Introduction

E-cadherin is an important adhesion molecule that is downregulated in embryogenic cells which undergo epithelial–mesenchymal transition (EMT). During this process, epithelial cell layers lose polarity and cell–cell contacts and undergo dramatic remodeling of the cytoskeleton, which is important for the correct development of the embryo (Thiery, 2002). The formerly epithelial cells acquire the expression of mesenchymal components after passing through EMT, and get a migratory phenotype. As a consequence of the mesenchymal phenotype, the adhesion molecule N-cadherin is expressed. Such a dramatic change in cadherin expression is not just a phenomenon in embryonic development, as EMT is an important event in the progression of many carcinomas. Here, loss of E-cadherin can be due to transcriptional regulation (Poser et al., 2001), promoter methylation (Grady et al., 2000; Melki et al., 2000) or mutations (Berx et al., 1995; Tamura et al., 1996; Guilford et al., 1998), as shown, for example, in hereditary diffuse gastric cancer (HDGC) where carriers of the CDH1 germline gene mutation develop an aggressive, diffuse, submucosal gastric cancer (Guilford et al., 1998).

The cadherin switch is also found in malignant melanoma development (Silye et al., 1998; Sanders et al., 1999; Poser et al., 2001). Normal melanocytes and the surrounding keratinocytes express E-cadherin for stable tissue assembly (Hsu et al., 1996; Furukawa et al., 1997). After the switch to N-cadherin expression, the cells acquire a migratory phenotype.

The homophilic cadherin contact of the melanocytes is mediated through the extracellular domain of cadherins. Additionally, several studies were focused upon the conserved cytoplasmic domain of cadherins which interact with intracellular proteins, termed catenins (Ozawa et al., 1990; Ozawa and Kemler, 1992, 1998; Stappert and Kemler, 1994). The catenins connect the cytoplasmic tail of cadherin to the actin cytoskeleton of the cells (Sommers et al., 1994; Knudsen et al., 1995; Rimm et al., 1995).

The biological activity maintained through cadherins can be indirect, for example, via signaling through small GTPases Rho, Rac, and Cdc42, which in turn link E-cadherin to a number of downstream signaling pathways (Fukata et al., 1999; Kaibuchi et al., 1999). Additionally, cadherin function is regulated through receptor tyrosine kinases (RTKs), which phosphorylate -catenin (Soler et al., 1998). In particular, E-cadherin has been proposed to account for tumor suppressor activity; however, the exact molecular mechanism has not been clearly established (Behrens et al., 1991; Schipper et al., 1991; Shiozaki et al., 1991; Sommers et al., 1991; Rasbridge et al., 1993).

The idea that adhesion alone may not be sufficient to suppress invasion is supported by the observations that enhanced adhesion does not always lead to reduced invasion (Navarro et al., 1993; Sommers et al., 1994). These findings suggest that the inhibition of invasion by E-cadherin may also occur through effects other than adhesion. Results of Wong and Gumbiner (2003) indicate that intracellular signaling, not adhesion, mediated E-cadherin's tumor suppressor function in breast and prostate cancer. They show that the extracellular domain of E-cadherin was neither necessary nor sufficient to stop invasive behavior.

Restoration of E-cadherin-mediated interaction in melanoma represents a legitimate strategy to reverse melanocytic malignancy (Herlyn et al., 2000; Li and Herlyn, 2000). As a consequence, our group is not only interested in cadherins as adhesion molecule but also in investigating new signaling pathways at the plasma membrane, accomplishing juxtacrine signaling events (Fagotto and Gumbiner, 1996; Buckley and Simmons, 1997; Hazan et al., 2000; Hsu et al., 2000a; Zhu and Watt, 1996).

As a mediator for E-cadherin signaling, NFB was analysed. NFB is a heterodimeric transcription factor that is predominantly composed of subunits of the Rel family. In resting cells, NFB is mainly retained in the cytoplasm by the IB family of proteins which mask the nuclear translocation signal of the transcription factor. Upon stimulation, IB proteins are phosphorylated, triggering ubiquitination and subsequent degradation by proteasomes. Hereby, the NFB proteins are released for translocation to the nucleus and induction of B-dependent genes. NFB modulates multiple basic cellular functions like cell growth, differentiation, inflammatory response, and immune response. More recently, NFB activation has been connected with multiple aspects of oncogenesis, including the control of apoptosis, differentiation, and cell migration (Baldwin, 2001). New publications reveal the fact that NFB is also required for the induction and maintenance of EMT in cancer models (Huber et al., 2004a, 2004b).

Recently, we could show that loss of E-cadherin in melanoma cells leads to induction of NFB activity and that re-expression of E-cadherin results in downregulation of NFB activity in melanoma cells (Kuphal et al., 2004). New findings presented here show that NFB activity also regulates further EMT processes in malignant melanoma development. Hence, we are interested in EMT signaling events affected through NFB.

Results

In most types of cancer of epithelial origin, E-cadherin-mediated cell–cell adhesion is lost concomitant with progression towards malignancy. By the loss of E-cadherin in melanoma development, cells escape the control of neighboring keratinocytes. In melanoma cells, de novo synthesis of N-cadherin has been observed (Li and Herlyn, 2000; Tomita et al., 2000).

Re-establishing the functional E-cadherin complex, for example, by re-expression of full-length E-cadherin, results in a reversion from an invasive, mesenchymal, to a benign, epithelial phenotype of cultured tumor cells (Birchmeier and Behrens, 1994; Vleminckx et al., 1991).

Influence of full-length E-cadherin on endogenous N-cadherin expression

The current study was designed to elucidate the influence of E-cadherin on the expression of N-cadherin and to analyse the molecular basis of this effect.

Quantitative real-time PCR of nine melanoma cell lines revealed that melanoma cells express N-cadherin strongly in a constitutive manner. The N-cadherin expression level was analysed in comparison to normal human epidermal melanocytes (NHEMs, which were set as 1), showing no N-cadherin expression (Figure 1a). Additionally, the E-cadherin expression level of the nine melanoma cell lines was analysed by quantitative real-time PCR. In comparison to NHEMs, the melanoma cell lines express almost no E-cadherin (Figure 1b).

Figure 1.

Quantitative real-time PCR to measure expression levels of N-cadherin and E-cadherin in nine melanoma cell lines. Quantitative real-time PCR, Western blot analysis, and promoter analysis were performed for transiently E-cadherin-transfected Mel Im cells. (a) Quantitative real-time PCR was performed to measure relative expression levels of N-cadherin in nine different melanoma cell lines. The melanoma cell lines are presented in comparison to NHEM showing no N-cadherin expression. NHEM was set as 1. (b) Quantitative real-time PCR was performed to measure the relative expression levels of E-cadherin in nine different melanoma cell lines. The melanoma cell lines are presented in comparison to NHEM set as 100. Two different charges of isolated melanocytes were used. (c) Quantitative real-time PCR was performed to analyse the relative expression level of N-cadherin after transfection of Mel Im cells with different concentrations of full-length E-cadherin. Mock-transfected cells were set as 1. The E-cadherin-transfected Mel Im cells show reduced N-cadherin expression in a dose-dependent manner. (d) Western blot analysis was used to detect N-cadherin level in the melanoma cell line Mel Im (mock) compared to Mel Im cells which were transiently transfected with full-length E-cadherin (0.5 g E-cadherin). The E-cadherin re-expressing cells show reduced N-cadherin protein level. -Actin was used to control for loading of equal protein amounts. (e) Reporter gene assays using a N-cadherin-luc promoter construct were performed to detect the N-cadherin promoter activity in Mel Im cells (mock) in comparison to full-length E-cadherin-transfected Mel Im cells (0.5 g E-cadherin). N-cadherin promoter activity of mock-transfected cells was set as 1. After transfection of E-cadherin, the N-cadherin promoter activity was downregulated.

Full figure and legend (127K)

Here, we first wanted to determine the effect of re-expression of E-cadherin on endogenous N-cadherin expression in melanoma cells. For this, transfection experiments with full-length E-cadherin constructs (Figure 2a) were performed. These experiments try to imitate the E-cadherin expression of healthy melanocytes. The result showed that, in transiently full-length E-cadherin-transfected melanoma cells, N-cadherin expression was found to be downregulated in a dose-dependent manner. This was revealed by quantitative real-time PCR and by Western blot analysis of mock-transfected cells in comparison to full-length E-cadherin-transfected Mel Im cells (Figure 1c and d). Additionally, analysis of N-cadherin promoter activity revealed a clear downregulation of N-cadherin gene transcription (Figure 1e). These first experiments led to the hypothesis that full-length E-cadherin can regulate the endogenous expression level of N-cadherin in melanoma cells.

Figure 2.

Schematic overview of the E-cadherin derivative constructs used. (a) The domain structure of full-length E-cadherin is shown. (b) Schematic overview of the E-cadherin-catenin construct with a deleted -catenin-binding site. (c) Schematic overview showing the E-cadherin CD construct. (d) Schematic overview showing the membrane (M)-anchored CD of E-cadherin (located at the plasma membrane through the lipid tag). TM, transmembrane domain; C, cytoplasmic domain; N, extracellular domain; black box, catenin-binding site.

Full figure and legend (54K)

Influence of the cytoplasmic domain of E-cadherin on N-cadherin expression

To separate the extracellular cell–cell adhesive functions of E-cadherin from intracellular adhesion to the cytoskeleton and signaling, two types of E-cadherin derivatives were used for the following experiments. Both have a deleted extracellular domain, thus averting cadherin–cadherin self-association. One construct used was the cytoplasmic domain of E-cadherin, which results in translation of a soluble fragment within the cytoplasm (Figure 2c). In the second construct, the intracellular domain of E-cadherin was membrane anchored by a myristoylation signal added to the cadherin intracellular domain (Figure 2d).

We established stable transfected cell clones for each cytoplasmic E-cadherin derivative in the melanoma cell line Mel Im. The stable transfected cell clones were controlled for E-cadherin expression by quantitative real-time PCR with specific primer against the intracellular domain of E-cadherin (Figure 3a and c), and in case of the membrane-anchored E-cadherin additionally with specific primer against the myristyolation signal (Figure 3b). Translation of the 20.2 kDa membrane-bound E-cadherin cytoplasmic domain and the 18.8 kDa soluble cytoplasmic E-cadherin was shown in Western blots (Figure 3d and e). Three stable transfected Mel Im cell clones for each cytoplasmic E-cadherin construct were chosen: E-cadherin CDc, E-cadherin CDd, E-cadherin CDe for expression of the soluble cytoplasmic domain and E-cadherin M2, E-cadherin M3, E-cadherin M7 for expression of the membrane-anchored cytoplasmic domain. The soluble cytoplasmic cadherin was only detectable in the cytoplasmic fraction of the cellular extract, whereas the membrane-anchored cytoplasmic E-cadherin was detectable in the membrane and the cytoplasmic fraction (Figure 3d and e), as previously shown by Nieman et al. (1999).

Figure 3.

Quantitative real-time PCR, Western blot analysis to prove expression of the CD of E-cadherin in the generated cell clones and functional analyses of the cell clones. (a, b) Quantitative real-time PCR using primers against the cytoplasmic domain of E-cadherin (a) and the myristoylation signal of the transmembrane construct (b) was performed to verify expression of the stably transfected membrane-anchored cytoplasmic E-cadherin construct. The designed stable cell clones were named E-cadherin M2, M3, and M7. (c) Quantitative real-time PCR using primers against the soluble CD of E-cadherin was performed to check for efficient expression. The soluble CD could be detected in the cell clones, named E-cadherin CDc, CDd, and CDe. (d, e) Extracts of the membrane fraction and the cytoplasmic fraction of stably transfected cells were used in Western blots to test if the E-cadherin derivatives are targeted to their respective location. (d) The soluble CD of E-cadherin (CDc, CDd, CDe) is located in the cytoplasm. (e) The membrane-anchored CD of E-cadherin (M2, M3, M7) is located at the plasma membrane and in the cytoplasm. (f, g) Invasion (left) and migration (right) of the E-cadherin M2, M3, and M7 cell clones (f) and the E-cadherin CDc, CDd, and CDe cell clones (g) were analysed in a Boyden Chamber model in comparison to mock-transfected Mel Im cells set as 100%.

Full figure and legend (195K)

To analyse the functional importance of cytoplasmic E-cadherin, we performed Boyden chamber experiments for migration and invasion of the stable transfected melanoma cells in comparison to mock-transfected melanoma cells. The experiments revealed reduced capability of the cell clones E-cadherin cytoplasmatic domain (CD) and E-cadherin M to migrate and invade the artificial basement membrane (Figure 3f and g). Proliferation of the cell clones was not influenced through cytoplasmic E-cadherin (data not shown).

In subsequent experiments, the endogenous level of N-cadherin in the stable transfected cell clones was determined. The E-cadherin derivative transfected cells M and CD showed a decrease in expression of endogenous N-cadherin on mRNA and protein level (Figure 4a–c), as revealed before for the full-length E-cadherin transfected cells. The N-cadherin promoter activity was measured by reporter gene assays after transfection of the cytoplasmic membrane-anchored and cytoplasmic soluble domain of E-cadherin. The N-cadherin promoter activity was downregulated after transfection of the derivatives (Figure 4d). These experiments revealed the effect of the cytoplasmic domain of E-cadherin on endogenous N-cadherin expression of melanoma cells.

Figure 4.

Quantitative real-time PCR and Western blot analysis were performed to measure the expression level of N-cadherin in Mel Im cells expressing the CD of E-cadherin. Additionally, reporter gene analysis using the N-cadherin promoter was conducted to detect its activity after transient transfection of the CD of E-cadherin in Mel Im cells. (a, b) The results shown were generated by quantitative real-time PCR for N-cadherin expression of stably transfected Mel Im cells. The cell clones E-cadherin M2, M3, and M7 (a) and CDc, CDd, and CDe (b) show downregulated N-cadherin expression in comparison to mock (empty vector)-transfected cells. (c) Western blot analysis showing the reduced N-cadherin level in Mel Im cells stably transfected with the soluble (CD) and membrane-anchored (M) CD of E-cadherin. Mock cells are transfected with an empty vector. (d) Reporter gene analysis was performed analysing a N-cadherin-luc promoter construct in the Mel Im cells transiently transfected with the soluble and membrane-anchored domain of E-cadherin. The N-cadherin promoter activity was downregulated in comparison to cells transfected with the empty vector (mock).

Full figure and legend (99K)

NFB as mediator for the cadherin switch in melanoma development

The search for a mediator of cadherin switch in melanoma cells resulted in the analysis of NFB.

In a recent study, we could show that nuclear extracts from malignant melanoma cells displayed strong NFB activity compared to melanocytes. Accordingly, electrophoretic mobility shift assay (EMSA) data showed that NFB DNA-binding activities were higher in malignant melanoma cell lines than in melanocytes. Furthermore, after re-expression of E-cadherin in melanoma cells, NFB activity was downregulated (Kuphal et al., 2004). These findings result in the new hypothesis that NFB could be the master regulator of N-cadherin in dependence on E-cadherin expression in the cells.

We, therefore, compared the N-cadherin promoter sequences of chicken and human and searched for putative NFB-binding motifs using the search engine Genomatix/MatInspector (www.genomatix.de). The NFB-binding sequences of the chicken and the human N-cadherin promoter are displayed in Figure 5a; both are about 460 nucleotides upstream of the translation start of N-cadherin. EMSA experiments were performed to prove NFB DNA binding. The nuclear extract of Mel Im cells displayed strong NFB-binding activity to a consensus sequence of NFB in the putative N-cadherin promoter of chicken (Figure 5b). The DNA binding was competable with two different concentrations of the classical NFB consensus sequence (5'-AGT TGA GGG GAC TTT CCC AGG C-3') (Figure 5b). Additionally, we detect NFB-binding activity to the human N-cadherin promoter in EMSA experiments (data not shown).

Figure 5.

EMSA of the binding capacity of nuclear NFB to the N-cadherin promoter sequence. (a) Schematic overview of the classical NFB-binding motif (top) and the binding sites of NFB in the chicken (middle) and human (bottom) N-cadherin promoter sequence (R: A or G; Y: C or T; H: A, C, or T; N: any nucleotide). (b) The EMSA analysis showed the binding capacity of nuclear extracts of Mel Im cells to the NFB-binding motifs (NFB site I, NFB site II) found in the chicken N-cadherin promoter sequence. Additionally, the shifts were competed with two different concentrations of unlabeled classical NFB consensus sequence to prove NFB binding.

Full figure and legend (96K)

After transfection of full-length E-cadherin and of the CDs and M, the relative NFB DNA-binding activity to its consensus sequence was measured in reporter gene assays (Figure 6a and b). All led to downregulation of NFB activity. Comparison to data shown in Figures 1e and 4d revealed that E-cadherin re-expression not only led to downregulation of NFB DNA-binding capacity but also to downregulation of N-cadherin promoter activity.

Figure 6.

Reporter gene analysis to quantify the NFB DNA-binding activity and the N-cadherin promoter activity. (a) Reporter gene analysis was performed to quantify the NFB DNA-binding activity after transfection of full-length E-cadherin. The NFB DNA-binding activity was downregulated after re-expression of full-length E-cadherin in Mel Im cells. Mel Im cells transfected with the vector pCDNA3 (mock) were set as 1. (b) Additionally, the NFB DNA-binding activity was checked for the soluble (CD) and membrane (M)-anchored CD of E-cadherin using an NFB-luc construct. The NFB activity was downregulated after re-expression of the cytoplasmic E-cadherin in Mel Im cells. Mel Im cells transfected with the vector pCDNA3 (mock) were set as 1. (c) Reporter gene analysis to quantify the N-cadherin promoter activity after transfection of the NFB subunits p50 and p65 of the Rel family into the melanoma cell lines Mel Im and Mel Ei and into the colon carcinoma cell lines SW480 and CaCo. The mock transfection with pCDNA3 was set as 1.

Full figure and legend (65K)

To verify the regulatory function of NFB on N-cadherin, the effect of NFB overexpression on N-cadherin promoter activity was analysed. We, therefore, transfected the NFB subunits p50 and p65 and the N-cadherin promoter construct transiently into two melanoma cell lines (Mel Im and Mel Ei) and, to prove cell type independence, into the colon carcinoma cell lines SW480 and CaCo. In all four cell lines, upregulation of N-cadherin promoter activity in comparison to mock-transfected cells was clearly seen (Figure 6c).

To further analyse the interplay between NFB signaling and N-cadherin as a target, we made adenoviral transduction experiments with Mel Im and Mel Ei melanoma cell lines. The transduction of the IB super-repressor (Ad5IB) provides a nondegradable IB, which leads to a cytoplasmic arrest of p65 and to an interrupted NFB signaling cascade (Hellerbrand et al., 1998). A -galactosidase-expressing adenoviral vector (Ad5LacZ) was used as a control. As a control for successful viral infection and inhibition of NFB activity, expression of the endogenous NFB-dependent gene interleukin 8 (IL-8) was measured by quantitative real-time PCR. The expression of IL-8 was reduced in the melanoma cells transduced with stable IB (data not shown). Further, expression of the NFB-independent gene MMP-1 was not influenced through the expression of the super-repressor IB (data not shown). After expression of stable IB and inhibition of NFB signaling, the N-cadherin expression was quantified and it was ascertained that N-cadherin expression was downregulated in comparison to -galactosidase-expressing cells after 24 and 48 h (Figure 7a). Additionally, the N-cadherin protein level of the Mel Im cells transduced with the super-repressor IB was analysed. Western blot analyses also revealed downregulation of N-cadherin in virus-transduced cells in comparison to the control (Figure 7b).

Figure 7.

Quantitative real-time PCR and Western blot analysis for the relative expression level of N-cadherin after inhibition of NFB signaling. (a) Cells were transduced with the control adenoviral vector (Ad5LacZ) containing the E. coli -galactosidase cDNA and the adenoviral vector Ad5IB, which leads to the expression of a super-repressor IB. The expression of the stable IB leads to blockade of NFB signaling. Transduced were the melanoma cell lines Mel Im and Mel Ei for 24 and 48 h. The expression level of N-cadherin in these cells was quantified by real-time PCR and the Ad5LacZ-transduced cells were set as 1. The expression level of N-cadherin was downregulated after blocking NFB. (b) The protein level of N-cadherin in Mel Im cells transduced with the adenoviral vector Ad5IB was shown in comparison to Mel Im cells transduced with the vector Ad5LacZ by Western blot. Like in (a) the N-cadherin expression level was downregulated after blocking NFB. (c) Treatment of Mel Im cells with the proteasome inhibitor MG-132 which reduces the degradation of ubiquitin-conjugated proteins (like IB) led to inhibition of NFB activation. After this treatment for 4 and 8 h, relative N-cadherin expression was determined by quantitative real-time PCR. Untreated cells were set as 1. (d, e) NHEMs were treated with LPS and TNF to stimulate NFB activity. The stimulated NFB activity led to the upregulation of N-cadherin expression (d) and IL-8 (e) in comparison to untreated melanocytes.

Full figure and legend (191K)

Additionally, NFB activity was inhibited by the treatment of Mel Im cells with the proteasome inhibitor MG-132, which blocks the degradation of IB and the nuclear translocation of p65 NFB. After the blockade of NFB signaling for 4 and 8 h, the N-cadherin expression was downregulated (Figure 7c).

It was recently published that E-cadherin regulates NFB activity (Kuphal et al., 2004). We, therefore, concluded that E-cadherin regulates NFB activity and proximate N-cadherin expression was directly regulated by NFB.

Vice versa, the stimulation of NFB activity with LPS and TNF in NHEM led to upregulation of N-cadherin expression measured by quantitative real-time PCR. NHEM treated with TNF showed a six-time upregulation of N-cadherin expression (Figure 7d). Quantification of the expression level of the endogenous NFB-dependent gene IL-8 in a control experiment showed the upregulation of the expression level through NFB (Figure 7e).

Taken together, the results revealed that full-length E-cadherin and the cytoplasmic fragments of E-cadherin downregulate the endogenous N-cadherin expression of melanoma cells through downregulation of NFB activity and the diminished influence of NFB to the N-cadherin promoter.

Influence of the cytoplasmic domain of N-cadherin on N-cadherin expression

The cytoplasmic domain of E-cadherin is very similar to the cytoplasmic domain of N-cadherin. The two adhesion molecules share up to 60% identity. The next experiments were designed to see if expression of cytoplasmic N-cadherin has an effect on endogenous N-cadherin expression in melanoma cells. We used two types of N-cadherin derivatives: the cytoplasmic domain of N-cadherin, which results in a soluble fragment within the cytoplasm, and the cytoplasmic domain of N-cadherin membrane anchored by the myristyolation signal.

It was important to differentiate between endogenous and transfected N-cadherin in melanoma cells. Quantitative real-time PCR using primer against the extracellular domain of N-cadherin was performed to detect the influence on endogenous N-cadherin. The result revealed that transfection of the cytoplasmic domain of N-cadherin influences expression of endogenous N-cadherin negatively similar to the cytoplasmic E-cadherin domain (data not shown). Additionally, we tested if the cytoplasmic domain of N-cadherin influences the NFB DNA-binding capacity and the N-cadherin promoter activity (Figure 8a and b). Like the cytoplasmic domain of E-cadherin, the cytoplasmic domain of N-cadherin downregulated both.

Figure 8.

Analysis of NFB DNA-binding and N-cadherin promoter activity after transient transfection of cytoplasmic N-cadherin fragments, and of the role of the -catenin-binding site of E-cadherin in the regulation of N-cadherin promoter activity. (a) Reporter gene analysis was performed to quantify the NFB DNA-binding activity after transfection of soluble cytoplasmic (CD) and membrane-bound cytoplasmic (M) N-cadherin. The NFB DNA-binding activity was downregulated after overexpression of cytoplasmic N-cadherin in Mel Im cells. Mel Im cells transfected with the vector pCDNA3 (mock) were set as 1. (b) Reporter gene analysis was performed, analysing a N-cadherin-luc promoter construct in the Mel Im cells transiently transfected with two different concentrations of the soluble (CD) and membrane-anchored domain (M) of N-cadherin. The N-cadherin promoter activity was downregulated in comparison to cells transfected with the empty vector (mock). (c) Reporter gene analysis was performed, analysing a N-cadherin-luc promoter construct in the Mel Im cells transiently transfected with three different concentrations of an E-cadherin construct with deleted -catenin-binding site (E-cadherin-catenin). Without sequestering -catenin, E-cadherin has no capacity for downregulating N-cadherin promoter activity.

Full figure and legend (62K)

This finding could be due to the homology of the cytoplasmic domains of E-cadherin and N-cadherin, leading to same effects on N-cadherin expression. We hypothesize that this effect was due to the -catenin-binding sites and -catenin sequestering through both cadherins.

To identify the region within the intracellular E-cadherin domain that is responsible for the signaling effect and to check if the -catenin-binding site of E-cadherin is involved in the regulation of endogenous N-cadherin of the melanoma cells, we used an E-cadherin construct with deleted -catenin-binding site (E-cadherin-catenin). The E-cadherin-catenin is not capable of downregulating endogenous N-cadherin expression (Figure 8c), leading to the suggestion that -catenin binding is involved in the regulatory effect of E-cadherin.

Discussion

Here, we were interested to find factors regulating the endogenous N-cadherin expression in melanoma cells during EMT and development and progression of the disease. We have shown in this study that full-length E-cadherin, but also the cytoplasmic domain of E-cadherin, can regulate expression of endogenous N-cadherin in melanoma cells, and that NFB is the mediator for E-cadherin signaling.

Previous studies have demonstrated the functional importance of cadherin molecules in normal skin homeostasis and in melanoma development (Hsu et al., 2000b; Li and Herlyn, 2000; Li et al., 2001). It has been observed that the function of E-cadherin during tumor progression is frequently replaced or even overruled by the expression of mesenchymal cadherins, such as N-cadherin (reviewed by Cavallaro et al., 2002). Re-establishing the functional cadherin complex, for example, by forced expression of full-length E-cadherin, results in a reversion from an invasive, mesenchymal, to a benign, epithelial phenotype of cultured tumor cells (Vleminckx et al., 1991; Birchmeier and Behrens, 1994). However, the impulse of the cadherin switch in malignant melanoma remains to be elucidated.

Our data demonstrate for the first time that the re-expression of the cytoplasmic domain of E-cadherin led to downregulation of endogenous N-cadherin expression in malignant melanoma cells. Additionally, this regulatory effect of cytoplasmic E-cadherin was achieved by downregulation of N-cadherin transcription.

Particularly, the cytoplasmic domain of E-cadherin was able to accomplish signaling effects identical to the full-length E-cadherin. Our data implicated that it is not important whether the cytoplasmic domain is attached to the cell membrane or is soluble in the cytoplasm. This is in contrast to a publication on human epidermoid carcinoma cells (A431), which demonstrated that expression of intracellular cadherin domains lead to decreased cell–cell interactions only when these are membrane-associated (Nieman et al., 1999).

The loss of E-cadherin in malignant melanoma is in particular regulated by the zinc-finger transcriptional repressor Snail (Batlle et al., 2000; Comijn et al., 2001; Poser et al., 2001). The functional block of E-cadherin in melanocytes by an antibody did not lead to re-expression of N-cadherin in these cells. Additionally, stable transfected antisense N-cadherin cell clones showed no re-expression of E-cadherin, suggesting that the sole loss of one cadherin subtype does not lead necessarily to expression of another new cadherin subtype.

Previously, other reports demonstrated that Twist, a master regulator of embryonic morphogenesis, is capable of inducing EMT and is associated with a migratory behavior in normal epithelial cells. Additionally, Twist plays an essential role in cancer metastasis. MDCK cells expressing Twist show N-cadherin upregulated and E-cadherin downregulated (Kang and Massague, 2004). Our experiments led initially to the impression that E-cadherin could influence Twist expression as NHEM and full-length E-cadherin re-expressing melanoma cells showed downregulated Twist expression in comparison to different melanoma cell lines which have high Twist expression. This correlates with constitutively upregulated N-cadherin expression detected in melanoma cells. However, the stably cytoplasmic E-cadherin-expressing cell clones showed no downregulated Twist expression, although they showed regulation of N-cadherin. In agreement with our results, the publication of Yang et al. (2004) indicated that E-cadherin does not influence Twist expression.

NFB was discussed as a transcription factor responsible for epithelial plasticity and EMT (Bates et al., 2004; Huber et al., 2004a, 2004b). Inhibition of NFB signaling prevented EMT in Ras-transformed epithelial breast cancer cells, while activation of this pathway promoted the transition to a mesenchymal phenotype. We found that the molecule E-cadherin downregulates NFB after re-expression in melanoma cells (Kuphal et al., 2004). The participation of E-cadherin in NFB regulation was supported by the experiments shown in this study.

We, therefore, questioned if NFB inactivation can influence endogenous N-cadherin expression level in melanoma cells and if NFB can participate in the switch of the cadherin class during melanoma development. After comparison of N-cadherin promoter sequences of chicken and human and finding of putative NFB-binding sites, EMSA data revealed NFB binding to the promoter sequence of the N-cadherin gene. Here, we demonstrate for the first time to our knowledge that N-cadherin exhibits an NFB-binding site in its promoter and its expression is directly regulated by NFB. Further experiments revealed that NFB was the mediator of E-cadherin-regulated N-cadherin expression in melanoma cells.

Evidence that NFB directly regulates the N-cadherin promoter was confirmed by transfection of NFB subunits p50 and p65 into the melanoma cell lines Mel Im and Mel Ei, and the colon carcinoma cell lines SW480 and CaCo, resulting in downregulation of N-cadherin promoter activity. Infection of the melanoma cells with the super-repressor IB (Ad5IB), which blocked the NFB translocation into the nucleus, resulted in reduced expression of N-cadherin (Figure 7a and b). Additionally, the proteasome inhibitor MG-132 led to downregulation of N-cadherin expression after NFB translocation into the nucleus was blocked. Vice versa, the stimulation of NFB activity in normal melanocytes through treatment of the cells with LPS and TNF led to the upregulation of N-cadherin expression in these cells.

We, therefore, conclude that E-cadherin negatively controls the activity of NFB and, as a consequence, N-cadherin expression in melanocytic cells. Loss of expression of E-cadherin in malignant melanoma cells results in upregulated NFB activity, leading to upregulation of the EMT marker N-cadherin. Normally, melanoma cells have a high NFB activity and a high endogenous N-cadherin expression rate (Kuphal et al., 2004).

We could show that the adhesive properties of E-cadherin are not important for E-cadherin signaling into the cell. The intracellular domain of E-cadherin is responsible for signaling to NFB and the N-cadherin promoter. Additionally, the results of Wong and Gumbiner (2003) indicate that intracellular signaling, not adhesion, mediated E-cadherin's tumor suppressor function in breast and prostate cancer. They showed that the extracellular domain of E-cadherin was neither necessary nor sufficient to stop invasive behavior.

To analyse if the cytoplasmic domain of N-cadherin has an influence on the endogenous N-cadherin expression of melanoma cells, we transfected the cell line Mel Im transiently with two N-cadherin constructs which are analogous to the E-cadherin cytoplasmic constructs. Like cytoplasmic E-cadherin, the cytoplasmic N-cadherin downregulates N-cadherin promoter and NFB DNA-binding activity. We hypothesize that the homology of the cytoplasmic domains of cadherins has the same regulating effect. This regulating effect is due to similarities of the -catenin-binding site of cadherins. Using an E-cadherin construct with a deleted -catenin-binding site, we could prove that sequestering of -catenin through the cytoplasmic domain of E-cadherin is important for downregulating the N-cadherin promoter activity. The transfection of the E-cadherin construct with the deleted -catenin-binding sequence cannot regulate N-cadherin promoter activity.

Boyden chamber experiments which analyse the importance of cytoplasmic E-cadherin for functionally aspects of cell behavior support the involvement of cytoplasmic E-cadherin in migration and invasion of the cell. The transfection of E-cadherin into melanoma cells leads to diminished migration and invasion potential of the formerly aggressive melanoma cells through the imitated basement membrane.

In summary, healthy melanocytes express high amounts of E-cadherin, which sequester -catenin. Here, NFB is not activated and no N-cadherin is expressed. Melanoma cells loss E-cadherin, which leads to upregulated NFB activity stimulating N-cadherin promoter activity, resulting in high levels of N-cadherin. Re-expression of E-cadherin in melanoma cells, therefore, leads to negative effects of E-cadherin on NFB activity, resulting in downregulation of the N-cadherin promoter activity and, as a consequence, in downregulation of N-cadherin gene transcription.

There is a suggestive evidence that experimental re-establishment of E-cadherin expression, specially the expression of the cytoplasmic domain of E-cadherin, could be therapeutically beneficial, as overexpression of E-cadherin reduces growth, invasion, and survival of melanoma cells (Haass et al., 2004).

Materials and methods

Cell lines and cell culture conditions

Further, melanoma cell lines Mel Im, Mel Ju, Mel Wei, Mel Juso, Mel Ho, Mel Ei, HTZ-19d, and SK-Mel 28 were used (Jacob et al., 1998). For tissue culture, the cells were maintained in DMEM supplemented with penicillin (400 U/ml), streptomycin (50 g/ml), L-glutamine (300 g/ml) and 10% fetal calf serum (FCS; Sigma, Deisenhofen, Germany) and splitted 1:5 every 3 days.

Human primary melanocytes derived from normal skin were cultivated in melanocyte medium MGM-3 (Gibco, Eggenstein, Germany) under a humidified atmosphere of 5% CO₂ at 37°C. Cells were used between passages 2 and 5.

For further experiments, the cells were treated with 2.5 g/ml of the proteasome inhibititor MG-132 (Sigma) for 4 and 8 h.

Migration and invasion assays were performed using Boyden Chambers containing polycarbonate filters with 8 m pore size (Costar, Bodenheim, Germany) coated with gelatine or Matrigel (diluted 1:3 in H₂O; Becton Dickinson, Heidelberg, Germany), respectively, essentially as described previously (Kuphal et al., 2004). The lower compartment was filled with fibroblast-conditioned medium as chemoattractant.

Molecular constructs

The cadherin constructs E-cadherin CD, E-cadherin M (membrane-anchored domain), N-cadherin CD and N-cadherin M (membrane-anchored domain) were generous gifts from Margaret J Wheelock and Keith R Johnson (Department of Biology, University of Toledo, USA).

To target the cadherin intracellular domain to the plasma membrane, E-cadherin M (Figure 2) and N-cadherin M were constructed by ligating the cytoplasmic domain of E-cadherin and N-cadherin to the myristoylation signal from the vector M (Aronheim et al., 1994). The myristoylation signal (MGSSKSKPKDPSQR) derived from src was added to the cadherin intracellular domain. The N-cadherin constructs are homolog to the E-cadherin cytoplasmic constructs and are generous gifts from Keith R Johnson and Margaret J Wheelock.

The E-cadherin-catenin construct was a generous gift from the group of Gumbiner (Wong and Gumbiner, 2003).

Transfection experiments and luciferase measurements

For transient transfections, 2 10⁵ cells were seeded into each well of a six-well plate and transfected with 0.5 g plasmid DNA using the lipofectamine plus method (Gibco). The cells were lysed 24 h after transfection and luciferase activity in the lysate was quantified (Promega Corp., Madison, USA). Transfection efficiency was normalized according to renilla luciferase activity by cotransfecting 0.1 g of the plasmid pRL-TK (Promega). All transfections were repeated at least four times. For transient transfection, the plasmids NFB-luc (Promega), pBAT-E-cadherin (generous gift from Gabriele Handschuh, Munich, Germany) and p50/p65 (generous gift from JA Schmid, Vienna, Austria) were used. Additionally, the N-cadherin chicken promoter-luc construct, first cloned by Li et al. (1997) and subcloned from Panda et al. (2001), was a generous gift from David Goltzman, (Montreal, Canada).

Further, a panel of Mel Im cell clones were established by stable transfection of pCMX E-cadherin CD and pCMX E-cadherin M under the control of a CMV promoter and cotransfected with the neo-selectable pcDNA3 plasmid (Invitrogen, Groningen, Holland). Mock-transfected cells are the Mel Im cells with the empty vector. Transfected cells were cultured under selective conditions using G418 (Sigma) in a concentration of 50 g/ml. After 25 days of selection, individual G418-resistant colonies were subcloned. E-cadherin expression levels of these clones were determined by Western blot analysis of the cell lysates.

Adenoviral vectors and infection of melanoma cells

The recombinant replication-deficient adenovirus Ad5IB expressing IB S32A/S36A, a super-repressor of NFB, was constructed by the method described previously (Hellerbrand et al., 1998). Ad5lacZ containing the Escherichia coli -galactosidase (-Gal) gene was used as a control virus throughout the study. Mel Im and Mel Ei cells were infected with Ad5IB or Ad5lacZ in serum-free medium at a multiplicity of infection (m.o.i.) of 100. The experiment was stopped after 24 and 48 h.

Western blotting

3 10⁶ cells were lysed in 200 l RIPA buffer (Roche, Mannheim, Germany). In all, 10 g of RIPA-cell lysate was loaded per lane, separated on 10% SDS–PAGE gels and subsequently blotted onto PVDF membrane. After blocking with 3% BSA/PBS, the membrane was incubated for 16 h with anti-E-cadherin, anti-N-cadherin (BD Transduction Laboratories), or anti--actin (Sigma) antibody. After washing with PBS, the membrane was incubated with an alkaline phosphate-coupled secondary anti-mouse or anti-rabbit antibody (AP303A, Chemicon, Hofheim, Germany). Finally, immunoreactions were visualized by NBT/BCIP (Sigma) staining.

Electrophoretic mobility shift assays

A double-stranded oligomeric binding site for NFB, specific for the human N-cadherin promoter (5'-GCTCTTGGGGAGCGCCATCCGCTC-3'), was phospholabeled. Additionally, two double-stranded oligomeric binding sites for NFB, specific for the chicken N-cadherin promoter (5'-GGAGAGGGGCGGGCACGATCCGGGC-3' and (5'-GCGTGCAGCAGGGCTGGGGCCGGGGC-3'), were phospholabeled and used for gel mobility shift assays. A double-stranded oligomeric binding site for the classical NFB (5'-AGTTGAGGGGACTTTCCCAGGC-3', Promega) was used for competition experiments.

Nuclear extracts were prepared from primary melanocytes and Mel Im cells and gel shifts were performed as described previously (Bosserhoff et al., 1996).

Quantitative real-time PCR for N-cadherin and E-cadherin

Quantitative real-time PCR was performed on a Lightcycler (Roche, Mannheim, Germany). In all, 2 l cDNA template, 2 l 25 mM MgCl₂, 0.5 l (20 mM) of forward and reverse primers and 2 l of SybrGreen LightCycler Mix in a total volume of 20 l were applied to the following PCR programs: N-cadherin CD/E-cadherin CD: 95°C for 30 s (initial denaturation); 20°C/s transition rate up to 95°C for 15 s, 60°C for 3 s, 72°C for 5 s, 85°C acquisition mode single, repeated for 40 times (amplification). The PCR reaction was evaluated by melting curve analysis and checking the PCR products on 1.8% agarose gels. -Actin was the control gene for the quality of the mRNA/cDNA which was used in the quantitative real-time PCR. -Actin primers were used for standardization of the concentration of the cDNA.

Subcellular proteome extraction kit

Protein isolation for membrane, cytoplasm, and nuclear fractions was performed using the extraction kit from Calbiochem as described by the manufacturer (Calbiochem, Darmstadt, Germany).

References

1. Aronheim A, Engelberg D, Li N, al-Alawi N, Schlessinger J & Karin M. (1994) Cell 78: 949–961. | Article | PubMed | ISI | ChemPort |
2. Baldwin AS. (2001) J Clin Invest 107: 241–246. | PubMed | ISI | ChemPort |
3. Bates RC, DeLeo MJ, III & Mercurio AM. (2004) Exp Cell Res 299: 315–324. | Article | PubMed | ISI | ChemPort |
4. Batlle E, Sancho E, Franci C, Dominguez D, Monfar M & Baulida J et al.. (2000) Nat Cell Biol 2: 84–89. | Article | PubMed | ISI | ChemPort |
5. Behrens J, Weidner KM, Frixen UH, Schipper JH, Sachs M & Arakaki N et al.. (1991) EXS 59: 109–126. | PubMed | ChemPort |
6. Berx G, Cleton-Jansen AM, Nollet F, de Leeuw WJ, van de Vijver M & Cornelisse C et al.. (1995) EMBO J 14: 6107–6115. | PubMed | ISI | ChemPort |
7. Birchmeier W & Behrens J. (1994) Biochim Biophys Acta 1198: 11–26. | Article | PubMed | ISI | ChemPort |
8. Bosserhoff AK, Hein R, Bogdahn U & Buettner R. (1996) J Biol Chem 271: 490–495. | Article | PubMed | ISI | ChemPort |
9. Buckley CD & Simmons DL. (1997) Mol Med Today 3: 449–456. | Article | PubMed | ISI | ChemPort |
10. Cavallaro U, Schaffhauser B & Christofori G. (2002) Cancer Lett 176: 123–128. | Article | PubMed | ISI | ChemPort |
11. Comijn J, Berx G, Vermassen P, Verschueren K, van Grunsven L & Bruyneel E et al.. (2001) Mol Cell 7: 1267–1278. | Article | PubMed | ISI | ChemPort |
12. Fagotto F & Gumbiner BM. (1996) Dev Biol 180: 445–454. | Article | PubMed | ISI | ChemPort |
13. Fukata M, Nakagawa M, Kuroda S & Kaibuchi K. (1999) J Cell Sci 112: 4491–4500. | PubMed | ISI | ChemPort |
14. Furukawa F, Fujii K, Horiguchi Y, Matsuyoshi N, Fujita M & Toda K et al.. (1997) Microsc Res Tech 38: 343–352. | Article | PubMed | ISI | ChemPort |
15. Grady WM, Willis J, Guilford PJ, Dunbier AK, Toro TT & Lynch H et al.. (2000) Nat Genet 26: 16–17. | Article | PubMed | ISI | ChemPort |
16. Guilford P, Hopkins J, Harraway J, McLeod M, McLeod N & Harawira P et al.. (1998) Nature 392: 402–405. | Article | PubMed | ISI | ChemPort |
17. Haass NK, Smalley KS & Herlyn M. (2004) J Mol Histol 35: 309–318. | Article | PubMed | ISI | ChemPort |
18. Hazan RB, Phillips GR, Qiao RF, Norton L & Aaronson SA. (2000) J Cell Biol 148: 779–790. | Article | PubMed | ISI | ChemPort |
19. Hellerbrand C, Jobin C, Iimuro Y, Licato L, Sartor RB & Brenner DA. (1998) Hepatology 27: 1285–1295. | Article | PubMed | ISI | ChemPort |
20. Herlyn M, Berking C, Li G & Satyamoorthy K. (2000) Melanoma Res 10: 303–312. | Article | PubMed | ISI | ChemPort |
21. Hsu M, Andl T, Li G, Meinkoth JL & Herlyn M. (2000a) J Cell Sci 113: 1535–1542. | PubMed | ISI | ChemPort |
22. Hsu MY, Meier FE, Nesbit M, Hsu JY, Van Belle P & Elder DE et al.. (2000b) Am J Pathol 156: 1515–1525. | PubMed | ISI | ChemPort |
23. Hsu MY, Wheelock MJ, Johnson KR & Herlyn M. (1996) J Investig Dermatol Symp Proc 1: 188–194. | PubMed | ChemPort |
24. Huber MA, Azoitei N, Baumann B, Grunert S, Sommer A & Pehamberger H et al.. (2004a) J Clin Invest 114: 569–581. | Article | PubMed | ISI | ChemPort |
25. Huber MA, Beug H & Wirth T. (2004b) Cell Cycle 3: 1477–1480. | ISI |
26. Jacob K, Wach F, Holzapfel U, Hein R, Lengyel E & Buettner R et al.. (1998) Melanoma Res 8: 211–219. | PubMed | ISI | ChemPort |
27. Kaibuchi K, Kuroda S, Fukata M & Nakagawa M. (1999) Curr Opin Cell Biol 11: 591–596. | Article | PubMed | ISI | ChemPort |
28. Kang Y & Massague J. (2004) Cell 118: 277–279. | Article | PubMed | ISI | ChemPort |
29. Knudsen KA, Soler AP, Johnson KR & Wheelock MJ. (1995) J Cell Biol 130: 67–77. | Article | PubMed | ISI | ChemPort |
30. Kuphal S, Poser I, Jobin C, Hellerbrand C & Bosserhoff AK. (2004) Oncogene 23: 8509–8519. | Article | PubMed | ISI | ChemPort |
31. Li B, Paradies NE & Brackenbury RW. (1997) Gene 191: 7–13. | Article | PubMed | ISI | ChemPort |
32. Li G & Herlyn M. (2000) Mol Med Today 6: 163–169. | Article | PubMed | ISI | ChemPort |
33. Li G, Satyamoorthy K & Herlyn M. (2001) Cancer Res 61: 3819–3825. | PubMed | ISI | ChemPort |
34. Melki JR, Vincent PC, Brown RD & Clark SJ. (2000) Blood 95: 3208–3213. | PubMed | ISI | ChemPort |
35. Navarro P, Lozano E & Cano A. (1993) J Cell Sci 105: 923–934. | PubMed | ISI | ChemPort |
36. Nieman MT, Kim JB, Johnson KR & Wheelock MJ. (1999) J Cell Sci 112: 1621–1632. | PubMed | ISI | ChemPort |
37. Ozawa M & Kemler R. (1992) J Cell Biol 116: 989–996. | Article | PubMed | ISI | ChemPort |
38. Ozawa M & Kemler R. (1998) J Biol Chem 273: 6166–6170. | Article | PubMed | ISI | ChemPort |
39. Ozawa M, Ringwald M & Kemler R. (1990) Proc Natl Acad Sci USA 87: 4246–4250. | PubMed | ChemPort |
40. Panda DK, Miao D, Lefebvre V, Hendy GN & Goltzman D. (2001) J Biol Chem 276: 41229–41236. | Article | PubMed | ISI | ChemPort |
41. Poser I, Dominguez D, de Herreros AG, Varnai A, Buettner R & Bosserhoff AK. (2001) J Biol Chem 276: 24661–24666. | Article | PubMed | ISI | ChemPort |
42. Rasbridge SA, Gillett CE, Sampson SA, Walsh FS & Millis RR. (1993) J Pathol 169: 245–250. | Article | PubMed | ISI | ChemPort |
43. Rimm DL, Koslov ER, Kebriaei P, Cianci CD & Morrow JS. (1995) Proc Natl Acad Sci USA 92: 8813–8817. | PubMed | ChemPort |
44. Sanders DS, Blessing K, Hassan GA, Bruton R, Marsden JR & Jankowski J. (1999) Mol Pathol 52: 151–157. | PubMed | ChemPort |
45. Schipper JH, Frixen UH, Behrens J, Unger A, Jahnke K & Birchmeier W. (1991) Cancer Res 51: 6328–6337. | PubMed | ISI | ChemPort |
46. Shiozaki H, Tahara H, Oka H, Miyata M, Kobayashi K & Tamura S et al.. (1991) Am J Pathol 139: 17–23. | PubMed | ISI | ChemPort |
47. Silye R, Karayiannakis AJ, Syrigos KN, Poole S, van Noorden S & Batchelor W et al.. (1998) J Pathol 186: 350–355. | Article | PubMed | ISI | ChemPort |
48. Soler C, Rousselle P & Damour O. (1998) Cell Adhes Commun 5: 13–25. | PubMed | ISI | ChemPort |
49. Sommers CL, Gelmann EP, Kemler R, Cowin P & Byers SW. (1994) Cancer Res 54: 3544–3552. | PubMed | ISI | ChemPort |
50. Sommers CL, Thompson EW, Torri JA, Kemler R, Gelmann EP & Byers SW. (1991) Cell Growth Differ 2: 365–372. | PubMed | ISI | ChemPort |
51. Stappert J & Kemler R. (1994) Cell Adhes Commun 2: 319–327. | PubMed | ISI | ChemPort |
52. Tamura G, Sakata K, Nishizuka S, Maesawa C, Suzuki Y & Iwaya T et al.. (1996) Jpn J Cancer Res 87: 1153–1159. | PubMed | ChemPort |
53. Thiery JP. (2002) Nat Rev Cancer 2: 442–454. | Article | PubMed | ISI | ChemPort |
54. Tomita K, van Bokhoven A, van Leenders GJ, Ruijter ET, Jansen CF & Bussemakers MJ et al.. (2000) Cancer Res 60: 3650–3654. | PubMed | ISI | ChemPort |
55. Vleminckx K, Vakaet L, Jr, Mareel M, Fiers W & van Roy F. (1991) Cell 66: 107–119. | Article | PubMed | ISI | ChemPort |
56. Wong AS & Gumbiner BM. (2003) J Cell Biol 161: 1191–1203. | Article | PubMed | ISI | ChemPort |
57. Yang J, Mani SA, Donaher JL, Ramaswamy S, Itzykson RA & Come C et al.. (2004) Cell 117: 927–939. | Article | PubMed | ISI | ChemPort |
58. Zhu AJ & Watt FM. (1996) J Cell Sci 109: 3013–3023. | PubMed | ISI | ChemPort |

Acknowledgements

We are indebted to Sibylla Lodermeyer for technical assistance, Gabriele Handschuh for providing the pBAT-E-cadherin-expression vector, David Goltzman for the chicken N-cadherin-luc vector, Barry M Gumbiner for the E-cadherin-catenin construct and Margaret J Wheelock and Keith R Johnson for the derivative E-cadherin and N-cadherin expression vectors. This work was supported by grants from the DFG to AK Bosserhoff.