2% formalin for 10 min at room temperature and washed a further twice with FACS buffer. Level of antigen expression was evaluated by flow cytometry (BD FACScan).

Adhesion assay

Ninety-six-well 'black' plates (Costar) were coated with monomer collagen and fibronectin as described above. HT1080 cells maintained for 3 days in culture on fibrillar collagen in the absence or presence of the PI mix were recovered by collagenase treatment as described above. Cells were washed twice in normal culture media and incubated with the fluorescent dye, Cell-Tracker green CMFDA (10 M) (Molecular Probe Europe) for 45 min at 37°C. Cells were washed once in normal culture media and incubated for a further 30 min at 37°C before resuspension in normal culture media at a concentration of 1 10⁵ cells/ml. 1 10⁴ cells were seeded in quadruplicate onto monomer collagen-coated and fibronectin-coated dishes for 30 min. Cells were washed twice with PBS and fluorescence emission of attached cells was quantified at 444 nm excitation and 538 nm emission wavelengths using a Flexstation fluorescent plate reader (Molecular Devices). The percentage of total number of cells that adhered was calculated by relating fluorescence of adhered cells to fluorescence emission of a 1 10⁴ cells/ml cell suspension.

References

1. Adachi Y, Itoh F, Yamamoto H, Iku S, Matsuno K, Arimura Y et al. (2001). Tumour Biol 22: 247–253. | Article | PubMed | ChemPort |
2. Bobak D, Moorman J, Guanzon A, Gilmer L, Hahn C. (1997). Oncogene 15: 2179–2189. | Article | PubMed | ISI | ChemPort |
3. Brakebusch C, Fassler R. (2003). EMBO J 22: 2324–2333. | Article | PubMed | ISI | ChemPort |
4. Brand K, Baker AH, Perez-Canto A, Possling A, Sacharjat M, Geheeb M et al. (2000). Cancer Res 60: 5723–5730. | PubMed | ISI | ChemPort |
5. Brunton VG, Avizienyte E, Fincham VJ, Serrels B, Metcalf III CA, Sawyer TK et al. (2005). Cancer Res 65: 1335–1342. | Article | PubMed | ISI | ChemPort |
6. Carragher NO, Fincham VJ, Riley D, Frame MC. (2001). J Biol Chem 276: 4270–4275. | Article | PubMed | ISI | ChemPort |
7. Carragher NO, Frame MC. (2004). Trends Cell Biol 14: 241–249. | Article | PubMed | ISI | ChemPort |
8. Carragher NO, Levkau B, Ross R, Raines EW. (1999). J Cell Biol 7: 619–630. | Article |
9. Carragher NO, Westhoff MA, Fincham VJ, Schaller MD, Frame MC. (2003). Curr Biol 13: 1442–1450. | Article | PubMed | ISI | ChemPort |
10. Chambers AF, Matrisian LM. (1997). J Natl Cancer Inst 89: 1260–1270. | Article | PubMed | ChemPort |
11. Cooke DW, Patel YM. (2005). Mol Cell Endocrinol 232: 37–45. | Article | PubMed | ISI | ChemPort |
12. Cuevas BD, Abell AN, Witowsky JA, Yujiri T, Johnson NL, Kesavan K et al. (2003). EMBO J 22: 3346–3355. | Article | PubMed | ISI | ChemPort |
13. Cukierman E, Pankov R, Yamada KM. (2002). Curr Opin Cell Biol 14: 633–639. | Article | PubMed | ISI | ChemPort |
14. Dourdin N, Bhatt AK, Dutt P, Greer PA, Arthur JS, Elce JS et al. (2001). J Biol Chem 276: 48382–48388. | PubMed | ISI | ChemPort |
15. Etienne-Manneville S, Hall A. (2002). Nature 420: 629–635. | Article | PubMed | ISI | ChemPort |
16. Falcioni R, Cimino L, Gentileschi MP, D'Agnano I, Zupi G, Kennel SJ et al. (1994). Exp Cell Res 210: 113–122. | Article | PubMed | ISI | ChemPort |
17. Fearon ER, Vogelstein B. (1990). Cell 61: 759–767. | Article | PubMed | ISI | ChemPort |
18. Franco SJ, Rodgers MA, Perrin BJ, Han J, Bennin DA, Critchley DR et al. (2004). Nat Cell Biol 6: 977–983. | Article | PubMed | ISI | ChemPort |
19. Friedl P. (2004). Curr Opin Cell Biol 16: 14–23. | Article | PubMed | ISI | ChemPort |
20. Friedl P, Borgmann S, Brocker EB. (2001). J Leukoc Biol 70: 491–509. | PubMed | ISI | ChemPort |
21. Friedl P, Wolf K. (2003). Nat Rev Cancer 3: 362–374. | Article | PubMed | ISI | ChemPort |
22. Glading A, Bodnar RJ, Reynolds IJ, Shiraha H, Satish L, Potter DA et al. (2004). Mol Cell Biol 24: 2499–2512. | Article | PubMed | ISI | ChemPort |
23. Glading A, Chang P, Lauffenburger DA, Wells A. (2000). J Biol Chem 275: 2390–2398. | Article | PubMed | ISI | ChemPort |
24. Goss KJ, Brown PD, Matrisian LM. (1998). Int J Cancer 78: 629–635. | Article | PubMed | ISI | ChemPort |
25. Hanke JH, Gardner JP, Dow RL, Changelian PS, Brissette WH, Weringer EJ et al. (1996). J Biol Chem 271: 695–701. | Article | PubMed | ISI | ChemPort |
26. Hennigan RF, Hawker KL, Ozanne BW. (1994). Oncogene 9: 3591–3600. | PubMed | ISI | ChemPort |
27. Huttenlocher A, Palecek SP, Lu Q, Zhang W, Mellgren RL, Lauffenburger DA et al. (1997). J Biol Chem 272: 32719–32722. | Article | PubMed | ISI | ChemPort |
28. Huttenlocher A, Sandborg RR, Horwitz AF. (1995). Curr Opin Cell Biol 7: 697–706. | Article | PubMed | ISI | ChemPort |
29. Hynes RO. (1992). Cell 69: 11–25. | Article | PubMed | ISI | ChemPort |
30. Jaspars LH, Bonnet P, Bloemena E, Meijer CJ. (1996). J Pathol 178: 36–43. | Article | PubMed | ISI | ChemPort |
31. Krylyshkina O, Kaverina I, Kranewitter W, Steffen W, Alonso MC, Cross RA et al. (2002). J Cell Biol 156: 349–359. | Article | PubMed | ISI | ChemPort |
32. Lokuta MA, Nuzzi PA, Huttenlocher A. (2003). Proc Natl Acad Sci USA 100: 4006–4011. | Article | PubMed | ChemPort |
33. Lozonschi L, Sunamura M, Kobari M, Egawa S, Ding L, Matsuno S. (1999). Cancer Res 59: 1252–1258. | PubMed | ISI | ChemPort |
34. Malliri A, Collard JG. (2003). Curr Opin Cell Biol 15: 583–589. | Article | PubMed | ISI | ChemPort |
35. Mamoune A, Luo JH, Lauffenburger DA, Wells A. (2003). Cancer Res 63: 4632–4640. | PubMed | ISI | ChemPort |
36. Mignatti P, Rifkin DB. (1993). Physiol Rev 73: 161–195. | PubMed | ISI | ChemPort |
37. Palecek SP, Huttenlocher A, Horwitz AF, Lauffenburger DA. (1998). J Cell Sci 111: 929–940. | PubMed | ISI | ChemPort |
38. Pavlaki M, Zucker S. (2003). Cancer Metast Rev 22: 177–203. | Article | ISI | ChemPort |
39. Petroll WM, Ma L. (2003). Cell Motil Cytoskeleton 55: 254–264. | Article | PubMed | ISI |
40. Reiske HR, Kao SC, Cary LA, Guan JL, Lai JF, Chen HC. (1999). J Biol Chem 274: 12361–12366. | Article | PubMed | ISI | ChemPort |
41. Rintoul RC, Sethi T. (2002). Clin Sci (London) 102: 417–424. | ChemPort |
42. Sahai E, Marshall CJ. (2003). Nat Cell Biol 5: 711–719. | Article | PubMed | ISI | ChemPort |
43. Scott LA, Vass JK, Parkinson EK, Gillespie DA, Winnie JN, Ozanne BW. (2004). Mol Cell Biol 24: 1540–1559. | Article | PubMed | ISI | ChemPort |
44. Shepherd FA, Giaccone G, Seymour L, Debruyne C, Bezjak A, Hirsh V et al. (2002). J Clin Oncol 20: 4434–4439. | Article | PubMed | ISI | ChemPort |
45. Sieg DJ, Hauck CR, Schlaepfer DD. (1999). J Cell Sci 112: 2677–2691. | PubMed | ISI | ChemPort |
46. Sternlicht MD, Werb Z. (2001). Annu Rev Cell Dev Biol 17: 463–516. | Article | PubMed | ISI | ChemPort |
47. Wagenaar-Miller RA, Gorden L, Matrisian LM. (2004). Cancer Metast Rev 23: 119–135. | Article | ISI | ChemPort |
48. Wang KK, Nath R, Posner A, Raser KJ, Buroker-Kilgore M, Hajimohammadreza I et al. (1996). Proc Natl Acad Sci USA 93: 6687–6692. | Article | PubMed | ChemPort |
49. Webb DJ, Donais K, Whitmore LA, Thomas SM, Turner CE, Parsons JT et al. (2004). Nat Cell Biol 6: 154–161. | Article | PubMed | ISI | ChemPort |
50. Wolf K, Mazo I, Leung H, Engelke K, von Andrian UH, Deryugina EI et al. (2003). J Cell Biol 160: 267–277. | Article | PubMed | ISI | ChemPort |
51. Yamada KM, Pankov R, Cukierman E. (2003). Braz J Med Biol Res 36: 959–966. | Article | PubMed | ISI | ChemPort |
52. Zamir E, Geiger B. (2001). J Cell Sci 114: 3583–3590. | PubMed | ISI | ChemPort |
53. Zhai J, Lin H, Nie Z, Wu J, Canete-Soler R, Schlaepfer WW et al. (2003). J Biol Chem 278: 24865–24873. | Article | PubMed | ISI | ChemPort |
54. Zucker S, Vacirca J. (2004). Cancer Metast Rev 23: 101–117. | Article | ISI | ChemPort |

Acknowledgements

We thank Jeff Evans for comments. This work was supported by Cancer Research UK by Beatson Institute for Cancer Resarch core grants to MCF and BWO, and by an AICR project grant to VGB.