mARVCF cellular localisation and binding to cadherins is influenced by the cellular context but not by alternative splicing

One of the most recently identified members of the p120(ctn) subfamily, the murine ARVCF protein (armadillo repeat gene deleted in velo cardio facial syndrome) comprises an N-terminal coiled-coil region and a central armadillo repeat region. This structure closely resembles that of p120(ctn) itself (Sirotkin et al., 1997; Kaufmann et al., 2000; Mariner et al., 2000; Anastasiadis and Reynolds, 2000). Not only do the armadillo repeat regions of both proteins share 56% homology but also exon-intron boundaries of the genes are very similar (Keirsebilck et al., 1998). The 35 amino acid N-terminal coiled-coil region of ARVCF (Sirotkin et al., 1997; Kaufmann et al., 2000) is generally known as a motif mediating protein-protein interactions, although such interactions have not yet been demonstrated for ARVCF or any other member of this subfamily. ARVCF’s armadillo repeat region is characterised by 10 of these repeats and a putative nuclear localisation signal (NLS) within this region, in addition to a putative nuclear export signal (NES) in the C-terminus of the protein (Sirotkin et al., 1997; Kaufmann et al., 2000). The armadillo motif, originally identified in a segment polarity gene in drosophila (Wieschaus and Rigglemann, 1987), consists of an imperfect series of 42 amino acids that form a positively charged groove (Riggleman et al., 1989).

The human gene maps to chromosome 22q11, the so-called DiGeorge critical region (Sirotkin et al., 1997; Bonne et al., 1998), which is hemizygous in 80-85% of DiGeorge patients and those with velo cardio facial syndrome (Desmaze et al., 1993; Kelly et al., 1993; Morrow et al., 1995). Human ARVCF appears to be more or less ubiquitously expressed, being found in a variety of tissues including heart, skeletal muscle, lung, brain, liver, pancreas and kidney (Sirotkin et al., 1997).

Murine and human ARVCF can associate with the membrane proximal amino acids in the cytoplasmic region of cadherins such as E-cadherin in epithelial cells and M-cadherin in muscle cells (Reynolds et al., 1994; Kaufmann et al., 2000; Mariner et al., 2000). This was shown in detail by binding assays using GST-fusion proteins comprising the cytoplasmic domain of M-cadherin plus several deletion mutants, demonstrating that the 55 membrane-proximal CPD amino acids of M-cadherin are necessary and sufficient for ARVCF binding. Vice versa, all ten armadillo repeats of ARVCF are necessary for efficient M-cadherin binding. Deletion of repeats 1 to 4 or 1 to 5 abolished the ability of ARVCF to colocalise with N-cadherin in rat ventricular cardiomyocytes, although such deletions still facilitated some interactions in vitro (Kaufmann et al., 2000). However, whether ARVCF directly connects the cadherin complex to the cytoskeleton or is involved in cadherin clustering is not yet clear.

In human ARVCF two alternative splicing events have been reported. One concerns the N-terminus leading to the removal of the coiled-coil domain and the use of an alternative start codon. The second splice event leads to the insertion of an 18 base pair exon in the armadillo region that alters the putative NLS (Sirotkin et al., 1997). It has also been shown for p120(ctn) and other members of the subfamily that different isoforms can arise by alternative splicing (Hatzfeld, 1999; Paulson et al., 2000). For p120(ctn) itself this applies to the N-terminus where alternative splicing leads to the use of different start codons. Furthermore, the armadillo repeat region and the C-terminus can be altered by using three alternative exons (Keirsebilck et al., 1998).

Cadherins are a multigene family of calcium-dependent transmembrane cell-cell adhesion glycoproteins that mediate homophilic interactions and are expressed in a tissue-specific manner (Ringwald et al., 1987; Takeichi, 1991; Geiger and Ayalon, 1992; Shapiro et al., 1995; Huber et al., 1996). Many of the cadherins have been classified according to the tissues from which they have been isolated, such as P-cadherin from placenta, E-cadherin first isolated from epithelial cells or M-cadherin from muscle. The classical cadherins (and M-cadherin) consist of an N-terminal extracellular domain, a short transmembrane region and a cytoplasmic domain (CPD) averaging 150-160 amino acids, which all exhibit a high degree of homology with each other (Chothia and Jones, 1997; Humphries and Newham, 1998; Kaufmann et al., 1999a). Most cadherins are known to form two distinct complexes with catenins via their CPD (Ozawa et al., 1989; Hirano et al., 1992; Aberle et al., 1994; Butz and Kemler, 1994; Hinck et al., 1994; Näthke et al., 1994; Knudsen et al., 1995; Hertig et al., 1996; Kuch et al., 1997; Finnemann et al., 1997; Yap et al., 1998; Allport et al., 2000). One complex is composed of the respective cadherin, β-catenin and α-catenin, a second complex contains cadherin, plakoglobin (also called γ-catenin) and α-catenin. α-catenin joins the complex by binding to β-catenin or plakoglobin and connects this cadherin-catenin complex to components of the cytoskeleton (Hirano et al., 1987; Tsukita et al., 1992). β-catenin interacts with the C-terminal part of cadherin’s CPD, whereas ARVCF and p120(ctn), for example, bind to the juxtamembrane region of the cadherin’s cytoplasmic tail, as discussed above. ARVCF and p120(ctn) compete for the same binding site in the CPD of cadherins (Mariner et al., 2000) but the different functions of the two molecules are as yet unknown.

Many proteins of the armadillo repeat family are known to enter the nucleus, although the mechanism and functional consequences of this have only been described for β-catenin. In addition to its interaction with cadherins, β-catenin can enter the nucleus alone or complexed with Tcf/Lef, a transcription factor of the Lef1/TCF family. Together with Tcf/Lef, β-catenin can stimulate the transcription of different target genes (Behrens et al., 1996; Molenaar et al., 1996; van de Wetering et al., 1997). Similary, ARVCF and p120(ctn) show a dual localisation at cell-cell junctions and under some circumstances in the nucleus (van Hengel et al., 1999; Kaufmann et al., 2000; Mariner et al., 2000). Their role in the nucleus, however, remains to be determined.

We report here the identification and cloning of novel N- and C-terminal isoforms of murine ARVCF resulting, for example, from the use of different start codons. By using RT-PCR, we show that the appearance of the isoforms varies depending on the cell line or tissue examined. Cloned as EGFP-fusion proteins and expressed in different cell lines we demonstrate that the localisation of mARVCF isoforms is not influenced by the N- or C-terminus of the protein but depends on the cellular context. Using the MOM recruitment assay we examined the ability of all isoforms to associate with M-, E- or N-cadherin in different cell types.

Mouse myoblasts, i28, derived from a primary satellite cell culture (Kaufmann et al., 1999b) were grown in Ham’s nutrient mixture F10 (Gibco) supplemented with 20% fetal calf serum (FCS; Sigma) in 5% CO₂ at 37°C. To initiate myogenic differentiation, growth medium was replaced by differentiation medium consisting of Dulbecco’s modified Eagle’s (DMEM) medium (Gibco) with 10% horse serum (Sigma, Buchs, Switzerland). MCF7 cells, HeLa cells, CMT cells, EJ28 cells, COS-7 cells and RT112 cells were grown in DMEM supplemented with 10% FCS. The cells were transfected with 2 μg plasmid DNA using SuperFect™ Transfection Reagent, or with 1.5 μg plasmid DNA using PolyFect™ Transfection Reagent, both from Qiagen (Hilden, Germany). For the MOM recruitment assay, cells were cotransfected with 1 μg pMOM-M/E/N-cadherin vector and 1 μg of either of the mARVCF EGFP-fusion plasmids (see plasmid constructions). The different cell types were subjected to immunofluorescence microscopy 24-48 hours after transfection. Cells used in addition to i28 were: MCF7 human breast carcinoma cells (ATCC HTB-22); COS-7 kidney cells from African green monkey (ATCC CRL-1651); HeLa human cervix carcinoma cells (ATCC CCL-2.1); RT112 human, non-invasive bladder carcinoma cells and EJ28 human, invasive bladder carcinoma cells (Gaetje et al., 1997).

The monoclonal antibody 4A6 described previously (Rüdiger et al., 1997) was used to identify the birch profilin (BP) tag. Monoclonal anti-GFP antibody was obtained from Clontech (Heidelberg, Germany). Polyclonal antibodies against the extracellular domain of M-cadherin were affinity-purified as described (Rose et al., 1994; Kaufmann et al., 1999b). Monoclonal pan-cadherin (clone CH-19) and monoclonal N-cadherin antibody (anti-A-CAM, clone GC-4) were obtained from Sigma (Buchs, Switzerland). E-cadherin antibody was obtained from Monosan (Germany). Secondary antibody (Alexa Fluor 568) was obtained from Molecular Probes (Leiden, The Netherlands).

Cells grown on coverslips were rinsed in PBS and fixed in 4% paraformaldehyde (PFA) in PBS at room temperature for 10 minutes. After fixation, cells were permeabilised by incubation with 0.2% Triton X-100 in PBS for 10 minutes, washed three times with PBS and incubated with the relevant antibodies diluted in PBS/10% FCS for 1 hour (RT). After washing three times with PBS, binding of the primary antibodies was detected by species-specific fluorochrome-conjugated secondary antibodies diluted in PBS/10% FCS. Controls in the absence of primary antibodies confirmed the specificity of the immunolabelling. Fluorescence was monitored with a Zeiss Axiophot microscope. Pictures were taken with 40×, 63× or 100× objectives. Kodak Elite 400 film (400 ASA; Eastman Kodak, Rochester, NY) was used for colour slides.

SDS-PAGE and immunoblots were performed as described for E-cadherin (Butz and Kemler, 1994) and M-cadherin (Kuch et al., 1997; Kaufmann et al., 1999b). Membranes were incubated with a primary GFP-antibody for 1 hour followed by incubation with alkaline phosphatase (AP)-conjugated secondary antibody (Dianova, Hamburg, Germany), which was visualised using the phosphatase substrates nitroblue-tetrazolium and 5-bromo-4 chloro-3-indolyl (NBT/BCIP) (Boehringer Mannheim, Germany).

The different splice variants of mARVCF were obtained by RT-PCR using Pfx-polymerase (Gibco). First strand cDNA was prepared from RNA templates extracted from differentiating i28 cells 30 hours after induction for fusion using the primers indicated in Fig. 1 (primer sequences in 5′→3′ orientation: 5′ UTR: GCCTGTCTTGGGGGCGGA; 6R: ACTCGGTCCAAGCTGCCC; Seq5-5: ATCGCGCTGCGCAACCTCTCA; Seq5-6: TGCAGAGGGATGGCTGGACGA; Ex19as: GGATACTGGCACACAGGTGG; 11R: TCTCCTACCACACAGCACC). Subsequently, 1 μl Taq-polymerase (Gibco) was added to produce 3′-overhanging adenine nucleotides for cloning the fragments into the vector pGEM™-T Easy (Promega). The fidelity of the amplified fragments was confirmed by DNA sequencing.

To clone the C-terminal splice variants of mARVCF as full length constructs, the pGEM™-T Easy vectors containing the four different C-terminal fragments (see previous section) and full length mARVCF (Kaufmann et al., 2000) were used as templates for PCR. Two fragments were produced with overlaps at the 3′-end of the N-terminal fragment and the 5′-end of the C-terminal fragment. These two PCR products were used in a linear amplification where the base-paired overlaps served as a primers. The product obtained was re-amplified with the outer primers containing an EcoRI restriction site in the sense primer and a SalI restriction site in the reverse primer; the resulting PCR products were inserted into the EcoRI and SalI restriction sites of the eukaryotic expression vector pEGFP-C2 (Clontech, Heidelberg, Germany). The mARVCF variants with alternative 5′-ends were similarly produced by PCR using a sense primer containing an EcoRI restriction site and a reverse primer containing a SalI restriction site and the corresponding full length construct as a template. The resulting PCR products were also inserted into the EcoRI and SalI restriction sites of the eukaryotic expression vector pEGFP-C2. All PCRs used the Pfu-polymerase (Promega).

The vector pMOM was used for the analysis of intracellular recruitment of ARVCF by different cadherins (for plasmid construction see Kaufmann et al.) (Kaufmann et al., 2000). Mouse M-cadherin (residues 626-784; GenBank accession no. M74541) or mouse N-cadherin cytoplasmic domain (residues 747-906; GenBank accession no. AB008811) were amplified either from a plasmid (M-cadherin) or by RT-PCR from mRNA of mouse heart tissue (N-cadherin) with primers containing a BamHI (sense primer) and EcoRI restriction site (reverse primer) for insertion of the PCR fragments into pMOM.

GST-fusion constructs were generated by cloning the cytoplasmic domain of the respective cadherin as a PCR-product into the prokaryotic expression vector pGEX-5X-1 (Pharmacia, Freiburg, Germany) using BamHI and EcoRI restriction sites.

mARVCF splice variants FL-C11 and FL-3/7 were cloned into the expression vector pcDNA3.1 (Invitrogen, Netherlands) and synthesised by in vitro transcription-translation in the presence of ³⁵S-methionine using the TNT™-coupled reticulocyte lysate (Promega, Mannheim, Germany). GST-fused cytoplasmic domain of the respective cadherin was expressed in and purified from E. coli strain BL21pLys.S. The in vitro GST binding assay was performed as described (Kaufmann et al., 2000).

All clones were sequenced by SeqLab (Göttingen, Germany). DNA sequence analysis and homology searches were performed using HUSAR from DKFZ Heidelberg and the Blast-program-packet from NCBI, USA.

Splice variants of murine ARVCF were obtained by RT-PCR using primer combinations chosen in such a way that they covered the entire mARVCF sequence (for primer sequences see Materials and Methods). Three of the primer combinations generated additional bands, which were presumed to arise from alternative splicing within both mARVCF’s N- and C-terminus. The positions of these three primer sets are indicated in Fig. 1A. PCR fragment sizes generated from the known mARVCF mRNA sequence and the sizes of those obtained from alternative splice variants are indicated in Table 1. Cloning of the novel PCR fragments into pGEM™-T Easy and subsequent DNA sequence analysis revealed one novel N-terminal splice variant of mARVCF and three new variants in the C-terminus (Fig. 1B,C). The alternative 5′-end of mARVCF (5′alt) lacks exon 3, which contains both the methionine start codon for the full-length (FL) construct and the coiled-coil domain. The first methionine in exon 4 is now the putative translation start.

The previously known C-terminus of mARVCF mRNA, called C11, was recently identified by a yeast two-hybrid assay using the cytoplasmic domain of M-cadherin as bait (Kaufmann et al., 2000). In the variant mARVCF 3/5, exon 19 (present in C11) is spliced out (Fig. 1B,C). This changes the reading frame between exon 18 and exon 20, generating an earlier stop codon and thus a truncated protein. In variant mARVCF 3/7, exon 19 is replaced by the 120 nucleotide long exon B, containing a stop codon just before exon 20. Analysis of the cDNA-derived amino acid sequence showed that exon B maintains the same reading frame as the rest of mARVCF. Exon B has been found in human ARVCF (hARVCF) (Sirotkin et al., 1997) and, as in hARVCF, introduces a putative PDZ-binding domain at the very C-terminus of the truncated mARVCF, which is not present in any other splice variant identified so far (Fig. 1C).

The longest open reading frame of mARVCF is encoded by the third new C-terminal isoform Y, which shows an in-frame insertion of 273 base pairs between exon 18 and 19 (Fig. 1B,C). Analysis of the extended peptide sequence has not revealed any homology to known protein domain motifs.

Finally, primer pairs were chosen to discover possible splice variants within the armadillo repeat region. Alternative splicing has been reported in this region for human p120(ctn), as well as human and xenopus ARVCF (Sirotkin et al., 1997; Keirsebilck et al., 1998; Paulson et al., 2000) resulting in an 18 base pair insertion in armadillo repeat six that converts the putative nuclear localisation motif present at this position into a shorter NLS. This alternative exon, or any other alternative exons within the armadillo repeat region, could not be detected within murine ARVCF using RT-PCR (data not shown).

Altogether, eight different mARVCF splice variants are possible by combining one out of four C-termini with either full length or the 5′ alternative N-terminus. Fig. 1B schematically summarises the N- and C-terminal variants and Fig. 1C indicates the amino acid sequences derived from the different putative mARVCF isoforms.

The next question was whether one splice variant is preferentially expressed in different cell lines or tissue types. In order to determine the relative quantity of the 5′- and the 3′- variants we performed RT-PCR with two pairs of oligonucleotide primers that either amplify both N-terminal variants or, alternatively, all four C-terminal variants simultaneously. Templates were cDNAs prepared from mRNA of differentiating i28 cells, CMT cells (derived from mouse colon carcinoma) and total mouse heart. As shown in Fig. 2A, the N-terminal full-length mARVCF mRNA is much more abundant than that encoding the 5′-alternative end missing the coiled-coil domain. However, at the 3′-end it is evident that the relative amounts of the four C-terminal isoforms are very similar in heart and CMT cells. Isoforms 3/5 and Y are expressed at low levels, whereas variants C11 and 3/7 are predominant. In i28 cells the mRNA of mARVCF splice variant Y is even less abundant, but can easily be amplified using exon Y specific primers (data not shown). Variant 3/5 mRNA is more abundant in this cell line than in CMT cells or heart tissue. Using this primer pair it was difficult to discriminate between isoform C11 and 3/7 because they differ in length by only 28 nucleotides. Therefore, one sense primer in exon 15 was used together with two antisense primers: one specific for the 3′-end of exon B and an other specific for the 5′-end of exon 19 generating bands that differ in length by 47 nucleotides. The exon 19-specific primer is also able to amplify isoform Y but this was not relevant owing to the very rare appearance of this variant (Fig. 2B). Fig. 2C shows that variant 3/7 is more abundant than variant C11 in i28 cells, heart tissue and CMT cells.

In order to show that all cell lines (i28, MCF7, RT112, EJ28, HeLa and COS-7) that were used for further experiments express endogenous ARVCF, we performed RT-PCR analyses. One pair of oligonucleotide primers amplified ARVCF in the region encoding the armadillo repeats that is not affected by alternative splicing. As an internal standard the housekeeping gene BIP (binding protein) was amplified in the same RT-PCR. As shown in Fig. 2D, ARVCF is expressed in each cell line examined and the endogenous level of ARVCF expression is comparable in the studied cell lines.

The eight possible splice variants of mARVCF were cloned as EGFP-fusion proteins and expressed in COS-7 cells. Protein extracts from the transfected cells were analysed by western blots using the monoclonal anti-GFP antibody (Fig. 3A). The results indicated that all eight constructs can be expressed and that the corresponding proteins appeared at the expected position in the blot. Furthermore, we expressed the most abundant splice variant full length 3/7 (FL-3/7) as a EGFP-fusion protein in each cell line used for further investigations. Western blot analysis of the protein extracts from the transfected cells lines indicates that EGFP-FL-3/7 protein is expressed at the correct size and in comparable amounts in all six cell lines (Fig. 3B).

As shown by Kaufmann et al., EGFP-ARVCF-C11 is able to interact with M-cadherin and E-cadherin in the MOM recruitment assay and colocalises with N-cadherin in rat ventricular cadiomyocytes, or E-cadherin in epithelial cells (Kaufmann et al., 2000). Furthermore, hARVCF can associate with the E-cadherin-catenin complex as shown by immonoprecipitation (Mariner et al., 2000). With regards to mARVCF, the armadillo repeat region is required for the interaction with cadherins (Kaufmann et al., 2000). To investigate whether the C- or N-termini of the novel mARVCF isoforms influence the cadherin-binding and/or cellular localisation of mARVCF, six different cell lines were transfected with each of the isoforms as an EGFP-fusion protein. Twenty-four to 48 hours after transfection the cells were fixed and analysed for EGFP-mARVCF expression. In addition to the mouse muscle myoblast cell line i28, these cells were human epithelial cancer cell lines MCF7 (mammary carcinoma), RT112 (non-invasive bladder carcinoma), EJ28 (invasive bladder carcinoma), HeLa cells (cervix carcinoma) and monkey COS-7 kidney cells, all of which express endogenous ARVCF as shown by RT-PCR (Fig. 2D).

i28 cells, the original source of the novel mARVCF splice variants, express M-cadherin, which localises to the cell membrane predominantly at cell-cell contacts (Fig. 4Ai). As described recently, mARVCF binds to the cytoplasmic domain of M-cadherin, which correlates with the fact that the ectopic fragment EGFP-ARVCF-C11 is localised at the membrane in i28 cells, although it is, to a certain extent, also found in the cytoplasm (Kaufmann et al., 2000). The membrane localisation is reproduced when the different mARVCF isoforms are expressed in i28 cells. The nuclei of the muscle cells were found to be free of the ectopically expressed mARVF isoforms whereas the sites of cell-cell contacts were clearly stained. Here the EGFP-fusion proteins colocalise with M-cadherin, as exemplified by the merged image (Fig. 4Am) with variant FL-3/7.

To investigate the cellular localisation of mARVCF splice variants in cells expressing endogenous E-cadherin, MCF7 cells (mammary carcinoma) and RT112 cells (bladder carcinoma), both of which are human non-invasive cell lines, were transfected with the cDNAs encoding the splice variants as EGFP-fusion proteins and fixed 24-48 hours after transfection. In both MCF7 and RT112 cells each of the isoforms clearly localised at the cell membrane together with E-cadherin and was never detected in the nucleus (Fig. 4B,C; and Fig. 4Bm,Cm). This indicated that not only mARVCF fragment C11 (Kaufmann et al., 2000) but also all of the splice products can associate with E-cadherin.

In contrast to MCF7 and RT112 cells, the human invasive bladder carcinoma cell line EJ28 does not express E-cadherin but is positive for N-cadherin (Fig. 4Di) (A. Zeitvogel, unpublished). When transfected into EJ28 cells none of the mARVCF isoforms appeared to colocalise with N-cadherin (Fig. 4D,Dm), although the adhesion molecule itself was found correctly at the plasma membrane (Fig. 4Di). All mARVCF isoforms were distributed equally in the EJ28 cells and no clear membrane-staining could be detected. However, after transfection of HeLa cells, which are also E-cadherin-negative but N-cadherin-positive (Fig. 4Ei), all of the EGFP-mARVCF isoforms clearly colocalised with N-cadherin at the sites of cell-cell contact (Fig. 4E,Em).

Finally, in monkey kidney COS-7 cells, which express endogenous cadherin(s) detectable with a pan-cadherin antibody (Fig. 4Fi), the mARVCF isoforms localised at cell-cell contacts (Fig. 4F). Thus, our results support the idea that the subcellular localisation of mARVCF splice variants depends not only on the presence or absence of an appropriate interaction partner, in this case the cadherins, but also on additional factors.

The results above merited closer investigation into the binding potential of E-, M- and N-cadherin for the eight mARVCF splice variants. The general ability of M-, E- and N-cadherin to interact with mARVCF was demonstrated by in vitro GST binding assays. mARVCF splice variants FL-C11 and FL-3/7 were cloned into the expression vector pcDNA3.1 and used as a template for in vitro transcription and translation in the presence of ³⁵S-methionine. The cytoplasmic domains of M-, E- and N-cadherin were expressed as GST-fusion proteins in bacteria. The results revealed that in vitro translated mARVCF splice variants FL-C11 and FL-3/7 can bind directly to the cytoplasmic domain of all three cadherins (Fig. 5).

Furthermore, we used the MOM recruitment assay in the different cell lines. This assay provides a means of testing the interaction between two proteins directly in mammalian cells (Kaufmann et al., 2000) (see also Materials and Methods). The cytoplasmic domains (CPD) of the cadherins were cloned into the MOM vector to produce a birch profilin (BP)-tagged fusion protein that is anchored to the mitochondrial outer membrane via the TOM70 protein anchor, and that shows a mitochondrial staining pattern in immunofluorescence assays (Kaufmann et al., 2000). Following co-transfection of the MOM-construct with each of the EGFP-mARVCF splice variants, these proteins also exhibit a mitochondrial fluorescence pattern if there is protein-protein interaction. As a control, the empty EGFP-vector is co-transfected with each of the MOM-constructs. No interaction between these two components means that the EGFP shows no mitochondrial localisation pattern (Fig. 6co, coi).

As exemplified in Fig. 6 for FL-C11 and MOM-M-cadherin, all mARVCF isoforms were able to interact comparably in the MOM recruitment assay with E- or M-cadherin in i28 myoblasts (Fig. 6A) and in the different carcinoma cell lines (Fig. 6C-F). In MCF7 cells, the membrane localisation of mARVCF did not disappear completely following co-transfection of either MOM-M-cadherin or MOM-E-cadherin (Fig. 6B). This suggests a competition between endogenous (membrane-located) and the mitochondrial-anchored cytoplasmic domains of E- or M-cadherin.

Mouse EGFP-ARVCF-C11 colocalises with N-cadherin in cardiomyocytes suggesting an association of these molecules (Kaufmann et al., 2000). In the experiments described here, the localisation of mARVCF isoforms was found to be distinct when EJ28 and HeLa cells were compared, although both cell types express N-cadherin but not E-cadherin (Fig. 4Di,Ei) (A. Zeitvogel, unpublished). Thus, it was of interest to similarly test the interaction of N-cadherin’s CPD with mARVCF isoforms in various cellular environments, including EJ28 cells. Co-transfection of the pMOM-N-cadherin and each EGFP-mARVCF isoform revealed that the interaction between each of the mARVCF splice variants and N-cadherin is possible, as exemplified by FL-3/7 in Fig. 7A-C. Also in EJ28 cells, such an interaction takes place between the transfected cytoplasmic N-cadherin domain and mARVCF. This finding suggests that there are no or few factors preventing this protein-protein interaction. But in contrast to the assays done with M- or E-cadherin, this interaction did not take place in every single cell, although both constructs were present and expressed (Fig. 7). Neighbouring cells, all of which expressed the MOM-N-cadherin construct and mARVCF, exhibited distinct interactions. For example, one cell showed interaction while the neighbour did not (Fig. 7). This inhomogeneous pattern of association between the cytoplasmic domain of N-cadherin and EGFP-mARVCF could be detected in every cell line examined with all the mARVCF isoforms. By counting 1.5×10³ cotransfected MCF7 cells we could show that 36.7% of these cells were positive for FL-3/7-MOM-N-cadherin interaction, whereas 63.3% of these cells showed no interaction of EGFP-FL-3/7 with MOM-N-cadherin, although both constructs were present (Fig. 7D). The results of the subcellular localisation and the MOM recruitment assays are summarised in Table 2.

ARVCF belongs to the p120(ctn) subfamily of armadillo repeat proteins and shows a high homology to the name-giving molecule. Both relatives can bind to the juxtamembrane region of cadherins (Kaufmann et al., 2000; Mariner et al., 2000) and are mutually exclusive for one another in E-cadherin complexes (Mariner et al., 2000). To investigate whether the inability of mARVCF to associate with MOM-N-cadherin in 63.3% of cells (Fig. 7D) is due to a competition between both armadillo repeat proteins for MOM-N-cadherin binding, we used p120(ctn) isoform 1A, which is the most homologous to mARVCF FL-3/7, in MOM recruitment assays. EGFP-p120(ctn) was able to interact with MOM-M- and-E-cadherin in all cell lines tested (data not shown). Interestingly, cells transfected with EGFP-p120(ctn) and MOM-N-cadherin showed the same inhomogenous pattern of association between both partners (Fig. 8) as it was observed in cells cotransfected with mARVCF and MOM-N-cadherin (Fig. 7). To study the idea of competition for MOM-N-cadherin binding, we co-transfected MCF7 cells with pMOM-N-cadherin, FL-3/7 fused to an Xpress-tag and EGFP-p120(ctn). Cells were stained with anti-Xpress antibodies to detect Xpress-tagged FL-3/7. Pictures were taken where both, FL-3/7 (Fig. 8Aa) and p120(ctn) (Fig. 8Ab) were located at the plasma membrane, hence did not interact with MOM-N-cadherin. To confirm that these cells were indeed positive for MOM-N-cadherin, the same coverslips were washed again several times with PBS and then stained with the BP antibody to detect MOM-N-cadherin. The positive MOM-N-cadherin staining is given in Fig. 8Ad, indicating that the binding of both, mARVCF and p120(ctn) to endogenous cadherins can occur in the presence of overexpressed MOM-N-cadherin. Taking into account the inhomogenous interaction pattern of the armadillo repeat proteins with MOM-N-cadherin, we also analysed cells where both FL-3/7 (Fig. 8Ba) and p120(ctn) (Fig. 8Bb) were able to bind to mitochondrial located MOM-N-cadherin. This showed that these two members of the p120(ctn) subfamily can bind to the overexpressed MOM-N-cadherin constructs in the same cell.

These data clearly show that the inhomogenous pattern of association of mARVCF and the cytoplasmic domain of N-cadherin, observed in the MOM recruitment assay, is not due to a competition between p120(ctn) and ARVCF. Both armadillo proteins can bind to endogenous cadherins at the plasma membrane even though MOM-N-cadherin is present. Alternatively, both proteins can bind to MOM-N-cadherin constructs in the same cell (Fig. 8).

In addition, the percentage of cells interacting with the cytoplasmic domain of N-cadherin in the MOM-recruitment assay was not altered significally by co-transfecting EGFP-p120(ctn). In the absence of p120(ctn), 36.7% of co-transfected cells showed an interaction of mARVCF and MOM-N-cadherin (Fig. 7D). In cells co-transfected with MOM-N-cadherin, FL-3/7 and p120(ctn), 32.9% showed an interaction of mARVCF with MOM-N-cadherin, and 35.2% of them exhibited an interaction of p120(ctn) with the CPD of N-cadherin (Fig. 8Af). Thus an interaction between mARVCF and MOM-N-cadherin is not affected significantly by the co-transfection of p120(ctn).

In this paper we report the isolation and characterisation of novel mARVCF splice variants. These isoforms were tested for their binding capacity to different cadherins in various cell contexts. In fact, all eight derivatives showed identical behaviour to each other in all assays performed (discussed later). However, the distinct binding pattern of the mARVCF isoforms to cadherins in the MOM recruitment assay suggests that the binding capacity of mARVCF to E-, M- and N-cadherin varies in vivo, although all three cadherins appear to bind equally well in vitro to mARVCF splice variants FL-C11 and FL-3/7. Although mARVCF isoforms reacted with MOM-M-cadherin and MOM-E-cadherin in all cell lines tested and in each individual co-transfected cell, their interaction with MOM-N-cadherin was rather inhomogeneous in a given cell population. The observation that transfected MOM-E-cadherin (AND/OR M-cadherin) protein did not completely abolish binding of EGFP-mARVCF isoforms to endogenous (junction-localised) E-cadherin in MCF7 cells implies a competition for mARVCF binding between the endogenous E-cadherin and the E- or M-cadherin CPD-MOM fusion protein. However, this is most likely influenced by additional cellular factors since this type of competition does not occur in RT112 cells that also express E-cadherin.

In contrast to M- and -E-cadherin, N-cadherin exhibits an inhomogenous pattern of association with the mARVCF isoforms, although it is generally able to interact with them. Not every cell expressing the MOM-N-cadherin fusion protein and a mARVCF isoform permits the protein interaction. ARVCF belongs to the p120(ctn) subfamily of armadillo proteins and shows a high homology to the namegiving molecule. Both relatives can bind to the juxtamembrane region of cadherins (Kaufmann et al., 2000; Mariner et al., 2000) and are mutually exclusive for one another in E-cadherin complexes (Mariner et al., 2000). Our results clearly demonstrate that the inhomogeneous pattern of interaction of mARVCF and the CPD of N-cadherin observed in the MOM recruitment assay, is not due to a competition between mARVCF and p120(ctn) for N-cadherin binding. The two members of the p120(ctn) subfamily can interact with endogenous, membrane-located cadherins in the presence of overexpressed MOM-N-cadherin. Alternatively, both mARVCF and p120(ctn) can be recruited to MOM-N-cadherin constructs in the same cell. Taking into account the results of Mariner et al. (Mariner et al., 2000), it is obvious that these two members of the p120(ctn) subfamily do not bind simultaneously to the same MOM-N-cadherin molecule. The inhomogeneous pattern of interaction might be the result of heterogeneity in the cellular context (e.g. expression profile of regulatory proteins) leading to differential modifications of the interaction partners in individual cells.

It is also interesting to note that the mARVCF isoforms cannot colocalise with N-cadherin in EJ28 carcinoma cells but do so in HeLa cells. Both cell types express N-cadherin and not E-cadherin (A. Zeitvogel, unpublished; and this paper). Furthermore, it could be shown by Kaufmann et al. (Kaufmann et al., 2000) that EGFP-ARVCF-C11 colocalises with N-cadherin in rat ventricular cardiomyocytes. This indicates that a given cellular background may determine the capacity of mARVCF to interact with its partner, in this case N-cadherin. In line with this, a previous report described the regulation of p120(ctn) binding to the cytoplasmic domain of cadherins by phosphorylation, thereby modulating cadherin-mediated adhesion in either a positive or negative manner. This process appears to depend on the cell context (Yap et al., 1998; Aono et al., 1999; Anastasiadis and Reynolds, 2000).

The evidence for the inhomogenous association of mARVCF isoforms with N-cadherin might also be relevant in the context of results that imply that expression of N-cadherin is associated with metastasis and cell migration (Nieman et al., 1999; Hazan et al., 2000). In particular, over-expression of N-cadherin in MCF7 cells leads to abnormal invasive behaviour of these cells in cell culture (Hazan et al., 2000). In turn, expression of E-cadherin in invasive cells abolishes invasion (Frixen et al., 1991). Furthermore, N-cadherin has been reported to be a path-finding molecule for migrating dermomyotomal cells during chicken embryogenesis (Brand-Saberi et al., 1996). Considering the concept that cell invasion and migration is incompatible with strong adhesion (as provided by E-cadherin) but still requires guidance mediated by low affinity cadherins (such as N-cadherin) that allow movement, mARVCF might play a modulatory role in these processes.

Apparently, the binding of the identified mARVCF isoforms to cadherins is generally unaffected by the altered N- and C-termini of the mARVCF proteins arising by differential splicing. This is in line with results described by Kaufmann et al. showing that the armadillo repeat region of mARVCF is required and sufficient for cadherin binding (Kaufmann et al., 2000). Alternative splicing events have been reported for most of the p120(ctn) subfamily members (Paffenholz and Franke, 1997; Keirsebilck et al., 1998; Hatzfeld, 1999) suggesting that the isoforms may be important modulators of the function of these proteins. Modulation might, for example, occur due to the presence (as in full length mARVCF) or absence (such as in 5′-alt mARVCF) of the N-terminal coiled-coil domain, which may allow the recruitment of different partners into the junctional cadherin-catenin complexes. Similarly, the expression of a putative PDZ-binding domain in mARVCF is regulated through alternative splicing and only occurs in variant 3/7.

The PDZ domain has been generally shown to mediate protein-protein interactions (Fanning and Anderson, 1996; Fanning and Anderson, 1998). For example, it has been demonstrated that one PDZ domain of the junction protein PAPIN is responsible for the interaction of this molecule with the armadillo repeat proteins δ-catenin or p0071 (Deguchi et al., 2000). Furthermore, β-catenin can associate with LIN-7 via its PDZ domain at cell junctions (Perego et al., 2000). Postulating a similar function of the potential PDZ-binding domain in mARVCF, it might be that this domain can modulate the interaction of the cadherin-catenin complex by recruiting additional proteins to the complex.

With human p120(ctn) alternative splicing leads to at least 32 potential isoforms. In the N-terminal region four different variants have been described that use four different start methionines (isoforms one to four, of which variants two to four lack the coiled-coil region). These can be combined with three alternative exons A, B, and C, in the armadillo repeat region or the C-terminus (Keirsebilck et al., 1998). So far, however, only the differential expression of exon B could be assigned to a defined mechanism, namely the insertion of a functional NES (nuclear export signal) directing the molecule out of the nucleus (van Hengel et al., 1999). Thus, the major roles for the isoforms of both ARVCF and p120(ctn) remain to be elucidated.

In summary, our data imply that the function of mARVCF may be regulated by multiple mechanisms including alternative splicing, cellular context and the cadherins themselves.

We thank Jean-Claude Perriard for critical reading of the manuscript, Beata Krebs for technical assistance, Heike Handrow-Metzmacher for help with the genomic ARVCF analysis and Frank Nonnenmacher for providing EGFP-p120(ctn). This work was supported by the Deutsche Forschungsgemeinschaft through SFB 474 (B2) to A.S.-P.