EGFR-dependent migration of glial cells is mediated by reorganisation of N-cadherin

Receptor tyrosine kinases of the EGF receptor (EGFR) family are implicated in multiple cellular functions. Different dimeric receptor complexes mediate the various effects on the cellular behaviour. The dimerisation leads to the activation of the intrinsic kinase domain of the receptors and consequently to phosphorylation of specific tyrosine residues, which serve as docking sites for adaptor and signalling molecules. The human epidermal growth factor 2 (ERBB2, also known as NEU, and hereafter referred to HER2) seems to be the preferred heterodimerisation partner of the other ErbB receptors, although no specific HER2 ligand has been identified so far. These heterodimers are responsible for strong and prolonged activation of downstream signalling pathways (Holbro et al., 2003). One of the most important cellular responses upon EGFR stimulation is the activation of the ERK1/2 (hereafter referred to as MAPK) and phosphoinositide 3-kinase (PI3K)/Akt pathway, which have important roles in cell migration (Brazil et al., 2004; Cantley, 2002; Jorissen et al., 2003). Thus, HER2 appears to be involved in EGFR-mediated migration, most probably mediated by EGFR-HER2 heterodimers (Dittmar et al., 2002).

The infiltration and migration of glioma cells into the surrounding brain tissue is a major clinical problem in human neuro-oncology. As in many other cancer types, including breast and lung cancer, members of the EGF receptor family – particularly EGFR and HER2 – are also implicated in the tumorigenesis and tumour progression of glioblastomas. Amplification of EGFR can be detected in ∼40% of all cases and it seems to be one of the most common genomic alterations in these tumours – leading to aberrant signal transduction in the ErbB receptor network, possibly resulting in enhanced migration and invasion (Louis, 2006). Overexpression of HER2 significantly increases early mortality of patients suffering from primary malignant brain tumours (Koka et al., 2003; Mineo et al., 2007; Potti et al., 2004).

Tumour progression, migration and invasion into surrounding tissue are accompanied by a reorganisation of the cytoskeleton, and the assembly and disassembly of cell-cell adhesions (Cavallaro and Christofori, 2004; Perego et al., 2002). These processes are controlled by members of the cadherin superfamily of cell surface glycoproteins involved in Ca²⁺-dependent cellular adhesion. The role of E-cadherin is well characterised in epithelial tumours, where it functions as a tumour suppressor that is downregulated in neoplastic transformation, thus decreasing the strength of cellular adhesion within a tissue and promoting tumour-cell migration (Birchmeier and Behrens, 1994). N-cadherin, another member of the cadherin family, which can be detected in neural cell types including astrocytes, is also expressed in glial tumour cells (Asano et al., 2000). N-cadherin can stimulate migration and invasion, but also adhesion, depending on the cell context (Derycke and Bracke, 2004). Therefore, its role in gliomagenesis is less clear, as the results obtained so far are controversial (Barami et al., 2006).

Ligand binding to the ErbB receptors induces the formation of homo- or heterodimeric receptor complexes. However, the impact of ErbB receptor expression, especially EGFR and HER2, on the aberrant cell behaviour of glial tumour cells is only poorly understood (Andersson et al., 2004). It has been shown that EGFR signalling regulates cadherins in other tumour types, and that this regulation might be important for cellular migration and tumour infiltration (Lu et al., 2003). The molecular mechanisms of the EGFR-cadherin crosstalk are, however, not completely understood.

Here, we show that HER2 affects EGFR-mediated signal transduction in human glioblastoma cells, and that the modulation of EGFR signalling by HER2 leads to alterations in the migratory phenotype of the cells, which is mediated by the reorganisation of N-cadherin expression and localisation.

Parental LN18 glioblastoma cells express high levels of endogenous EGFR, but only barely detectable levels of HER2 receptor (Fig. 2A). Selected clones from stably transfected LN18 cells express high amounts of HER2 receptor, either the wild-type (HER2N), the constitutively active (HER2VE) or the dominant-negative (HER2VEKA) receptor variant (Fig. 1). Under serum-free conditions and without EGF stimulation, high levels of tyrosine phosphorylation of HER2VE was detected using an antibody specifically recognising P-HER2 phosphorylated at Tyr1248. However, we found only low levels of P-HER2VE after transfection with wild-type HER2N, and no phosphorylation in untransfected parental cells and in cells expressing the dominant-negative HER2VEKA. Stimulation with EGF resulted in enhanced phosphorylation of HER2N, but no further increase of the constitutive activation in HER2VE-cells was detected. No phosphorylation after EGF treatment was observed in parental LN18 cells or in HER2VEKA cells. Expression levels of endogenous EGFR were unaffected by the transfections, and only cells that expressed the HER2VE construct showed low levels of constitutive phosphorylation of EGFR at Tyr1086 under serum starvation and without EGF stimulation. After stimulation with EGF, cells expressing HER2N and constitutively active HER2VE exhibited a strong phosphorylation of EGFR that was much stronger than in untransfected LN18 control cells. Cells transfected with kinase deficient HER2VEKA exhibited EGFR phosphorylation comparable to control cells (Fig. 2A). These data show that HER2 interferes with EGFR activation in human glioblastoma cells.

EGFR-dependent activation of the PI3K/Akt- and the MAPK-pathway was investigated in the different LN18 HER2 cell clones after serum starvation and stimulation with EGF for 5 minutes. Without EGF stimulation, only cells expressing HER2N or HER2VE showed a constitutive phosphorylation of Akt and MAPK. This was more pronounced in cells stably expressing HER2VE. After stimulation with EGF, all clones exhibited phosphorylation of Akt – detected using an antibody specifically recognising P-Akt (phosphorylated at Ser473), which was, however, clearly diminished in HER2VEKA cells. Activation of MAPK after EGF treatment was very strong in cells expressing HER2N or HER2VE. Parental cells and HER2VEKA-expressing cells showed weaker albeit comparable signals in western blot analysis using an antibody recognising P-MAPK (ERK1 and ERK2 phosphorylated at Thr202 and Tyr204, respectively). There was no difference in the overall expression of Akt and MAPK detected with specific antibodies, respectively (Fig. 2B). These data show, that HER2VEKA acts in a dominant-negative fashion on the activation of the EGFR-dependent intracellular signal transduction pathways, whereas the HER2VE receptor enhances phosphorylation even without EGF stimulation.

The migratory behaviour of the different LN18 clones was tested after serum starvation in wound-healing assays. After 48 hours, migration was abolished in LN18 cells transfected with the dominant-negative HER2VEKA construct. By contrast, migration and wound closure was increased in LN18 cells expressing the constitutively active HER2VE receptor. LN18 cells overexpressing the wild-type HER2N receptor showed migration rate that was similar to that of LN18 control cells (Fig. 3A,B). These data indicate a functional role of the HER2 receptor for glial tumour cell migration, because the expression of the dominant-negative HER2VEKA construct completely prevents migration.

Also, the morphology of the cells changed. Cells expressing the dominant-negative HER2VEKA-receptor mutant grew in clusters and showed tight intercellular junctions in an epithelial phenotype. Expression of the constitutively active HER2VE receptor mutant resulted in a complete loss of cell-cell adhesion and the cells appeared more fibroblastic. Finally, cells expressing the wild-type HER2N receptor exhibited a phenotype that was similar to that of the parental LN18 cells, but with slightly more-dissociated colonies (Fig. 3C). These data indicate that the presence of HER2 within the cells affects the cell adhesion system.

Changes in the cellular morphology and migration indicate alterations in the cell adhesion system. Since N-cadherin seems to have a crucial role in migration and invasion of glioblastoma cells (Barami et al., 2006), we analysed its role in HER2-mediated cell migration. Western blot analysis revealed N-cadherin expression in untransfected LN18 control cells. In HER2N- and HER2VE-expressing cells, N-cadherin expression was downregulated compared with expression in control cells. In the HER2VEKA clone, however, a pronounced upregulation, and by far the strongest N-cadherin expression, could be detected (Fig. 4A).

Furthermore, we examined the localisation pattern of N-cadherin in the HER2-transfected cell clones. Immunofluorescence staining revealed the presence of N-cadherin at the cell surface at a high density only in the non-migrating HER2VEKA cells (Fig. 4B). By contrast, all other cell clones showed no membrane staining but, instead, a predominant cytoplasmic distribution of N-cadherin (Fig. 4B). The data suggest that the EGFR-HER2-receptor-dependent migration of glioblastoma cells is mediated by changes in the cell adhesion system, and that upregulation of N-cadherin at the cell surface correlates with diminished migration.

The importance of N-cadherin for the migratory behaviour of LN18 glioma cells was confirmed by the downregulation of N-cadherin when using small interfering RNA (siRNA). When non-migrating HER2VEKA cells were transfected with siRNA specifically targeting N-cadherin, the migratory phenotype could be re-established to a degree comparable with that in HER2VE-cells (Fig. 5A,B). The protein levels of N-cadherin in HER2VEKA cells was reduced by 80% to a level comparable to that in migrating HER2VE-cells (Fig. 5C). Also, the cells showed a decrease in N-cadherin localisation at the cell surface (Fig. 5D).

To further elucidate the impact of the individual receptors on the regulation of N-cadherin and the changes in migration, we inhibited the receptors by using small tyrosine kinase inhibitors (TKIs) either mono-specific for EGFR (AG1478) or HER2 (AG825) alone, or by using the bi-specific drug AEE788, which binds both receptors with the same affinity and equally blocks the signals from both receptors (Traxler et al., 2004). Experiments were carried out in the constitutively activated HER2VE cells for 24 hours. Western blot analysis of treated HER2VE cells revealed that only inhibition of both receptors together leads to an increase in N-cadherin expression that is comparable with its levels in HER2VEKA cells. Blocking either EGFR or HER2 alone did not change N-cadherin expression compared with that in untreated HER2VE-cells (Fig. 6A). Furthermore, only inhibition with bi-specific AEE788 led to cell-surface expression of N-cadherin comparable with its pattern in HER2VEKA cells (Fig. 6B). These data and the effects of the dominant-negative HER2VEKA construct seem to indicate that signal transduction from EGFR-HER2 heterodimers but not from either of the receptors alone regulate the expression and localisation of N-cadherin.

Wound healing assays of HER2VE-cells after inhibitor treatment confirmed the relevance for the migratory behaviour, because only the AEE877 treatment completely abolished migration. The cells displayed the same clustered cell morphology as HER2VEKA cells. In case of treatment with either tyrphostin AG825 or AG1478, the number of migrating cells was similar to that in untreated control HER2VE cells (Fig. 6C). These results were confirmed in the glioblastoma cell lines LN229 (Ishii et al., 1999) and G139 (Schlegel et al., 2000), which possess high levels of both, EGFR and HER2 receptors. By using the bi-specific TKI, we observed an inhibition of migration that was accompanied by of N-cadherin upregulation and its re-localisation to the cell membrane (data not shown).

To get further insights into the mechanisms, by which EGFR-HER2 might mediate changes in N-cadherin expression and cellular distribution, we investigated direct interactions between EGFR-HER2 and N-cadherin–β-catenin complexes. By western blot analysis after immunoprecipitation of HER2 or EGFR, we found an N-cadherin immunoreaction in HER2VE cells but not in HER2VEKA cells. Consequently, HER2 or EGFR was detectable by western blotting only in HER2VE cells after immunoprecipitation by using antibodies against N-cadherin (Fig. 7A). These data indicate a direct interaction of activated EGFR-HER2-receptor complexes with N-cadherin. Moreover, western blot analysis, using anti-phospho-tyrosine antibodies after immunoprecipitation of β-catenin to detect tyrosine phosphorylated β-catenin, showed a stronger immunoreaction in HER2VE cells, and western blot analysis using antibody against phosphorylated β-catenin (P-β-catenin; phosphorylated at Thr41 and Ser45) showed a stronger immunoreaction in HER2VEKA cells. These data seem to indicate that, EGFR-HER2-dependent phosphorylation of β-catenin at tyrosine residues leads to the elimination of N-cadherin–β-catenin complexes from the plasma membrane, and prevention of β-catenin Thr41/Ser45 phosphorylation in HER2VE cells is most likely to be inhibited by PI3K/Akt-dependent phosphorylation of GSK3β (Fig. 7B).

Although initially considered as the prototypic mitogen pathway, EGFR signalling is also involved in other cellular functions, dependent on the type of preferred multimeric receptor complexes. Invasion is the hallmark of malignancy in many cancers, and EGFR-dependent signalling has been shown to be involved in cell migration and/or invasion in a variety of human cancers (Holbro and Hynes, 2004). The different signal transduction of HER2/ErbB2 leading to various cellular behaviours, e.g. mitosis vs migration, has only begun to be resolved at its molecular level. In several types of tumour, HER2 has been implicated in the shift towards an invasive phenotype (Yarden, 2001), but the impact of ErbB receptors – especially HER2 – in tumorigenesis of primary brain tumours is not well understood (Schwechheimer et al., 1998; Thomas et al., 2003). The invasive phenotype of these tumours, however, is of highest biological impact because their infiltration of surrounding brain tissue is the most pressing problem in the clinical management of glial tumours. The aim of this study was to investigate of the impact of HER2- on EGFR-related signalling in migratory and invasive phenotypes of malignant glial tumour cells. We therefore transfected different HER2 constructs, including the wild-type receptor (HER2N), a constitutive active variant (HER2VE) and a dominant-negative construct (HER2VEKA) (Messerle et al., 1994), into EGFR-overexpressing LN18 glioblastoma cells.

It has been shown in several studies that HER2 is expressed in malignant human gliomas and that its overexpression correlates with poor prognosis (Koka et al., 2003; Potti et al., 2004). The role of HER2 in glial cell tumour biology, however, is not clear, because no evidence for genetic alterations in gliomas has been found so far. The strong correlation of HER2 expression with that of EGFR has therefore assumed to be owing to a general co-expression of both receptors, and has been viewed as an indicator for the interaction of HER2 with EGFR signalling (Dietzmann and von Bossanyi, 1994; Schlegel et al., 1994). This opinion is corroborated by the results of our present study. The stable transfection of HER2 constructs resulted in an increased phosphorylation of EGFR (at Tyr1086) after EGF stimulation of HER2N- and HER2VE-expressing glioma cells that exceeded the levels of EGFR activation in parental and HER2VEKA cells. HER2N- and HER2VE-cells also exhibited high levels of P-HER2 (phosphorylated at Tyr1248) following stimulation with EGF. Moreover, HER2VE cells showed EGFR phosphorylation even without EGF stimulation. LN18 cells stably transfected with dominant-negative HER2VEKA showed neither HER2 activation, nor activation of EGFR exceeding endogenous activation levels of EGF-treated parental cells. Our data indicate that the increased EGFR phosphorylation is mediated by EGFR-HER2 heterodimers, because only cells that express functional HER2 were able to increase EGFR phosphorylation, thereby overcoming levels of endogenous activation.

The impact of the different heteromeric ErbB-receptor complexes on specific cellular reactions in glial cells is still poorly understood. In epithelial cells, signalling that involves HER2 seems to mediate cellular motility and invasion (Wolf-Yadlin et al., 2006; Yarden, 2001). Moreover, the phosphorylation site Tyr1248 of the HER2 receptor (investigated in this study) seems to be necessary for the induction of migration (Dittmar et al., 2002).

Furthermore, we observed morphological changes and alterations in the cell adhesion system after transfecting cells with the different HER2 receptor constructs. HER2VE-transfected cells lost cell-cell adhesion, whereas HER2VEKA-expressing cells grew in tight clusters. HER2-mediated changes in cell morphology were accompanied by changes in the motility of glioma cells, which also seemed to be affected by HER2 receptors because constitutively active HER2VE caused increased migration and dominant-negative HER2VEKA blocked migration. These results imply that overexpression of a functionally active HER2 receptor can modulate the EGFR-HER2 signal transduction in a direction that leads to an increase of migratory properties.

This assumption was also confirmed by blocking the receptors using TKIs. Only the concomitant block of EGF and HER2 using TKI AEE788 (Traxler et al., 2004) was able to abolish migration of HER2VE-cells to the extent of non-migrating HER2VEKA cells. Furthermore, HER2VE-cell morphology changed to the tight-clustered phenotype of the HER2VEKA cells. The fact, that we were not able to achieve similar effects by blocking only one of the two investigated receptors, seems to point towards the EGFR-HER2 heterodimer as being the most potent receptor configuration regarding modulation of migration within the investigated cell system.

It has been shown that EGFR-dependent signalling pathways, namely the PI3K/Akt and the MAPK pathway, are – among others – involved in cell migration and adhesion (Demuth and Berens, 2004; Jorissen et al., 2003). We observed the activation of Akt and MAPK in unstimulated HER2VE-expressing cells, which indicates that the constitutively active HER2VE receptor can initiate downstream signalling. After stimulation with EGF, phosphorylation of Akt and MAPK was clearly attenuated in HER2VEKA-expressing cells. All other stably transfected cell clones showed increased, but almost equal levels of P-Akt and P-MAPK. Thus, the PI3K/Akt pathway seems to be involved in the regulation of migratory properties of glioma cells. However, activation of Akt and MAPK showed no significant differences within the migrating HER2N- and HER2VE-expressing clones, indicating that other factors might also be involved in the motile phenotype of glioma cells. Otherwise, the basic activation levels of Akt and MAPK in unstimulated HER2VE-cells may be sufficient to enhance migratory properties, because the corresponding wound-healing assays were also carried out without EGF and showed clear effects.

Differences in the migratory behaviour and changes in cell morphology were mediated by N-cadherin that was upregulated and recruited to the plasma membrane in the non-migrating HER2VEKA cells. The function of N-cadherin in human cancer, however, is still viewed as being controversial because N-cadherin can stimulate migration and invasion of tumour cells, as well as adhesion – depending on the cell-type and -context (Angst et al., 2001; Derycke and Bracke, 2004). It has been shown by several authors that N-cadherin is expressed in human malignant gliomas (Barami et al., 2006). Here, again, its impact on tumorigenesis and tumour progression is not clear. There is evidence, that N-cadherin expression levels correlate with the degree of malignancy in human gliomas (Utsuki et al., 2002). However, no direct correlation has been found between N-cadherin expression levels and invasiveness (Shinoura et al., 1995). One study, which has been using astrocytic tumours as a model system, described that increased N-cadherin levels correlate with decreased invasiveness, indicating the importance of this N-cadherin for the infiltrative phenotype of gliomas (Asano et al., 2004). This finding is in accordance with the results of our study. We found a direct correlation between N-cadherin expression levels, its localisation at the membrane and the absence of migration in LN18 cells, which was dependent on the HER2-receptor status of the cell. In addition, our data show that not only the expression levels, i.e. the amount of N-cadherin in the cell, are important for the migratory behaviour of cells but, particularly, its distribution along the cell surface.

The inhibition of EGFR and HER2 in HER2VE-expressing cells by using TKIs, leading to abolishment of migration, also influences expression levels of N-cadherin and its localisation pattern. Again, only bi-specific block of both receptors yields an increase in expression and localisation of N-cadherin at the cell surface.

An additional consideration might be that predominantly EGFR-HER2-containing heterodimers modulate the migratory behaviour of the cells by reorganisation of N-cadherin. We investigated two other glioma cell lines, LN229 and G139, that exhibit moderate to high levels of endogenous EGFR and HER2 receptors. Both types of cell are highly motile and show only weak staining for N-cadherin in their membrane. Again, only the EGFR and HER2 receptor inhibition with AEE788 increased N-cadherin levels at the cell membrane, and stopped cell migration completely (our unpublished data).

In conclusion, our data show that HER2 modulates the migratory behaviour of human malignant glioma cells and that this effect is mediated by reorganisation of N-cadherin. The molecular mechanisms, however, that regulate N-cadherin are only poorly understood. We found a direct interaction of EGFR-HER2-receptor complexes with N-cadherin in HER2VE cells, which showed N-cadherin downregulation in the plasma membrane and exhibited a migratory phenotype. Our data are in accordance with previous studies, which also showed a direct interaction between the EGFR and cadherins (Hoschuetzky et al., 1994). In HER2VEKA-expressing cells, in which N-cadherin is localised at the membrane, no interaction with EGFR was found by immunoprecipitation studies. Moreover, β-catenin was intensely phosphorylated at Thr41 in HER2VE cells indicating that N-cadherin–β-catenin complexes are cleaved in the plasma membrane through the phosphorylation of activated receptor complexes. Phosphorylation of β-catenin at Thr41 and Ser45 was only present in HER2VEKA cells. These data indicate that PI3K/Akt-mediated inhibition of GSK3β in HER2VE cells prevents β-catenin phosphorylation, and that its phosphorylation in HER2VEKA cells may lead to the degradation of β-catenin. Alterations of β-catenin levels that affect the subcellular distribution of the molecule seem to have an important role in brain tumours most importantly in medulloblastomas. Our data corroborate previous studies that have connected the distribution of β-catenin and the migration of glioblastomas (Perego et al., 2002).

In conclusion, our data show that the EGFR signal transduction pathway is involved in glial tumour cell migration by reorganising the localisation of N-cadherin, and that this process might be mediated by a direct interaction of the receptor complexes with N-cadherin (Fig. 8).

The human glioblastoma cell line LN18 (Diserens et al., 1981) was cultured in Dulbecco's modified Eagle's Medium (DMEM) supplemented with 10% FCS, 2 mM glutamine, 100 U/ml penicillin and 100 μg/ml streptomycin (Invitrogen, Karlsruhe, Germany) under standard cell culture conditions at 37°C and 5% CO₂. The glial phenotype of LN18 cells has been regularly confirmed by GFAP expression.

The following primary antibodies were used: anti-N-cadherin (BD Transduction Laboratories, Heidelberg, Germany); anti-β-catenin, anti-α-tubulin (DM1A; Sigma, Taufkirchen, Germany). Antibodies against total MAPK, phosphorylated MAPK (P-MAPK; phosphorylated p42 and p44 at Thr202 and Tyr204, respectively), total Akt, phosphorylated Akt (P-Akt; phosphorylated at Ser473), phosphorylated β-catenin (P-β-catenin; phosphorylated at Thr41 and Ser45), phosphorylated GSK3β (phosphorylated at Ser9 and Ser21) and anti-phospho-tyrosine (clone PY20) were from (New England Biolabs, Frankfurt, Germany), antibody against phosphorylated EGFR (P-EGFR; phosphorylated at Tyr1086) was from (Invitrogen, Karlsruhe, Germany). Secondary antibodies coupled to horseradish peroxidase (HRP) were obtained from (New England Biolabs), and antibodies coupled to fluorochromes (Cy3, FITC) from (Invitrogen). Cells were treated with 100 ng/ml EGF (Upstate, Schwalbach, Germany), 5 μM AEE788 (Novartis, Nürnberg, Germany), 10 μM tyrphostin AG1478, 1 μM tyrphostin AG825 (Calbiochem, Darmstadt, Germany) (Bowers et al., 2001; Osherov et al., 1993), all dissolved in DMSO.

Plasmids containing different HER2-receptor mutants, either the wild-type receptor (HER2N), a constitutively activated one with a point mutation in the transmembrane domain (HER2VE) or a dominant-negative receptor (HER2VEKA) that possesses an additional point mutation in the ATP-binding site of the kinase domain (Fig. 1) (Messerle et al., 1994), were transfected into LN18 cells using Lipofectin® (Invitrogen, Karlsruhe, Germany). Briefly, cells were grown in six-well tissue culture dishes overnight and incubated with plasmid-Lipofectin complex in OptiMEM (Invitrogen, Karlsruhe, Germany) for 24 hours, followed by replacement with serum-containing medium for an additional 24 hours. Stable transfectants were selected in 1 mg/ml G418 (Biochrom, Berlin, Germany).

LN18 cells were grown in six-well tissue culture dishes until 70% confluence. Monolayers were wounded with a pipette tip. Cells were washed at least three times with PBS to remove dead cells. Cells were then cultured in serum-free DMEM with or without pharmacological tyrosine kinase inhibitors. Photos of the wound were taken immediately after scratching, and 24 hours and 48 hours later in the same (marked) location of the dish by using a Zeiss Axiovert 135 microscope equipped with a digital camera and AxioVision LE Rel. 4.5 software. Distances of the wound edges were measured on the images from the initial scratches and after 24 hours and 48 hours. The mean differences of the distance after 24 hours and 48 hours were compared with the initial distance and calculated as percentage of the initial distance.

Cells were lysed in lysis buffer (New England Biolabs, Frankfurt, Germany) supplemented with 1 mM PMSF (Roth, Karlsruhe, Germany). Equal amounts of protein (5-10 μg) were separated by SDS-PAGE and transferred to Immobilion membranes (Millipore, Schwalbach, Germany). Blocking of unspecific binding sites was done using 5% (w/v) non-fat dry milk in TBST. Membranes were incubated with polyclonal primary antibodies diluted in TBST for 12-14 hours at 4°C. HRP-conjugated immunoglobulins (diluted 1:10,000 in 5% non-fat dry milk/TBST) served as secondary antibodies and were probed for 1 hour at room temperature. Immunoreactivity was visualised by exposure to high-performance chemiluminescence film (Amersham, UK).

For immunoprecipitation cell lysates containing equal amounts of protein (500 μg) were used. Immunoprecipitation was performed using the Catch and Release Kit (Millipore, Schwalbach, Germany) following the manufacturer's protocol.

Cells were grown on coverslips for 24 hours. After washing with PBS (supplemented with Ca²⁺ and Mg²⁺), cells were fixed for 30 minutes at room temperature in 4% formaldehyde. Fixed cells were permeabilised for 5 minutes at room temperature using 0.25% Triton X-100, blocked for 30 minutes using 5% goat serum/PBS and were then immunostained using primary antibodies overnight at 4°C. After washing with PBS the Cy-3 conjugated secondary antibody (Invitrogen, Karlsruhe, Germany) was applied for 1 hour at room temperature in the dark. Nuclei were stained with DAPI (1 μg/ml) before mounting cells with VectaShield® (Vector Laboratories, Burlingame, CA). Fluorescence images were captured with a Zeiss ApoTome epifluorescence confocal microscope equipped with a digital camera and acquisition software AxioVision LE Rel 4.5 (Zeiss).

LN18 cells expressing the dominant-negative HER2VEKA receptor were seeded in 12-well tissue culture dishes (2×10⁵) shortly before transfection in DMEM containing serum and antibiotics. A mixture of 18 μl HiPerFect Transfection Reagent (Qiagen, Hilden, Germany) and 200 ng small interfering RNA (siRNA) was added. siRNA targeting N-cadherin (gene accession number NM_001792) was from Qiagen (order number 1603686). Cells were incubated as usual, and wound healing assay or protein extraction was performed as described after 24, 48 and 72 hours.

Alternatively, HER2VEKA cells were seeded in six-well tissue culture dishes (1×10⁵) for 24 hours to 70% confluence. RNA interference (RNAi) of PTEN was performed as described in HiScribe™ RNAi Transcription Kit (New England Biolabs, Frankfurt, Germany). A mixture of 4 μl Oligofectamine™ and 20 μg double-stranded RNA (dsRNA) in OptiMEM was added, and cells were incubated for 72 hours before seeded for protein extraction and wound healing assay.

We thank Nancy Hynes, FMI Basel, for helpful comments, and Birgit Luber for the critical reading of the manuscript and for providing antibodies. We are very grateful to Peter Traxler (Novartis AG, Basel) for providing us with AEE788. This study was supported by Graduiertenkolleg 333 of the Deutsche Forschungsgemeinschaft (German Research Foundation).