Introduction

Cadherins represent a family of calcium-dependent transmembrane glycoproteins which are essential to initiate and stabilize cell–cell contacts (Kemler, 1993; Gumbiner, 1996). Cadherins are characterized by an extracellular domain that mediates homophilic interactions between molecules on neighbouring cells. The cytoplasmic portion of cadherins is connected with the actin cytoskeleton via different catenins (Kemler, 1993). The cadherin molecule binds directly to - and -catenin (plakoglobin) which in turn interact with the actin-binding protein -catenin (Herrenknecht et al., 1991; Rimm et al., 1995) and thereby connects cadherin to the actin-based cytoskeleton. E-cadherin-mediated cell–cell adhesion is crucial for formation and maintenance of differentiated epithelial tissues in multicellular organisms. Numerous studies have shown that loss of E-cadherin correlates with tumour cell invasion, whereas re-expression of E-cadherin leads to suppression of invasiveness of tumour cells (Frixen et al., 1991; Schipper et al., 1991; Birchmeier et al., 1996; Chen et al., 1997). Therefore, E-cadherin is considered as a 'suppressor of invasion' (Mareel et al., 1997). One suggested mechanism is that the cadherin/catenin complex forms a mechanical connection between neighbouring tumour cells and prevents their detachment. In addition to a total loss of E-cadherin, - or -catenin expression, some mechanisms have been described to be involved in the regulation of cadherin-based adhesiveness. These include failure in cellular mechanisms regulating the strength of cell–cell adhesion such as phosphorylation of catenins or clustering of cadherins (Gumbiner, 2000). In this context, p120^ctn, another member of the catenin family, was described. This protein is involved in the regulation of cadherin adhesion complex assembly (Anastasiadis and Reynolds, 2000). Originally it was characterized as a substrate of the cellular kinase src but p120^ctn can directly interact with cadherins. The binding site for p120^ctn is located in the juxtamembrane domain of cadherins and is different from that for - and -catenin (Ozawa and Kemler, 1998; Yap et al., 1998). Functionally, p120^ctn influences E-cadherin-mediated cell–cell adhesion by regulating the lateral clustering of cadherin molecules, which seems to be necessary for strong cellular adhesion (Yap et al., 1998).

Pancreatic adenocarcinomas are among the most fatal cancers, because of their extensive invasion into surrounding tissues and metastasis to distant organs, even at an early stage of tumour progression (Poston et al., 1991; Kern et al., 2001). In contrast to the invasive growth, immunohistochemical analyses of pancreatic tumours revealed that only a fraction of tumours and their metastases are completely devoid of E-cadherin (Pignatelli et al., 1994; Weinel et al., 1996; Karayiannakis et al., 2001). In addition, many cell lines derived from pancreatic tumours contain E-cadherin, exhibit calcium-dependent cell–cell adhesion and are able to form metastases in in vivo invasion assays.

To examine the role of different cadherins in suppressing metastases of pancreatic cancer cells, we chose the pancreatic carcinoma cell line MIA PaCa-2 as model system. These cells, derived from an undifferentiated human pancreatic carcinoma, are devoid of E- and N-cadherin and possess a high metastatic and invasive potential after injection into nude mice. Cell clones were generated which stably express full-length E-cadherin, a C-terminal deletion mutant of E-cadherin or wild-type N-cadherin. The reconstitution of cellular adhesion by E-cadherin re-expression induced an epithelial transition and suppressed cell migration and invasion in an in vivo model. These effects could not be detected after N-cadherin expression. Furthermore, we demonstrate that different isoforms of p120^ctn associated with the cadherin molecules. E-cadherin was found complexed with the short isoform, whereas N-cadherin was coupled to the long isoforms of p120^ctn. In addition, only N-cadherin-associated p120^ctn was tyrosine phosphorylated.

Results

In this study, we investigated the contribution of p120^ctn to N- and E-cadherin-mediated cellular adhesion and invasion of pancreatic cancer cells.

For this purpose, we stably expressed wild-type cDNAs of N-cadherin or E-cadherin in MIA PaCa-2 cells, a pancreatic carcinoma cell line which lacks E-cadherin as well as N-cadherin expression. In addition to full-length human E-cadherin (E-cad WT), we stably introduced an E-cadherin mutant with a C-terminal truncation of 35 C-terminal amino acids, which comprises most of the -catenin-binding site (E-cadC). To verify the proper transcription of the introduced constructs, we performed Northern blot hybridization analysis of RNA as well as protein studies using immunoblots. The presence of specific signals for N-cadherin or E-cadherin on protein level is shown in Figure 1a. Northern blot hybridization demonstrated the presence of the corresponding mRNAs (data not shown). Control cells transfected with an empty vector expressed neither E-cadherin nor N-cadherin. E-cadherin of a reduced molecular weight, which is due to its C-terminal truncation, was detectable in MIA PaCa-2 cells expressing E-cadC (Figure 1a).

Figure 1.

Pancreatic MIA PaCa-2 cells were stably transfected with cDNAs encoding for full-length E-cadherin (E-cad WT), a C-terminal deletion mutant (E-cadC) or N-cadherin (N-cad WT). (a) Proteins of the appropriate constructs were detected in the transfected cell clones in Western blot analysis. Mock-transfected MIA PaCa-2 served as controls. (b) The morphology of mock-transfected cells (control), E-cadC-, full-length E-cadherin- and N-cadherin-expressing MIA PaCa-2 was analysed by phase-contrast microscopy. (c) Immunocytochemical staining of E-cadherin or N-cadherin and costaining of the same cells with -catenin. Bottom: -catenin staining in the mentioned cell clones. Bar, 30 m

Full figure and legend (242K)

The morphological characterization of the different cell clones revealed that only those cells transfected with full-length E-cadherin grew in colonies and exhibited a typical cobblestone-like epithelial morphology with extended cell–cell contacts (Figure 1b). Cells expressing N-cadherin or E-cadC exhibited a spindle-shaped, irregular morphology with few cell–cell contacts comparable to mock-transfected cells (Figure 1b). The cellular localization of ectopically expressed cadherins was analysed by immunohistochemistry and confocal laser scanning microscopy. As shown in Figure 1c, in N-cadherin-, wild-type E-cadherin- as well as E-cadC-expressing cells, cadherins were predominantly localized in the cell membrane. Analysis of the distribution of catenins in cells of the different clones demonstrated that only wild-type E-cadherin-expressing cells exhibited enhanced fluorescence signals of - and -catenin (Figure 1c). Both, - and -catenin co-localized together with E-cadherin in areas of cell–cell contacts at the cell membrane. The small dots visible in cells stained for -catenin, especially in E-cadC-cells, did not represent the nucleus as analysed by DNA costaining with DAPI (not shown).

Expression of N-cadherin does not lead to upregulation of - and -catenin and re-establishment of the cadherin/catenin adhesion complex

We examined the effects of different cadherin constructs on catenin gene expression and protein levels. Northern and Western blots were performed showing that mRNA and protein concentrations of - and -catenin were increased only in cells expressing full-length E-cadherin (Figure 2), whereas MIA PaCa-2 cells positive for N-cadherin, E-cadC or E-cadp120 did not show significant differences in the expression of either - or -catenin (Figure 2a and b) as compared to mock-transfected controls. The expression of -catenin was not changed in cell clones analysed in this study (data not shown).

Figure 2.

Protein and RNA analysis of MIA PaCa-2 cells transfected with E- or N-cadherin. Protein lysate or total RNA was prepared from mock-transfected MIA PaCa-2, cell clones transfected with the E-cadherin mutant (E-cadC or E-cadp120), with wild-type E-cadherin (E-cad WT) or N-cadherin (N-cad WT). (a) Total RNA was separated and hybridized with ³²P-labeled probes to detect - or -catenin. 18S rRNA amounts are shown to document equal loading. (b) Protein blots were probed with anti-- or anti--catenin antibodies. Equal loading of the blots is shown by analysis of -actin. (c) The association of the E-cadherin/catenin complex with the cytoskeleton was examined using Triton X-100 fractionated samples. E-cadherin as well as - and -catenin were detected in the insoluble fraction only in cells expressing wild-type E-cadherin. -Actin was stained to prove equal loading of the blots. Representative assays out of three are shown

Full figure and legend (158K)

We next addressed the question, if the above-mentioned increase in - and -catenin results in assembly of E-cadherin/catenin adhesion complexes and their connection with the actin cytoskeleton. Cell fractionation by Triton X-100 was used to prove the attachment of the E-cadherin/catenin complex to the cytoskeleton. In Figure 2c, we demonstrate that significant amounts of wild-type E-cadherin, - and -catenin were localized in the Triton X-100-insoluble fraction, which indicates their association with the cytoskeleton. In contrast, in E-cadC-expressing cells, neither the truncated E-cadherin nor - or -catenin were connected with the cytoskeleton (Figure 2c). All proteins were exclusively localized in the Triton X-100-soluble fraction, indicating that these proteins were not linked to the cytoskeleton. Coimmunoprecipitation studies further supported this view because - and -catenin coprecipitated only with full-length E-cadherin (data not shown). Interestingly, immunoprecipitation of N-cadherin from lysates of the appropriate cell clone revealed low amounts of -catenin associated with N-cadherin, but virtually no -catenin was detectable (Figure 3a). Since it was described that N-cadherin can form cadherin/catenin adhesion complexes in mesenchymal cells (Hamaguchi et al., 1993), we performed coimmunoprecipitation experiments of L-cells as well as epithelial CHO cells transiently expressing N-cadherin. As demonstrated in Figure 3b, ectopically expressed N-cadherin was associated with - and -catenin in L-cells and also CHO but not in AsPC-1 cells, another pancreatic carcinoma cell line. These results may implicate that in pancreatic carcinoma cells, N-cadherin cannot substitute for E-cadherin in mediating cell–cell adhesion and - and -catenin expression. To examine if the regulatory protein p120^ctn was involved in the upregulation of - and -catenin expression, we expressed an E-cadherin protein with a mutated p120^ctn-binding site in MIA PaCa-2 cells. Figure 3c shows that cells expressing this mutated E-cadherinp120 did not exhibit an increase in - or -catenin concentrations compared to cells expressing wild-type E-cadherin, suggesting a role of p120^ctn in this process. To prove the expression level of the transfected cadherins, immunostaining of E- or N-cadherins was performed (Figure 3c).

Figure 3.

Analysis of catenins associated with N-cadherin. (a) Immunoprecipitations were performed with a N-cadherin antibody using 3 mg of total protein lysates. The coprecipitated - and -catenin was analysed by Western blotting. (b) To analyse if the N-cadherin can induce catenins in a cell type-dependent manner, we transfected N-cadherin in fibroblastic L-cells, epithelial CHO cells and pancreatic AsPC-1. - and -catenin coprecipitated with N-cadherin was analysed by Western blotting. (c) MIA PaCa-2 cells transfected with an E-cadherin, containing a mutated p120^ctn-binding site, failed to induce - and -catenin and behaved like N-cadherin. Immunoblots with 30 g of total lysate were stained for - and -catenin. Staining of E- or N-cadherin was performed to document equal amounts of transfected proteins. Restaining with anti -actin antibody served to document equal loading. Representative blots out of three experiments are shown

Full figure and legend (75K)

Expression of E-cadherin but not N-cadherin induces cellular aggregation and reduces invasiveness

To examine if the observed assembly of E-cadherin/catenin adhesion complexes contributed to a modified cellular behaviour of MIA PaCa-2 cells, we performed several assays such as aggregation, migration and proliferation. In cell aggregation assays, wild-type E-cadherin-expressing cells formed large, tight cell clusters, whereas the other cell clones formed small aggregates which were poorly associated. Cell aggregation was quantified by counting the number of aggregates and calculating an aggregation index. Mock-transfected MIA PaCa-2 had an aggregation index of A=0.12. N-cadherin-expressing MIA PaCa-2 cells revealed only a slightly increased aggregation index of A=0.29. The index of A=0.22 for E-cadC-expressing cells as demonstrated in Figure 4a indicates that these cells displayed hardly any aggregation. In contrast, wild-type E-cadherin-transfected cells exhibited a calculated aggregation index of A=0.85, pointing towards strong cell–cell adhesion. Removal of calcium ions by addition of EDTA/EGTA to wild-type E-cadherin-transfected cells decreased the aggregation index to A=0.23, which is comparable to the one for cells lacking E-cadherin. The application of an antibody, which inhibits homophilic interaction between E-cadherins (DECMA-1) or N-cadherins (GC-4), reduced the aggregation index to A=0.36 for E-cad WT- or A=0.18 for N-cad-expressing cells, respectively (Figure 4a). These results demonstrate clearly that the ectopically expressed E-cadherin was responsible for the observed cellular aggregation.

Figure 4.

Functional characteristics of E-cadherin-transfected MIA PaCa-2 cells. (a) Estimation of cell aggregation capacities. Addition of EDTA/EGTA confirmed that calcium-dependent adhesion molecules were involved and addition of functionally neutralizing cadherin antibodies (DECMA-1 for E-cadherin or GC-4 for N-cadherin) verified that the aggregation was E- or N-cadherin mediated. Mean values s.e.m. of three independent assays performed in quadruplicate are shown. (b) Cell migration was estimated using transwell migration assays. At 24 h after seeding, the number of migrated cells was estimated. (c) The invasive potential of MIA PaCa-2 cells expressing different cadherins was analysed by orthotopic transplantation of the cells into the pancreas of nude mice. After 4 weeks, the numbers of metastases in liver and lung were estimated macroscopically. Meanss.d. are shown of three independent experiments with two mice each

Full figure and legend (128K)

Next, we studied the influence of different cadherins on cell migration and proliferation. Using transwell migration assays, we demonstrate that increased cellular aggregation induced by expression of wild-type E-cadherin led to a reduced migration rate. As shown in Figure 4b, only a small number of E-cad WT-transfected cells migrated through the porous membrane 24 h after seeding (E-cad WT: 7.65.8 migrated cells per visual field). In contrast, N-cadherin-, E-cadC- as well as mock-transfected clones showed significantly higher numbers of migrated cells (approximately 10-fold more cells) and there was no obvious difference between these clones (Figure 4b). Cells expressing E-cadherin, which was unable to bind p120ctn, exhibited only a slightly reduced migration rate. These results were supported by wound-healing studies, in which wild-type E-cadherin-expressing cells failed to reduce a gap in contrast to mock-, E-cadC- or N-cadherin-transfected MIA PaCa-2 cells (data not shown).

Since cell proliferation can also be influenced by altered cell–cell adhesion, the proliferation rate of transfected MIA PaCa-2 cells was analysed by measuring [³H]thymidine incorporation. Neither full-length E-cadherin- nor N-cadherin- or E-cadC-containing MIA PaCa-2 cells exhibited significant alterations of DNA synthesis compared to vector-transfected controls (data not shown).

E-cadherin-expressing MIA PaCa-2 cells form less metastases in nude mice

In addition to the described cell culture-based assays, we examined the malignant properties of the different cell clones in vivo. Therefore, we transplanted the individual cell clones into the pancreas of nude mice. Examination of the surrounding organs 4 weeks later revealed that MIA PaCa-2 as well as mock-transfected cells formed numerous metastases in liver and lung (Figure 4c). Histological examination, as shown in Figure 5, demonstrated an aggressive infiltration of the pancreatic tissue by the tumour cells. Injection of cells expressing wild-type N-cadherin or the C-terminal deletion mutant of E-cadherin produced an equivalent number of metastases in adjacent organs and led to comparable tumour growth in the pancreatic tissue compared to controls. Although, the transplantation of full-length E-cadherin-expressing MIA PaCa-2 cells resulted in tumour growth, the cells grew as solid tumours without remarkable infiltration of the pancreatic tissue (Figure 5), indicating a reduced metastatic potential of E-cadherin-positive MIA PaCa-2 cells.

Figure 5.

Tumours formed by injected cell clones were histologically analysed. Semithin sections of tumours induced by transplanted control cells, wild-type E-cadherin-, E-cadC- as well as N-cadherin-expressing cells were H&E stained. Bar, 20 m

Full figure and legend (488K)

Distinct p120^ctn isoforms are associated with different cadherins

In order to address the question about the molecular mechanisms that contribute to the different abilities of E- and N-cadherin to recruit proteins into adhesion complexes, we expanded our investigations on p120^ctn. As shown in Figure 6a, immunostaining for p120^ctn in cell clones used here revealed that it was located at the membrane in all cadherin-expressing cell clones in contrast to the mock-transfected controls. The overall protein level of p120^ctn remained unchanged in cells expressing E- or N-cadherin as analysed by immunoblotting of total protein lysates (Figure 6b). Several tissue-specific isoforms of p120^ctn are described in the literature resulting from differential ATG-usage and alternative splicing (Yap et al., 1998). Considering these data, we studied which isoform of p120^ctn was associated with E- or N-cadherin by immunoprecipitation of cadherins from isolated cell membrane fractions.

Figure 6.

Cadherin-associated p120^ctn isoforms and their tyrosine phosphorylation. (a) Immunostaining of transfected MIA PaCa-2 cells with anti-p120^ctn antibody demonstrated its localization at the cell membrane. Bar, 30 m. (b) The overall concentration of p120^ctn and of the different isoforms was not changed in the analysed cell clones. In addition, the phosphorylation of p120^ctn isoforms was not significantly altered. (c) Isoforms and phosphorylation state of p120^ctn associated with transfected cadherins was analysed by Western blotting after immunoprecipitation of E- or N-cadherin. The phosphorylation status of p120^ctn was estimated by incubation with a phosphotyrosine-specific antibody. To correlate the resulting bands with p120^ctn, the blots were reprobed with an anti-p120^ctn antibody. Representative assays out of three independent experiments are shown

Full figure and legend (150K)

The results, shown in Figure 6c, demonstrate that N-cadherin was associated with a longer isoform of about 120 kDa, whereas E-cadherin was predominantly associated with a markedly shorter form of about 95 kDa, which was not found in N-cadherin protein complexes (Figure 6d). These data, in conjunction with the results from the aggregation studies, support the hypothesis that association of the shorter p120^ctn isoform with E-cadherin is correlated with strong cell–cell adhesion. To examine if the binding of different p120^ctn isoforms to E- or N-cadherin is specific, we coexpressed E- and N-cadherin in MIA PaCa-2 cells. In immunofluorescence analyses, we confirmed that most cells exhibited signals for both cadherins. As documented in Figure 7a, immunoprecipitation of E- or N-cadherin revealed that the expression of the additional cadherin did not alter the p120^ctn isoform associated with E- or N-cadherin. E-cadherin still bound the 95 kDa isoform and N-cadherin was associated with the 120 kDa variant of p120^ctn. The importance of p120^ctn binding to E-cadherin in cell–cell adhesion was further analysed in aggregation assays using MIA PaCa-2 cells transfected with the E-cadherin mutant lacking the p120^ctn-binding site. The results, shown in Figure 7b, document that cells expressing E-cadherin, which was not associated with p120^ctn, did not form large aggregates and behaved like mock- or N-cadherin-transfected cells.

Figure 7.

Simultaneous expression of E- and N-cadherin in MIA PaCa-2 cells did not change the differential isoforms associated with E- or N-cadherin (a), as shown after precipitation of E- or N-cadherin from transfected MIA PaCa-2 cells. A representative blot out of three is shown. (b) Expression of an E-cadherin mutant that lacks the p120ctn-binding site was not able to induce full cell aggregation as compared to wild-type E-cadherin-expressing Mia PaCa-2 cells. Means.d. of a typical assay is shown out of three independent experiments

Full figure and legend (146K)

E-cadherin-associated p120^ctn is not tyrosine phosphorylated

Beside the different isoforms of p120^ctn which influence cellular adhesion, it is known that the phosphorylation status of p120^ctn also modulates cell aggregation capacities (Roura et al., 1999). To analyse the p120^ctn phosphorylation in the presence of different cadherins, we examined the distribution and phosphorylation of p120^ctn associated with E- and N-cadherin in our cell clones. As shown in Figure 6b, neither the amount of p120^ctn isoforms, nor their phosphorylation differ significantly between E- or N-cadherin-expressing MIA PaCa-2 cells. Next, the tyrosine phosphorylation of p120^ctn, which was coprecipitated with E- or N-cadherin, was analysed by immnoblot experiments. As shown in Figure 6c, p120^ctn was clearly phosphorylated at tyrosine residues when associated with N-cadherin, which did not contribute to intercellular adhesion. However, p120^ctn coprecipitated with E-cadherin was barely phosphorylated (Figure 6d). Nonphosphorylated p120^ctn could be observed after transfection of full-length E-cadherin as well as after transfection of the deletion construct missing the -catenin-binding domain (E-cadC). These data indicate that not only the association of different p120^ctn isoforms with cadherin but also the phosphorylation status at tyrosine residues of p120^ctn can be correlated with cadherin function with regard to cellular adhesion.

Discussion

Considering the extremely high rate of metastases of tumours derived from the exocrine pancreas, we were interested in mechanisms leading to dysregulation of the E-cadherin complex in the development of pancreatic carcinoma. In this study, we show that N-cadherin cannot substitute for E-cadherin in mediating efficient cell–cell adhesion of pancreatic carcinoma cells. Although both analysed cadherins were present in the membrane, only re-expression of E-cadherin resulted in increased - and -catenin concentrations and the assembly of the cadherin adhesion complex. In addition, both cadherins showed different association with p120^ctn. Only E-cadherin was associated with the short and nonphosphorylated isoform of p120^ctn, whereas N-cadherin was coupled with the longer, highly phosphorylated isoform of p120^ctn.

A few studies have been carried out analysing the role of E-cadherin in pancreatic cancer, however, these studies focused on histological analyses of the E-cadherin/catenin complex in pancreatic tumour samples (Pignatelli et al., 1994; Weinel et al., 1996; Karayiannakis et al., 1998). In summary, these studies showed that only 32–60% of pancreatic tumours and their metastases showed defects in E-cadherin expression with a loss or reduction of E-cadherin protein and sometimes of catenins. In an elegant approach, Christofori and co-workers demonstrated the principle importance of E-cadherin loss in the transition from adenoma to invasive carcinoma in a model of pancreatic -cell carcinogenesis (Perl et al., 1998). To determine the importance of E-cadherin and N-cadherin for the assembly of cadherin/catenin complexes and cell–cell adhesion in the exocrine pancreas, we stably re-expressed E-cadherin in cells of an E-cadherin-deficient pancreatic carcinoma cell line. Most important, re-expression of E-cadherin resulted in increased expression of catenins and in re-assembly of the adherens junctions. These results are in agreement with the data published before by other groups characterizing E-cadherin as invasion suppressor (Frixen et al., 1991; Behrens et al., 1993; Lowy et al., 2002). Moreover, we and others have shown that re-expression of E-cadherin was sufficient to suppress metastatic events after orthotopic transplantation of transfected cells into the pancreas of nude mice (Vleminckx et al., 1991; Bruns et al., 1999). Cells expressing the E-cadC- as well as the E-cadp120-construct did not show elevated - and -catenin levels and showed no cellular aggregation. Similar results were shown by Nagafuchi and Takeichi (1988). In addition to earlier reports which showed that cadherin expression stabilized the protein levels of catenins, especially -catenin (Finnemann et al., 1997), we demonstrate that re-expression of E-cadherin was followed by upregulation of - and -catenin gene expression. The transcriptional regulators by which E-cadherin modulates catenin gene expression are presently unknown. Our data indicate that the -catenin- and the p120^ctn-binding site of E-cadherin is necessary to induce transcription of catenin genes, because both mutants of E-cadherin failed to do so. This led to the hypothesis that the presence of the whole complex or its association with the actin cytoskeleton contributes to the increased transcription of - and -catenin genes. This may be supplemented by an additional upregulation of - and -catenin on protein level (Nagafuchi et al., 1991; Papkoff, 1997).

The incorrect expression of nonepithelial cadherins by epithelial cells represent an additional mechanism for facilitating invasion and metastasis of tumour cells (Rasbridge et al., 1993; Hazan et al., 1997). Interaction between different classical cadherins can contribute to reduced cell adhesion and to metastasis. The mechanisms involved in this process are currently not understood (Christofori, 2003). Although, there is a high structural homology between the different classical cadherins, N-cadherin was not able to contribute to cellular adhesion in pancreatic MIA PaCa-2 and AsPC-1 cells. N-cadherin was predominantly localized in the cell membrane but there was only a negligible increase in - or -catenin concentration in contrast to the observed increase mediated by E-cadherin. In conclusion, N-cadherin cannot substitute for E-cadherin in the upregulation of catenins and in the mediation of tight cellular adhesion in pancreatic carcinoma cells. In agreement with our data, comparative analyses of different breast carcinoma cell lines revealed that neither the presence of P-cadherin nor N-cadherin was sufficient to induce an epithelial phenotype or to reduce cell motility (Nieman et al., 1999; Tran et al., 1999). In addition, the model of pancreatic -cell carcinogenesis revealed that the loss of E-cadherin was necessary for progression to carcinoma, whereas N-cadherin levels remained unchanged (Perl et al., 1998). Different studies analysed the interaction between VE-cadherin and N-cadherin mainly in endothelial cells (Navarro et al., 1998; Jaggi et al., 2002) In this model, N-cadherin did not seem to inhibit VE-cadherin-mediated cellular adhesion. Although both cadherins were associated with - and -catenin, VE-cadherin was able to exclude N-cadherin from cell–cell junctions. Navarro et al. in 1998 showed that a cytoplasmic region different from the -catenin-binding site was necessary to induce N-cadherin exclusion.

Concerning the mechanisms which allow E-cadherin but not N-cadherin to mediate cell adhesion, we studied the influence of p120^ctn on cadherin-complex assembly. Several recent studies underline the importance of p120^ctn for cell adhesion (Kintner, 1992; Yap et al., 1998) as well as regulation of cell motility (Chen et al., 1997). There is growing evidence that p120^ctn can influence lateral clustering of cadherin molecules and the assembly of cadherin adhesion complexes itself or its connection with the actin cytoskeleton. However, the data concerning the consequences of p120^ctn association with cadherins are controversial. Thus, Yap et al. (1998) reported that binding of p120^ctn to E-cadherin supported cadherin clustering and promoted strong cell–cell adhesion. Thoreson et al. (2000) showed that selective uncoupling of p120^ctn by mutation of the E-cadherin p120^ctn-binding site disrupted strong adhesion in cultured cells. In addition, Ireton et al. (2002) provided evidence that re-expression of p120^ctn increased E-cadherin concentration and induced a mesenchymal–epithelial transition in colon cancer cells. In agreement with the results shown in this study, Ireton and co-workers described that phosphorylation of p120^ctn interfered with strong cell adhesion. In contrast to these data, which suggest a positive influence of p120^ctn, other groups described that the binding of p120^ctn to E-cadherin could regulate cell adhesion in a negative manner (Aono et al., 1999; Ohkubo and Ozawa, 1999). Ohkubo and Ozawa (1999) reported that prevention of p120^ctn binding to E-cadherin activated E-cadherin and stimulated aggregation of epithelial cells. Therefore, it seems that the juxtamembrane domain of cadherins and associated p120^ctn modulate cell adhesion in a positive or negative manner depending on the cellular situation.

We could demonstrate that p120^ctn indeed binds ectopically expressed E- and N-cadherin. However, the finding that only E-cadherin was connected with the cytoskeleton led to the conclusion that the presence or association of p120^ctn with cadherin alone is not sufficient to induce strong cell–cell aggregation. One explanation may be the phosphorylation state of p120^ctn which has been reported to play a fundamental role in the regulation of the E-cadherin adhesion complex (Shibamoto et al., 1995; Keilhack et al., 2000). In agreement with our results that p120^ctn was tyrosine phosphorylated when it was associated with the nonadhesive N-cadherin, recent observations showed that tyrosine-phosphorylated p120^ctn destabilized the E-cadherin/catenin adhesion complex (Ozawa and Ohkubo, 2001). The authors demonstrated that the cellular kinase src phosphorylated p120^ctn and as a consequence led to reduced cellular aggregation. These data point towards a link between src-induced p120^ctn phosphorylation and regulation of the cadherin adhesion complex assembly. Additionally, it was reported that different isoforms of p120^ctn differentially influenced cellular adhesion (Montonen et al., 2001; Piedra et al., 2003; reviewed in Anastasiadis and Reynolds, 2000). The longer splice variants (1A and 2A of 120 or 110 kDa, respectively), which are predominantly present in mesenchymal cells, were correlated with reduced cellular adhesion and promotion of cell migration (Mo and Reynolds, 1996; Thoreson et al., 2000). Our data support this view, because N-cadherin, which did not contribute to cell–cell adhesion, was predominantly associated with the longer isoforms of p120^ctn. In contrast, the smaller isoform of p120^ctn of about 95 kDa was associated with E-cadherin in wild-type E-cadherin-expressing cells, which exhibited enhanced cell aggregation and reduced invasive capacities in vitro and in vivo. Although accumulating data exist that p120^ctn isoforms and its phosphorylation are involved in the regulation of cadherin-mediated cellular adhesion, we cannot exclude that jet unidentified changes in - or -catenin in MIA PaCa2-cells or additional regulatory proteins contribute to the effects observed in this study. Other authors suggest that despite the high degree of similarity between the cytoplasmic tail of different classical cadherins, these region exhibits different functional properties in E- or N-cadherin. In another study, it has been shown that a deletion of 11 amino acids in the proximal region of E-cadherin was able to induce cell–cell aggregation in contrast to an exchange of three amino acids in this region (Ozawa, 2003). Ozawa assumed from these data that other proteins binding close to the p120ctn-binding site of E-cadherin may be responsible for regulation of E-cadherin-induced cellular adhesion.

The molecular mechanisms by which different p120^ctn isoforms control E-cadherin/catenin interaction are currently unknown and need to be further analysed. It is conceivable that p120^ctn physically stabilizes the E-cadherin complex, competitively blocks interaction of cadherins with other proteins or can regulate its function by modulation of Rho-GTPases. It was recently shown that p120^ctn can inhibit RhoA and activate Rac1 or Cdc42, small GTPases of the Rho-family and well-known modulators of the actin cytoskeleton (Tapon and Hall, 1997; Grosheva et al., 2001; Yap and Kovacs, 2003). These GTPases coordinate the organization of the actin cytoskeleton and may contribute to anchoring the E-cadherin complex and promote E-cadherin clustering (Braga et al., 1997; Jou and Nelson, 1998). Moreover, our data indicate that different p120^ctn isoforms associated with E- or N-cadherin can be correlated with different expression of - and -catenin. This may be related to the recently described interaction of p120^ctn with the transcription factor Kaiso, thereby opening a new field of p120^ctn activities (Daniel and Reynolds, 1999). It was speculated that the smaller p120^ctn isoforms binds Kaiso more efficiently than the longer isoform (Daniel and Reynolds, 1999). In addition, Aho et al. showed that overexpression of the smaller p120^ctn isoforms in HaCaT cells downregulates the expression of E-cadherin (Aho et al., 2002). These findings indicate that p120^ctn isoforms may directly or indirectly influence the expression of genes, which are involved in cell–cell adhesion.

In conclusion, in this study, we show that E-cadherin re-expression in a pancreatic carcinoma cell line led to re-establishment of a complete cadherin/catenin complex. The formation of the cadherin/catenin complex resulted in recovery of strong cell–cell adhesion and reduction of invasion. Furthermore, we demonstrated that N-cadherin did not substitute for E-cadherin in pancreatic epithelial cells with regard to cellular aggregation. The different behaviour of classical cadherins was associated with binding of different isoforms of p120^ctn. The long variant was preferentially associated with nonadhesive N-cadherin, whereas adhesive E-cadherin was coupled to the shorter p120^ctn isoform 3. In addition, the p120^ctn phosphorylation state at tyrosine residues played a pivotal role in the regulation of E-cadherin-mediated cellular adhesion. These data imply that p120^ctn has both positive and negative effects on cadherin-mediated adhesion and thus contributes to the modulation of E-cadherin-mediated adhesion.

Materials and methods

Plasmids, antibodies and cell culture

Human full-length cDNA and a C-terminal-truncated cDNA of E-cadherin were kindly provided by D Rimm (Yale University, CT, USA), mouse wild-type N-cadherin cDNA was kindly provided by M Takeichi (Kyoto University, Japan). The cDNAs were subcloned into pcDNA3.1 expression vector (Invitrogen, Breda, The Netherlands). An E-cadherin mutation within the p120^ctn-binding domain (replacement of amino acids 764–766 by AAA according to Thoreson et al., 2000) was generated using an in vitro mutagenesis kit (Stratagene). The correct sequence of all constructs was verified by sequencing (GATC, Konstanz, Germany). Monoclonal antibodies against E- (C20820), N-cadherin (C70320), -catenin (C21620), -catenin (C19220), -catenin (C26220), occludin (O79120), p120^ctn (P17920) and phosphotyrosine (PY20, P11625) were obtained from BD Bioscience (San Diego, CA, USA). Antibodies against pan-cadherin (C-3678), - (C-2081) and -catenin (C-2206) and vinculin (V-9131) were purchased from Sigma Biochemicals (Taufkirchen, Germany).

MIA PaCa-2 cells (CRL 1420) were obtained from the American Type Culture Collection (Manassas, VA, USA). For stable protein expression, MIA PaCa-2 cells were transfected using 15 g DNA per 100 mm dish and DMRIE-C reagent according to the manufacture's protocol (Gibco Invitrogen, Karlsruhe, Germany). For coexpression of E- and N-cadherin, 7 g of each plasmid was used. MIA PaCa-2 cells transfected with the empty vector (pcDNA3.1) served as controls. Stably transfected cell clones exhibited not less than 80% cadherin-positive cells as verified by immunfluorescence analysis. For each construct, three cell clones were analysed, whereof one representative clone is shown in the results.

Western blot and immunoprecipitation

Polyacrylamide gel electrophoresis (SDS–PAGE) was performed according to standard procedures as described previously (Menke et al., 2001). For total protein analysis, cells were lysed in NOP buffer (10 mM Tris-HCl pH 7.8, 150 mM NaCl, 1 mM MgCl₂, 1 mM CaCl₂, 1% Nonidet P40) containing proteinase inhibitors (5 M aprotinin, 1 mM pefabloc, 0.1 mM, 0.01% soja trypsin inhibitor (STI), 10 M pepstatin, 10 g/ml leupeptin). Immunoreactive proteins were localized with a second horseradish peroxidase-coupled antibody (Pierce, Rockford, IL, USA) and visualized using enhanced chemiluminescence (Pierce).

For immunocoprecipitation, 0.3–5 mg of NOP lysates were used. The antigen–Ig complexes were precipitated by incubation with protein-G-sepharose. Immunoprecipitates were analysed by Western blotting as described above.

Triton X-100 and subcellular fractionation

To obtain Triton X-100-soluble and -insoluble fractions, cells were incubated with Triton buffer (1% Triton X-100, 0.3 M sucrose, 25 mM HEPES pH 7.4, 100 mM NaCl, 4.7 mM KCl, 1.2 mM KH₂PO₄, 1.2 mM MgCl₂, 5 M aprotinin, 1 mM pefabloc, 0.01% STI, 10 M pepstatin, 10 g/ml leupeptin) for 15 min on a rocking platform. After centrifugation, the supernatant (Triton X-100-soluble fraction) was collected. The cell pellet was resuspended in SDS lysis buffer (20 mM Tris pH 7.5, 2.5 mM EDTA, 1%SDS, 0.5 g/ml DNase I, 1 mM pefabloc, 0.01% STI, 5 M aprotinin). Both fractions were reconstituted to equal volumes. For immunoblot analyses, 15 l of each fraction were analysed by SDS–PAGE. Membrane and cytosolic lysates were made by using a hypotonic buffer as described in Thoreson et al. (2000) with little modifications (10 mM Tris-HCl pH 7.4, 1 mM MgCl₂, 5 M aprotinin, 1 mM pefabloc, 0.1 mM STI, 10 M pepstatin, 10 g/ml leupeptin). For immunoblot analyses, 15 g of membrane fractions together with equal volumes of the corresponding cytosolic fractions were applied. For immunoprecipitation experiments, 1 mg of membrane fraction was used.

Immunocytochemical analysis

To study protein localization, cells were grown on glass slides. Cells were fixed and permeabilized in cold acetone–methanol. Primary antibodies were visualized by incubation with a secondary Cy-3- (Biomol, Hamburg, Germany) or Alexa-488-conjugated antibody (Molecular Probes, Eugene, OR, USA). DNA was stained with DAPI (Serva, Heidelberg, Germany). The staining was examined by fluorescence microscopy (Axiophot, Zeiss, Oberkochen, Germany) and confocal laser microscopy (TCS-4, Leica, Wetzlar, Germany).

Northern-blot analysis

RNA was obtained using the RNeasy extraction kit (Qiagen, Hilden, Germany). In total, 20 g of total RNA were analysed per lane. Northern blotting and hybridization was carried out as described previously (Menke et al., 1997).

Migration, proliferation and aggregation assay

Migration assays were performed using cell culture inserts with porous membranes (8.0 m pore size, BD Bioscience, Heidelberg, Germany) as described in Giehl et al. (2000). After 12, 24 and 48 h of incubation, the number of migrated cells was estimated by counting five independent visual fields in microscope using a 20 objective. Three independent assays were performed in quadruplicates. For wound-healing migration assays, cells were grown to confluence and wounding was performed by scraping a small split. To examine the wound-healing motility, the reduction of the cell-free split was photographed and measured at 0, 24 and 48 h monitoring the same area each time. The reduction of the cell-free split was expressed in per cent of the distance at 0 h. Five areas were measured in three independent assays. Cell proliferation was determined by measuring DNA synthesis by [methyl-³H]thymidine incorporation as described in Giehl et al. (2000). To determine the Ca²⁺-dependent cell–cell adhesion, aggregation assays were performed as described by Ozawa and Kemler (1990). The extent of cell aggregation was calculated by the formula A=(N_o-N_e)/N_o, with N_o representing the total particle number at the start and N_e being the total particle number after incubation for 20 min. To examine the calcium dependency, EDTA and EGTA were added to a final concentration of 5 mM each. E- or N-cadherin dependence was proved by adding a neutralizing antibody against E- or N-cadherin (DECMA-1 or GC-4, respectively, Sigma Biochemicals) at a final concentration of 4 g/ml assay medium. To remove preservatives of the antisera, the solutions were replaced by TBS using centricon centrifugation columns (Filtron, Northborough, MA, USA). Three independent assays were performed in quadruplicate.

Tumour growth in nude mice

To determine the metastatic potential, 100 l of a single-cell suspension containing 2 10⁶ cells were injected into the pancreas of narcotized NMRI nu-mice. After 4 weeks, the animals were killed, the tumour size in the pancreas was estimated macroscopically and the numbers of metastases in surrounding organs were counted. The pancreas, liver and lung were excised and fixed in 4% formaldehyde. For histological examination, semithin sections were stained with haematoxylin–eosin (H&E). Serial sections were stained with anti-E-cadherin antibody as described before. Three to five animals were used for each cell clone. The animal care protocol and the experimental design was approved by a governmental animal care review committee.

Statistics

The mean values and standard errors were calculated out of at least three experiments, each performed in triplicate or quadruplicate. For statistical analysis, the Wilcoxon nonparametric test was used and P<0.02 was considered significant.

References

1. Aho S, Levansuo L, Montonen O, Kari C, Rodeck U and Uitto J. (2002). J. Cell Sci., 115, 1391–1402. | PubMed | ISI | ChemPort |
2. Anastasiadis PZ and Reynolds AB. (2000). J. Cell Sci., 113, 1319–1334. | PubMed | ISI | ChemPort |
3. Aono S, Nakagawa S, Reynolds AB and Takeichi M. (1999). J. Cell Biol., 145, 551–562. | Article | PubMed | ISI | ChemPort |
4. Behrens J, Vakaet L, Friis R, Winterhagen E, Van Roy F, Mareel M and Birchmeier W. (1993). J. Cell Biol., 120, 757–766. | Article | PubMed | ISI | ChemPort |
5. Birchmeier C, Birchmeier W and Brand Saberi B. (1996). Acta Anat. Basel., 156, 217–226. | PubMed |
6. Braga VM, Machesky LM, Hall A and Hotchin NA. (1997). J. Cell Biol., 137, 1421–1431. | Article | PubMed | ISI | ChemPort |
7. Bruns CJ, Harbison MT, Kuniyasu H, Eue I and Fidler IJ. (1999). Neoplasia, 1, 50–62. | Article | PubMed | ChemPort |
8. Chen H, Paradies NE, Fedor-Chaiken M and Brackenbury R. (1997). J. Cell Sci., 110 (Part 3), 345–356. | PubMed | ISI | ChemPort |
9. Christofori G. (2003). EMBO J, 22, 2318. | Article | PubMed | ISI | ChemPort |
10. Daniel JM and Reynolds AB. (1999). Mol. Cell. Biol., 19, 3614–3623. | PubMed | ISI | ChemPort |
11. Finnemann S, Mitrik I, Hess M, Otto G and Wedlich D. (1997). J. Biol. Chem., 272, 11856–11862. | Article | PubMed |
12. Frixen UH, Behrens J, Sachs M, Eberle G, Voss B, Warda A, Löchner D and Birchmeier W. (1991). J. Cell Biol., 113, 173–185. | Article | PubMed | ISI | ChemPort |
13. Giehl K, Skripczynski B, Mansard A, Menke A and Gierschik P. (2000). Oncogene, 19, 2930–2943. | Article | PubMed | ISI | ChemPort |
14. Grosheva I, Shtutman M, Elbaum M and Bershadsky AD. (2001). J. Cell Sci., 114, 695–707. | PubMed | ISI | ChemPort |
15. Gumbiner BM. (1996). Cell, 84, 345–357. | Article | PubMed | ISI | ChemPort |
16. Gumbiner BM. (2000). J. Cell Biol., 148, 399–404. | Article | PubMed | ISI | ChemPort |
17. Hamaguchi M, Matsuyoshi N, Ohnishi Y, Gotoh B, Takeichi M and Nagai Y. (1993). EMBO J., 12, 307–314. | PubMed | ISI | ChemPort |
18. Hazan RB, Kang L, Whooley BP and Borgen PI. (1997). Cell Adhes Commun., 4, 399–411. | PubMed | ISI | ChemPort |
19. Herrenknecht K, Ozawa M, Eckerskorn C, Lottspeich F, Lenter M and Kemler R. (1991). Proc. Natl. Acad. Sci. USA, 88, 9156–9160. | Article | PubMed | ChemPort |
20. Ireton RC, Davis MA, van Hengel J, Mariner DJ, Barnes K, Thoreson MA, Anastasiadis PZ, Matrisian L, Bundy LM, Sealy L, Gilbert B, van Roy F and Reynolds AB. (2002). J. Cell Biol., 159, 465–476. | Article | PubMed | ISI | ChemPort |
21. Jaggi M, Wheelock MJ and Johnson KR. (2002). Cell Commun. Adhes., 9, 103–115. | Article | PubMed | ISI | ChemPort |
22. Jou TS and Nelson WJ. (1998). J. Cell Biol., 142, 85–100. | Article | PubMed | ISI | ChemPort |
23. Karayiannakis AJ, Syrigos KN, Chatzigianni E, Papanikolaou S, Alexiou D, Kalahanis N, Rosenberg T and Bastounis E. (1998). Anticancer Res., 18, 4177–4180. | PubMed | ISI | ChemPort |
24. Karayiannakis AJ, Syrigos KN, Polychronidis A and Simopoulos C. (2001). Anticancer Res., 21, 4127–4134. | PubMed |
25. Keilhack H, Hellman U, van Hengel J, van Roy F, Godovac-Zimmermann J and Bohmer FD. (2000). J. Biol. Chem., 275, 26376–26384. | Article | PubMed |
26. Kemler R. (1993). Trends Genet, 9, 317–321. | Article | PubMed | ISI | ChemPort |
27. Kern S, Hruban R, Hollingsworth MA, Brand R, Adrian TE, Jaffee E and Tempero MA. (2001). Cancer Res., 61, 4923–4932. | PubMed | ISI | ChemPort |
28. Kintner C. (1992). Cell, 69, 225–236. | Article | PubMed | ISI | ChemPort |
29. Lowy AM, Knight J and Groden J. (2002). Surgery, 132, 141–148. | Article | PubMed | ISI |
30. Mareel M, Boterberg T, Noe V, Van Hoorde L, Vermeulen S, Bruyneel E and Bracke M. (1997). J. Cell Physiol., 173, 271–274. | Article | PubMed | ISI | ChemPort |
31. Menke A, Philippi C, Vogelmann R, Seidel B, Lutz MP, Adler G and Wedlich D. (2001). Cancer Res., 61, 3508–3517. | PubMed | ISI | ChemPort |
32. Menke A, Yamaguchi H, Gress TM and Adler G. (1997). Gastroenterology, 113, 295–303. | Article | PubMed | ISI | ChemPort |
33. Mo YY and Reynolds AB. (1996). Cancer Res., 56, 2633–2640. | PubMed |
34. Montonen O, Aho M, Uitto J and Aho S. (2001). J. Histochem. Cytochem., 49, 1487–1496. | PubMed |
35. Nagafuchi A and Takeichi M. (1988). EMBO J., 7, 3679–3684. | PubMed | ISI | ChemPort |
36. Nagafuchi A, Takeichi M and Tsukita S. (1991). Cell, 65, 849–857. | Article | PubMed | ISI | ChemPort |
37. Navarro P, Ruco L and Dejana E. (1998). J. Cell Biol., 140, 1475–1484. | Article | PubMed | ISI | ChemPort |
38. Nieman MT, Prudoff RS, Johnson KR and Wheelock MJ. (1999). J. Cell Biol., 147, 631–643. | Article | PubMed | ISI | ChemPort |
39. Ohkubo T and Ozawa M. (1999). J. Biol. Chem., 274, 21409–21415. | Article | PubMed | ISI | ChemPort |
40. Ozawa M. (2003). J. Biol. Chem., 278, 46014–46020. | Article | PubMed | ISI | ChemPort |
41. Ozawa M and Kemler R. (1990). J. Cell Biol., 111, 1645–1650. | Article | PubMed | ISI | ChemPort |
42. Ozawa M and Kemler R. (1998). J. Cell Biol., 142, 1605–1613. | Article | PubMed | ISI | ChemPort |
43. Ozawa M and Ohkubo T. (2001). J. Cell Sci., 114, 503–512. | PubMed | ISI | ChemPort |
44. Papkoff J. (1997). J. Biol. Chem., 272, 4536–4543. | PubMed | ISI | ChemPort |
45. Perl AK, Wilgenbus P, Dahl U, Semb H and Christofori G. (1998). Nature, 392, 190–193. | Article | PubMed | ISI | ChemPort |
46. Piedra J, Miravet S, Castano J, Palmer HG, Heisterkamp N, Garcia de Herreros A and Dunach M. (2003). Mol. Cell. Biol., 23, 2287–2297. | Article | PubMed | ISI | ChemPort |
47. Pignatelli M, Ansari TW, Gunter P, Liu D, Hirano S, Takeichi M, Klöppel G and Lemoine NR. (1994). J. Pathol., 174, 243–248. | Article | PubMed | ISI | ChemPort |
48. Poston GJ, Gillespie J and Guillou PJ. (1991). Gut, 32, 800–812. | PubMed |
49. Rasbridge SA, Gillett CE, Sampson SA, Walsh FS and Millis RR. (1993). J. Pathol., 169, 245–250. | Article | PubMed | ISI | ChemPort |
50. Rimm DL, Koslov ER, Kebriaei P, Cianci CD and Morrow JS. (1995). Proc. Natl. Acad. Sci. USA, 92, 8813–8817. | Article | PubMed | ChemPort |
51. Roura S, Miravet S, Piedra J, de Herreros AG and Dunach M. (1999). J. Biol. Chem., 274, 36734–36740. | Article | PubMed | ISI | ChemPort |
52. Schipper JH, Frixen UH, Behrens J, Unger A, Jahnke K and Birchmeier W. (1991). Cancer Res., 51, 6328–6337. | PubMed | ISI | ChemPort |
53. Shibamoto S, Hayakawa M, Takeuchi K, Hori T, Miyazawa K, Kitamura N, Johnson KR, Wheelock MJ, Matsuyoshi N, Takeichi M and Ito F. (1995). J. Cell Biol., 128, 949–957. | Article | PubMed | ISI | ChemPort |
54. Tapon N and Hall A. (1997). Curr. Opin. Cell. Biol., 9, 86–92. | Article | PubMed | ISI | ChemPort |
55. Thoreson MA, Anastasiadis PZ, Daniel JM, Ireton RC, Wheelock MJ, Johnson KR, Hummingbird DK and Reynolds AB. (2000). J. Cell Biol., 148, 189–202. | Article | PubMed | ISI | ChemPort |
56. Tran NL, Nagle RB, Cress AE and Heimark RL. (1999). Am. J. Pathol., 155, 787–798. | PubMed | ISI | ChemPort |
57. Vleminckx K, Vakaet Jr L, Mareel M, Fiers W and Van Roy F. (1991). Cell, 66, 107–119. | Article | PubMed | ISI | ChemPort |
58. Weinel RJ, Neumann K, Kisker O and Rosendahl A. (1996). Int. J. Pancreatol., 19, 25–30. | PubMed |
59. Yap AS and Kovacs EM. (2003). J. Cell Biol., 160, 11–16. | Article | PubMed | ISI | ChemPort |
60. Yap AS, Niessen CM and Gumbiner BM. (1998). J. Cell Biol., 141, 779–789. | Article | PubMed | ISI | ChemPort |

Acknowledgements

We thank Klaudia Giehl, Yukiko Imamichi and Alexander Koenig for helpful discussions and Felicitas Genze for excellent technical assistance. This study is supported by the Deutsche Forschungsgemeinschaft (SFB 518).