The metastatic subline of a rat pancreatic adenocarcinoma differs from the non-metastasizing subline by overexpression of 5 membrane molecules: CD44 variant isoforms, EpCAM, the tetraspanin D6.1A, an uPAR-related molecule and, as described here, the α6β4 integrin. An antibody-defined molecule was identified by mass spectrometry and cloning as α6β4 integrin. Transfection-induced expression of α6β4 in the non-metastasizing subline did not support migration on laminin 5 or tumor progression. However, when the non-metastasizing subline was doubly transfected to express α6β4 and the D6.1A tetraspanin, intraperitoneally injected tumor cells frequently formed liver metastasis. For the following reasons we assume that metastasis formation is supported by an interaction between α6β4 and D6.1A. (i) The 2 molecules can associate and co-localize. (ii) Co-localization is strengthened by PKC stimulation. (iii) PKC stimulation, which induces a migratory phenotype, leads to a redistribution of α6β4/D6.1A complexes. In resting cells, the molecules co-localize at the trail of the cell; during PKC stimulation they become transiently internalized and are (re-)expressed in the leading lamella. Thus, in the appropriate milieu, i.e. intraperitoneally, α6β4 changes from an adhesion-supporting towards a migration-supporting molecule by its association with a tetraspanin. The findings provide a convincing experimental explanation for the repeatedly described involvement of α6β4 in tumor progression.
Metastasis formation is a multistep process that requires partly opposing properties from the metastasizing tumor cells, such as, detachment from the primary tumor tissue and adhesion to vessel endothelium. It is supposed that the metastasizing tumor cell fulfills these distinct requirements by silencing and upregulation of expression of a whole set of genes, which as such may well be unaltered (Lozahic et al., 2000; Mareel et al., 1993; Poste and Fidler, 1980; MacDonald and Steeg, 1997). Though it is generally agreed that the metastasizing phenotype cannot be explained by alteration, upregulation or silencing of a single gene, not much is known about possible cooperation between `metastasis-associated' gene products and alterations of their functional activity by such an association.
We have described five monoclonal antibodies (mAb) that recognize surface molecules on metastasizing, but not on non-metastasizing rat tumor lines (Claas et al., 1996; Matzku et al., 1983). The molecules have been identified as CD44v4-v7 (Günthert et al., 1991), the tetraspanin D6.1A (Claas et al., 1998), C4.4A, a molecule with homology to uPAR (Rösel et al., 1998), D5.7A, the rat homologue of EpCAM (Würfel et al., 1999) and, described in this report, the α6β4 integrin. Thus, all five molecules are presumably unaltered gene products and physiological functions of some of them are well described. Interestingly, all five are also known for their involvement in metastasis formation (reviewed by Zöller, 1998). The α6β4 integrin is a receptor for laminin, preferentially for laminin 5 (Belkin and Stepp, 1999; Falk-Marzillier et al., 1998; Lee et al., 1992; Nguyen et al., 2000; Stahl et al., 1997) and is a major component of hemidesmosomes (Dowling and Fuchs, 1996; Jones et al., 1998; Niessen et al., 1997; Nievers et al., 1998). There are numerous reports that, dependent on the cell type, upregulation or downregulation of α6β4 is required for tumor progression (reviewed by Mercurio and Rabinovitz, 2001; Zutter et al., 1998). In addition, increased motility and invasiveness is linked to PKC activation via the EGFR, which is accompanied by disintegration of hemidesmosomes (Gambaletta et al., 2000; Maniero et al., 1996; Rabinovitz et al., 1999; Rigot et al., 1998), the formation and stabilization of actin-containing motility structures (Rabinovitz and Mercurio, 1997) and the stimulation of MMP-2 secretion (Daemi et al., 2000; Sugiura and Berditchevski, 1999).
Tetraspanins have been described as molecular facilitators (Claas et al., 2001; Maecker et al., 1997; Todres et al., 2000). They form protein complexes that are mostly composed of different tetraspanins and integrins, but can also contain members of other protein families (Fitter et al., 1999; Horvath et al., 1998; Indig et al., 1997; Lozahic et al., 2000; Mannion et al., 1996; Rubinstein et al., 1997; Scherberich et al., 1998; Serru et al., 1999; Sincock et al., 1999; Tiwari-Woodruff et al., 2001; Yauch et al., 1998). The strongest association between tetraspanins and integrins has been described for CD151 and the α3 and α6 integrins (Sincock et al., 1999; Yauch et al., 1998; Yauch et al., 2000). For the rat tetraspanin D6.1A, which is closely related to CD9, we originally noted an association with the α3 and the α6β1 integrin (Claas et al., 1996). Meanwhile, an association with the α6β4 integrin was described for 2 tetraspanins (Baudoux et al., 2000; Jones et al., 1996; Sterk et al., 2000). CD151 associates with α6β4 within hemidesmosomes (Sterk et al., 2000), whereas an association between CD9 and α6β4 has only been observed outside of hemidesmosomes (Baudoux et al., 2000). The association between tetraspanins and integrins may be accompanied by changes in adhesiveness versus motility (Berditchevski and Odintsova, 1999; Domanico et al., 1997; Hinterman et al., 2001; Penas et al., 2000). As possible underlying mechanisms shedding, internalization and redistribution on the cell membrane have been discussed (Bretscher, 1992; Friedl et al., 1998; Gaietta et al., 1994; Nath et al., 2000). Tetraspanins are also known to contribute to the metastatic process (Charrin et al., 2001; Claas et al., 1998; Odintsova et al., 2000; Testa et al., 1999) (reviewed by Levy et al., 1998).
We have cloned the rat α6β4 integrin, which is highly expressed on several metastasizing rat tumor lines, and show that α6β4 expression on locally growing tumor lines does not confer the metastatic phenotype. Instead, there is evidence that co-expression of the α6β4 integrin and the D6.1A tetraspanin contributes to the hematogeneous spread of tumor cells. Metastatic spread is supported by the acquisition of a motile phenotype via transient internalization and re-expression of the α6β4 integrin and D6.1A at the leading edge of the tumor cell.
Materials and Methods
Rats and tumor lines
BDX rats were obtained from WIGA, Sulzfeld, Germany, and were kept under conventional conditions. They were used for experiments at the age of 8-10 weeks. The pancreatic adenocarcinoma lines BSp73ASML (metastasizing) and BSp73AS (low metastasizing) are sublines from a tumor spontaneously arising in the BDX rat strain (Matzku et al., 1983). Regressor and Progressor are two related colon carcinoma lines, which metastasize via the lymphatic system (Reisser et al., 1993). The bladder carcinoma line 804G was a kind gift from Dr J. Jones, Northwestern University Medical School Chicago, IL, USA (Riddelle et al., 1991). 804G cells form hemidesmosomes in vitro and secrete laminin 5 (Ln5). BSp73AS cells, which express the α6 integrin chain were transfected with the β4 integrin chain cDNA using the pIRESneo vector and selection in G418 (BSp73AS-β4). Where indicated, the BSp73AS-β4 cells were cotransfected with the D6.1A cDNA in the pKEX2XR vector and selected in hygromycin (BSp73AS-db). Cells were transfected by electroporation (Eurogentech Easyject, 260V, 1050μF). Positive clones (defined by FACS analysis) were recloned under limiting dilution conditions. BSp73AS cells transfected with the D6.1A cDNA (BSp73AS-D6.1A) have been described elsewhere (Claas et al., 1998). D6.1A cDNA cloned from pcDNA3 vector was fused at its C-terminal to EYFP (enhanced yellow fluorescence protein) cDNA in the EYFP fusion vector (Clontech) and was used for transient transfection of 804G cells. All cells were cultured in RPMI 1640 supplemented with 10% FCS. Confluent cultures were trypsinized and split.
Antibodies and staining procedures
The mAb A2.6 (anti-CD44v6), D6.1 (anti-D6.1A), C4.4 (anti-C4.4A), D5.7 (anti-D5.7A) and B5.5 (anti-α6β4), A8.10 (undefined specificity) have been described previously (Matzku et al., 1989). Ox50 (anti-rat CD44), B2C11 (anti-rat CD9) and Ralph3.1 (anti-rat α3) were obtained from the European Collection of Animal Cell Cultures. Hybridoma culture supernatants were purified by protein G-Sepharose FPLC (flow pressure liquid chromatography. Where indicated, purified antibodies were labeled with biotin or fluoresceinisothiocyanate (FITC) or Texas Red (TxR). The rat integrin-specific antibodies anti-β1, -β2, -β4, anti-α1, -α2, -α3, -α4 and -α6, anti-rat CD9 and unlabeled, biotin- and horseradish peroxidase (HRP)-conjugated as well as dye-labeled (FITC, phycoerythrin (PE), Cy-2 and TxR) secondary antibodies and dye-labeled streptavidin as well as TRITC-labeled phalloidin were obtained commercially (Becton Dickinson, Heidelberg, Germany).
Flow cytometry followed routine procedures using 3-5×105 cells per sample. Trypsinized cell were allowed to recover for 2 hours at 37°C in RPMI 1640, 10% fetal calf serum (FCS). Samples were analyzed by FACSCalibur (Becton Dickinson, Heidelberg, Germany).
For immunofluorescence microscopy cells were seeded on cover slides, which had been pretreated with extracellular matrix (ECM) substrates. After spreading, slides were washed, cells were fixed in 4% paraformaldehyde (w/v in PBS) and, where indicated, were permeabilized (4 minutes in 0.1% Triton X-100). After washing and blocking [0.2% gelatin, 0.5% bovine serum albumin (BSA) in PBS], cells were incubated with the primary antibody at pretested concentrations (5-10 μg/ml) in PBS/BSA for 60 minutes at 4°C. Slides were rinsed and subsequently incubated for 60 minutes at 4°C with a fluorochrome-conjugated secondary antibody. After washing, cells were incubated with an excess of unlabeled mouse IgG to block potentially free binding sites of the dye-labeled anti-mouse IgG. Unlabeled mouse IgG was also added during incubation with the second, dye-labeled antibody. Cells were stained with a directly labeled second antibody or with phalloidin-TRITC (0.5 μg/ml) (F-actin) for an additional 60 minutes incubation at 4°C. For cross-linking experiments, cells were incubated at 37°C for 15 minutes with the primary antibody and for 20 minutes with an excess (10 μg/ml) of the secondary, dye-labeled antibody. Cells were washed with ice-cold PBS and all consecutive steps were performed at 4°C. After staining and washing 3 times in PBS and once in H2O, slides were mounted in Elvanol. Digitized images were generated using a Leica DMRBE microscope equipped with a SPOT CCD camera from Diagnostic Instruments Inc. and Software SPOT2.1.2, or using a confocal microscope. Where indicated, cells had been pretreated with phorbolmyristate acetate (PMA) (10–8 M) for 2 hours.
Fibronectin (Fn) and collagen IV (Col IV) were obtained commercially (Sigma, Deisenhofen, Germany). Recombinant human laminin-5 (Ln5) was expressed in HEK293 cells and purified from the conditioned medium by immuno-affinity chromatography as reported recently (Kariya et al., 2002). In some experiments 804G supernatant was used as a source of Ln5 (Riddelle et al., 1991). Plates were coated with 10 μg/ml Fn and Col IV, 0.3 μg/ml recombinant Ln5 or undiluted 804G supernatant. After coating, free binding sites were blocked by incubation with PBS/1% BSA.
Purification and cloning of the rat α6 and β4 integrin
Progressor cells (5×108) were lysed in 25 mM Hepes, pH 7.2, 1% Triton X-114, 500 mM NaCl, 1 mM CaCl2, 1 mM PMSF, diluted to 50 ml and was run over a B5.5-coated Sepharose 4B column. B5.5-bound material was eluted with 50 mM glycine buffer, pH 2.7, 500 mM NaCl, 0.1% Triton X-100 and was neutralized with 1 M Tris-HCl, pH 8.0. Fractions were analyzed by SDS-PAGE. Protein-containing fractions were pooled and precipitated by methanol/chloroform before being loaded on SDS-PAGE gels and stained with Coomassie Blue. Two bands at 200 kDa and 130 kDa were cut out, digested with trypsin and subjected to mass spectrometry.
To define the β4 isoforms expressed in the metastasizing cells lines BSp73ASML, Progressor and Regressor, an expression analysis by RT-PCR was performed using primers to detect the β4B (53aa insertion) and β4C (70aa insertion) isoforms (AGAGGCCCAGCGTTTCAG and <reverse> TACCCGGAACACATAGGAGTG) the primers spanning base pairs 4293-4957 and the β4D (7aa deletion) isoform (CTCTCTGCAGCTGAGCTGG and <reverse> TCACTGAGGCCAGGAACC) the primers spanning base pairs 5122-5280 (nucleotide numbers are from the rat β4 sequence u60096, which includes the +53 variant). Total RNA was extracted by the guanidine isothiocyanate/acid phenol method of Chomzynski and Sacchi (Chomzynski and Sacchi, 1987). cDNA was synthesized from 2 μg total RNA and subjected to PCR amplification using the indicated primers. Amplification was performed at 60°C for 30 cycles.
From the rat α6 integrin chain two short sequences were known, EST AA955091 and a 265 bp fragment kindly provided by Dr L. Feltri, Institute San Rafaele, Milano, Italy (Feltri et al., 1997). Primers binding to these sequences were used together with pcDNA3-specific primers in an attempt to clone the full-length cDNA from a library of the Regressor line (Claas et al., 1998). Since the 5′ end of the gene was not present in the library, it was cloned by the RACE technique using terminal transferase from 804G cells (submitted to EMB database, accession numbers AJJ12933 and AJJ12934).
For the transfection of BSp73AS cells with the β4 integrin chain the cDNA was cloned into the pIRESneo vector. Three clones spanning bp 0-3143 (F14), bp 1875-5455 (L7) and bp 3684-5896 (L5) in the pBluescript SK+ vector were kindly provided by Dr L. Feltri (Institute San Rafaele, Milano, Italy). F14 was cut out with EcoRV and HindIII, was blunted and ligated into pIRESneo, which had been digested with EcoRV. L7 was digested with ClaI and BamHI (bp1946-3626) and cloned into pIRESneo-F14. L5 was digested with BamHI (bp 3626-5896) and was ligated into the pIRESneo-F14/L7part. Integrity of the cDNA was controlled by re-cleaving after ligation and association of the clone with the α6 integrin and staining with B5.5 (see below).
For transfecting BSp73AS-β4 cells with D6.1A cDNA, pcDNA3-D6.1A and pKEX-2XR were digested with EcoRI and XbaI, blunted with Klenow fragment and digested again using NotI. Transfection was done by electroporation.
Cells (5×106) were lysed in ice-cold lysis buffer (25 mM Hepes, 100 mM NaCl; 1 mM CaCl2, 1 mM MgCl2, pH 7.2) containing 1% CHAPS. Lysis was performed for 1 or 4 hours at 4°C. Lysis buffers contained a protease inhibitor cocktail (Boehringer Mannheim). After centrifugation for 30 minutes at 13000 rpm cell lysates were subjected to immunoprecipitation.
Lysates were precleared by the addition of 10 μg/ml control antibody for 60 minutes followed by incubation with 1/10 volume protein G Sepharose for 2 hours at 4°C. Precleared lysates were incubated for 60 minutes at 4°C with 2 μg of antibody or control IgG. Protein G Sepharose was added for an additional 60 minutes. Immune complexes were washed 4-6 times with lysis buffer. Immunoprecipitated proteins were analyzed by SDS-PAGE, followed by Western blotting.
Lysates were resolved on 10-12% SDS-PAGE under reducing or non-reducing conditions and the proteins were transferred to Immobilon P at 90 V for 90 minutes. After blocking the membranes with 3% BSA, immunoblotting was performed by using the indicated antibodies, followed by donkey anti-mouse-HRP or donkey anti-rabbit-HRP. Blots were developed with the enhanced chemiluminescence detection system.
Cell spreading, adhesion, migration and proliferation
Cell spreading was induced by seeding the tumor cells (1×105/ml in RPMI 1640, 10% FCS) in Petri dishes or on cover slides, which had been precoated with Col IV or Fn or recombinant Ln5 or supernatant of the 804G cell line, which contains Ln5. Cells were incubated at 37°C for various times. Spreading was analyzed by light microscopy.
In adhesion assays, cells were incubated with [3H]thymidine for 16 hours. Cells were washed and seeded in triplicates on substrate-coated flat-bottom 96-well plates. Cells were incubated for 20-120 minutes. Plates were washed vigorously. The remaining adherent cells were detached by incubation in 0.2% trypsin. Plates were harvested using an automatic harvester and were counted in a β-counter. Where indicated, 10 μg/ml antibodies were added during incubation.
Cell migration was evaluated using a modification of the scratch assay. Petri dishes were coated with substrates as described above. Thereafter, the central area of the Petri dishes was covered with a cover slide (6 mm diameter). Tumor cells were seeded on substrate-coated Petri dishes in RPMI supplemented with 10% FCS. Upon reaching subconfluency, the cover slide was removed. Medium was sucked off, plates were washed to remove non-adherent cells and RPMI supplemented with 1% FCS and, where indicated, 10 μg/ml antibodies were added. Plates were incubated for 72 hours, washed, fixed and stained with Hematoxylin-Eosin. Mean values and standard deviations of the number of cells migrating in the originally cell-free area were evaluated by counting 10 fields of 1 mm2 directly at the boundary towards the originally cell-free area using an inverted microscope at 24, 48 and 72 hours after removal of the cover slide. Fast migrating cells were evaluated by counting cells in the central area of the removed cover slide. Values represent the mean of 3 independently performed experiments.
Proliferative activity was determined by [3H]thymidine incorporation (10 μCi/ml) after seeding 104 tumor cells in triplicates in 96-well flat-bottom plates and incubation for 8 hours at 37°C. Thereafter cells were centrifuged, medium was discarded and adherent cells were detached by trypsin treatment. Cells were harvested and [3H]thymidine incorporation was determined as described above.
Tumor cells (2×105) were injected either intrafootpad (i.f.p.) or intraperitoneally (i.p.). When tumors in the footpad reached a mean diameter of 0.5 cm, they were excised (Animal license 089/98). Animals were regularly controlled for the development of lymph node metastases and a palpable mass in the peritoneal cavity. Animals were sacrificed and analyzed macroscopically for the presence of metastases when they became anemic, short breathing or lost weight, or when a palpable mass in the peritoneal cavity or in lymph nodes reached a mean diameter of 2-3 cm.
Significance of differences was evaluated using the two-tailed Student's t-test.
Cloning of the rat α6 integrin chain
The mAb B5.5 recognizes a surface molecule on metastasizing rat tumor lines. Since attempts to clone the molecule by screening of an expression library of a metastasizing rat tumor line (Regressor) failed, we speculated that B5.5 may recognize a conformational epitope (see below). Hence, we isolated the molecule by affinity chromatography and used mass spectrometry for its identification.
Lysates of the metastasizing tumor line Progressor were loaded on a B5.5-coated Sepharose 4B column. Eluates were run on SDS-PAGE and showed two bands of 130 kDa and 200 kDa (Fig. 1A). Bands were cut out, digested with trypsin, and peptides were subjected to mass spectrometry. The 200 kDa band revealed 26 peaks, of which 18 matched the rat β4 integrin chain. From the rat α6 integrin cDNA only two short fragments were known. Yet, the 130 kDa band revealed 29 peptides, of which 16 matched the murine α6 integrin chain, suggesting the 130 kDa band to be the rat α6 integrin.
Two known rat α6 integrin cDNA fragments, ETS AA955091 and a 265 bp fragment, were obtained from the German Resource Center, Berlin, Germany, and from L. Feltri, Institute San Rafaele, Milano, Italy (Feltri et al., 1997). Primers designed from these fragments were used in combination with pcDNA3-specific primers to amplify the 5′ and 3′ ends of the α6 cDNA from the above mentioned library of the Regressor line (see Materials and Methods). The 5′ end of the gene was not present in the library and was cloned from 804G cells (see Materials and Methods). The full-length cDNA of the α6 integrin chain was sequenced (accession no. AJJ12933 and AJJ12934). The rat α6 integrin chain shows 94% and 86% identity to mouse and human cDNA, respectively. The homology at the protein level is 96% to mouse and 89% to human α6 (Fig. 2).
Expression of β4 in the tumor line from which the extracts were derived and in two additional metastasizing lines was verified by PCR. Primers for the PCR (see Materials and Methods) had been chosen such that splice variants of the β4 chain would have become apparent. All three metastasizing lines expressed the β4A isoform (DeMelker and Sonnenberg, 1999). The β4 integrin was not expressed by BSp73AS, a non-metastasizing subline of BSp73 (Fig. 1B).
To prove that the mAb B5.5 recognizes α6β4, the non-metastasizing BSp73AS line was transfected with the β4 cDNA. BSp73AS cells are not recognized by B5.5, but express α6β1, which implies that B5.5 does not recognize α6. After transfection of BSp73AS cells with β4 cDNA several clones were obtained that were stained by B5.5 (Fig. 1C). Expression of β1 was reduced in BSp73AS-β4 clones. This might be due to the efficient association of β4 with α6, a phenomenon already described (Shaw et al., 1996). It should be mentioned that the staining of BSp73AS-β4 cells with B5.5 does not allow any prediction on the binding epitope because β4 is only expressed in association with α6. Hence, B5.5 might recognize an epitope on β4 or an epitope made up by α6 and β4.
Expression of α6β4 on the non-metastasizing BSp73AS line
Functional activity of α6β4 was first explored in BSp73AS cells transfected with β4 cDNA. Expression of α6β4 on BSp73AS cells was accompanied by a change in cell shape. The epitheloid-like BSp73AS cells became spindle shaped, and staining of the actin cytoskeleton revealed the formation of stress fibers (Fig. 3A).
However, expression of α6β4 in BSp73AS cells had no impact on metastasis formation after intrafootpad tumor cell application (data not shown). Also, the change in cell shape was not accompanied by altered adhesion to either plastic, Fn or Col IV. Adhesion to Ln5, the major ligand of α6β4, was only slightly improved as compared to BSp73AS cells and only after a short incubation period. However, the metastasizing lines adhered better to Ln5 than to plastic (Fig. 3B). Also, α6β4 expression on BSp73AS cells had no impact on migration on Ln5. However, in the presence of B5.5, migration of BSp73AS-β4 cells was slightly improved. The metastasizing Progressor cells readily migrated on Ln5-coated plates. Particularly at later time points, the migratory capacity on Ln5 was strengthened in the presence of B5.5 (Fig. 3C, Table 1). Thus, the question arose of why expression of α6β4 on BSp73AS-β4 cells, distinct from BSp73ASML and Progressor cells, had little bearing on adhesion to and migration on Ln5.
Co-localization of α6β4 with the tetraspanin D6.1A
Because expression of α6β4 on BSp73AS-β4 cells had no impact on metastasis formation or adhesion to Ln5, we speculated that in metastasizing lines α6β4 may exert functional activity in concert with additional molecules not present in the non-metastatic BSp73AS line. We first evaluated by antibody cross-linking whether α6β4 would co-cluster with additional molecules highly overexpressed on metastasizing tumors. Cells were seeded on cover slides and allowed to adhere overnight. Thereafter, cells were incubated with either A2.6 (anti-CD44v6), C4.4 (recognizing the uPAR-related molecule C4.4A), D6.1 (recognizing the tetraspanin D6.1A), D5.7 (anti-EpCAM) or anti-CD9. Bound antibodies were cross-linked with an excess of FITC-labeled anti-mIgG. Thereafter, cells were washed and fixed. Free binding sites of FITC-labeled anti-mouse IgG were blocked by incubation with an excess of unlabeled mouse IgG. Cells were then stained with TxR-labeled B5.5. In fact, cross-linking of CD44v6 and of the tetraspanin D6.1A was associated with coclustering of α6β4 on BSp73ASML (data not shown) and Progressor cells. Very few α6β4 molecules were detected in clusters of EpCAM (D5.7A) or the uPAR-related molecule C4.4A. Interestingly, α6β4 also hardly co-clustered with the tetraspanin CD9 (Fig. 4,Fig. 4). In contrast, after cross-linking of α6β4, CD44v6 and D6.1A, but not EpCAM or C4.4A co-clustered with α6β4. However, it should be mentioned that a considerable amount of D6.1A was not detected in α6β4 clusters (data not shown).
Because tetraspanins associate in particular with integrins (reviewed by Maecker et al., 1997), we focussed on the co-localization of α6β4 and D6.1A. These studies were performed with metastasizing lines and BSp73AS cells transfected with β4 and D6.1A cDNA (BSp73AS-db). Several BSp73AS-db clones were established. The integrin and tetraspanin expression profile of one characteristic clone in comparison to BSp73AS, BSp73AS-β4, BSp73AS-D6.1 and the metastasizing lines BSp73ASML, Progressor and 804G is listed in Table 2. All lines express α3, α6 and β1 at a high level. Only the metastasizing lines (and transfected lines) express α2 and β4. With the exception of a downregulation of β1 in BSp73AS-β4 cells and a slight upregulation of α3 in BSp73AS-db cells, transfection of the non-metastasizing line with β4 and D6.1A cDNA did not influence integrin expression. Only the metastasizing lines (and by transfection the BSp73AS-D6.1A, BSp73AS-db lines) express the tetraspanin D6.1A. But, all lines express the tetraspanins CD151 (unpublished finding) and CD9.
To test, whether D6.1A and α6β4 also co-localize in the BSp73AS-db line, cells were seeded on Ln5-coated cover slides and were allowed to spread for 48 hours before staining with B5.5/anti-mIgG-FITC and D6.1-TxR or D6.1/anti-mIgG-FITC or anti-CD9/anti-mIgG-FITC and B5.5-TxR. In BSp73AS-db cells, α6β4 mostly co-localized with D6.1A. D6.1A also co-localized with α6β4, although a considerable proportion of D6.1A was seen outside of α6β4 clusters. Only very few α6β4 molecules co-localized with CD9 (Fig. 5).
We next tested whether co-localization of D6.1A and α6β4 could be verified by co-immunoprecipitation. Progressor cells were lysed for 1 hour at 4°C in the mild detergent CHAPS. Lysates were precipitated with anti-β4, anti-α3, anti-β1 and D6.1 and were blotted with D6.1. D6.1A clearly co-precipitated with β4 and also with α3 and β1, which has been shown before (Claas et al., 1998) (Fig. 6A). To control for the specificity of co-precipitation, Progressor cells were biotinylated and precipitated with C4.4 (anti-C4.4A <uPAR-related>), D5.7 (anti-EpCAM), Ox50 (anti-panCD44), anti-CD9, anti-β4, anti-α6, B5.5, D6.1 and A8.10. CD44, EpCAM and C4.4A are highly expressed on metastasizing lines. mAb A8.10, which recognizes a molecule only detected on BSp73AS cells, served as a negative control. After gel separation, all precipitates were blotted with streptavidin, D6.1 and anti-CD9. D6.1A co-precipitated with α6β4 (B5.5), β4 and CD9. Smaller amounts of D6.1A were recovered in the α6 and the CD44 precipitate. It should be mentioned that Progressor cells hardly express the CD44 standard isoform, but large amount of CD44v6, i.e. small amounts of D6.1A co-precipitate with CD44 variant isoforms. There was a very faint co-precipitate with D5.7A (EpCAM), but none with C4.4A and with the control antibody A8.10. A different picture emerged when the precipitates were blotted with anti-CD9. CD9 was only recovered in D6.1 and anti-CD9 precipitates (Fig. 6B). Notably, α6β4 and D6.1A did not co-immunoprecipitate after lysis in strong detergents like Triton X-100, which argues against a direct association of the two molecules (data not shown).
Since the B5.5 and the anti-β4 precipitates contained only part of D6.1A, the specificity of co-precipitation of the two molecules was controlled by several means. First, using lysates of Progressor cells (CHAPS, 4 hours, 4°C) the β4 could be re-precipitated from precipitates with B5.5 or D6.1 but not Ox50 or A8.10, although high amounts of precipitate were required to reach the detection limit. Second, when mixing BSp73AS-D6.1A and BSp73AS-β4 cells, the B5.5 precipitate of the lysate (CHAPS, 4 hours, 4°C) did not contain D6.1A. Yet, it did so when BSp73AS-db cells were lysed (data not shown).
Thus, a small proportion of α6β4 and D6.1A can be co-precipitated when using mild detergent. To obtain further information on the conditions for this association, we compared co-localization of the two molecules by immunofluorescence microscopy under resting and stimulatory conditions.
Non-random co-localization and redistribution of α6β4 and D6.1A after PKC activation
PKC was described to regulate integrin-dependent cell motility (Ng et al., 1999) and to facilitate cell migration (Maniero et al., 1996). Also, the motility of keratinocytes was modulated when α6β4 associated with CD9 (Baudoux et al., 2000). Thus, it became of interest to determine whether PKC activation would be accompanied by changes in co-localization of α6β4 and D6.1A. The distribution/co-distribution of α6β4 and D6.1A was evaluated by immunofluorescence in metastatic lines and BSp73AS-db cells. Cells were seeded on Ln5-coated cover slides and were allowed to spread for 48 hours. Thereafter cells were treated for 30 minutes to 2 hours with PMA. Cells were fixed and permeabilized for between 30 minutes and 10 hours after PMA addition. Cell were stained immediately after fixation and permeabilized with either phalloidin-TRITC or were stained with D6.1 or anti-CD9 and were counterstained with B5.5.
PMA treatment was accompanied by changes in the actin cytoskeleton. Growing actin bundles were already seen 1 hour after PMA addition in BSp73AS-db cells. Cells developed long filipodia and spikes. Later on, actin bundles were seen at the front of the leading lamella. Long actin fibres were no longer seen 10 hours after PMA treatment, but organization in fibres was still more pronounced than in untreated cells. Although actin bundle formation was not seen in PMA-treated Progressor cells, the actin distribution clearly changed after PMA treatment. The equal distribution seen in untreated cells was lost, actin was enriched in filipodia and later on it clustered in the body of the cells. From there it moved towards the leading lamella. At 6 hours after PMA treatment actin was strongly enriched in the leading lamella and at 10 hours after PMA treatment at the front of the leading lamella (Fig. 7A).
D6.1A and α6β4 were more or less equally distributed after spreading of Progressor cells on Ln5 (Fig. 6B1). This accounted also for BSp73ASML and BSp73AS-db cells (data not shown). During PMA treatment, Progressor cells developed long filipodia, which were strongly stained by B5.5 and D6.1, whereas the body of the cells became almost devoid of integrin and tetraspanin. After PMA removal, the tetraspanin and α6β4 were transiently seen in the body of the cells (4 and 6 hours). Ten hours after PMA treatment, lamellae were brightly stained by B5.5 and D6.1. The apparent internalization was not restricted to α6β4 and D6.1A. Membrane staining of EpCAM was strongly reduced (data not shown) and CD9 rapidly and completely disappeared from the cell membrane. CD9 became strongly enriched in intracellular bodies for up to 8 hours. When PMA-treated and permeabilized Progressor cells were stained with anti-CD9 and B5.5, yellow areas were seen in digital overlays, which could indicate co-localization of CD9 and α6β4. B5.5 staining of the leading lamella was more pronounced at 10 hours than at 8 hours after PMA treatment. At 10 hours after PMA treatment, CD9 had also reached the front of the leading lamella, however, α6β4 and CD9 did not generally co-localize at the migratory front of the cells (Fig. 7B2). EpCAM became equally redistributed in the cell membrane and there was no evidence for a selective co-distribution of EpCAM and the α6β4 integrin (data not shown).
A similar redistribution of a6β4 and D6.1A as described for Progressor cells was seen in BSp73ASML and BSp73AS cells. BSp73ASML cells, which appear round and `unstructured' developed protrusions and lamellae. Protrusions and lamellae were first largely devoid of the integrin and the tetraspanin, which were concentrated intracellularly. Already 6 hours after PMA treatment, lamellae were stained by B5.5 and D6.1. In PMA-treated BSp73AS-db cells co-localization of the integrin and the tetraspanin was mainly seen in spikes and filipodia. Later on, both molecules disappeared from the cell surface and co-localization in the leading lamella was seen only after 10 hours (data not shown).
The redistribution of α6β4 and D6.1A after PMA treatment was also seen in 804G cells, which were transiently transfected with D6.1A-EYFP cDNA. In the resting state, only a very minor part of D6.1A and α6β4 co-localized in filipodia. After 30 minutes PMA treatment D6.1A and α6β4 readily colocalized. After 2 hours of treatment, cells started to migrate with long filipodia at the trailing edge, which was strongly stained by B5.5. D6.1A co-localized with α6β4 in the trailing edge and in the main body of the cell. As revealed by confocal microscopy at the level of adhesion and 1 μm above (data not shown) and by a sagittal section, the α6β4-D6.1A complexes forming after PMA treatment were found intracellularly (Fig. 7C). Evidence for cell migration (traces left behind by α6β4-containing filipodia) (data not shown). Unlike Progressor cells, extension of lamellae could already be seen 2 hours after PMA treatment in 804G cells, at which time the leading lamellae were still mainly devoid of α6β4.
It was also of interest to see whether α6β4 and D6.1A would re-distribute in PMA-treated BSp73AS cells expressing either α6β4 or D6.1A. PMA-induced morphological changes of BSp73AS-D6.1A and BSp73AS-β4 cells were less pronounced than of BSp73AS-db cells. Yet, it appeared that D6.1A was enriched intracellularly after PMA treatment even in the absence of α6β4. BSp73AS-β4 cells developed thin filipodia, which were transiently stained by B5.5. During the recovery period α6β4 was mainly detected intracellularly. Notably, even 12 hours after PMA treatment, α6β4 did not become enriched in filipodia, protrusions or lamellae as it was seen in Progressor, BSp73ASML and BSp73AS-db cells (data not shown).
Thus, PKC stimulation induced changes in morphology and was accompanied by a redistribution of α6β4 and of α6β4-D6.1A `complexes'. The redistribtion of α6β4-D6.1A `complexes' started with a transient and co-ordinated `internalization' of the molecules. Internalization was also seen in BSp73AS-β4 and BSp73AS-D6.1A cells and PMA-induced internalization was not restricted to α6β4 and D6.1A, i.e. CD9 also became internalized. However, re-expression of α6β4 in the leading lamella was only seen in metastasizing lines and BSp73AS-db cells and was accompanied by co-localization with D6.1A.
D6.1 interferes with functional activity of α6β4
If our hypothesis holds true that the α6β4-tetraspanin complex is involved in cell motility, blockade of D6.1A could possibly interfere with adhesive and migratory functions of α6β4.
It has already been shown that expression of β4 on BSp73AS cells hardly influenced binding to and migration on Ln5. The same was true for BSp73AS-D6.1A cells (data not shown). Yet, BSp73AS-db cells adhered significantly more strongly to Ln5 than to BSA. Because BSp73AS-db cells express α3 at a slightly increased level as compared to BSp73AS, we first determined whether the increased Ln5 binding could be α3 mediated. Ln5-binding of BSp73ASML, BSp73AS-db, and also of AS-β4 and AS-D6.1 was slightly reduced in the presence of anti-α3. Yet, even in the presence of anti-α3, an increased percentage of BSp73AS-db adhered to Ln5-coated wells. Furthermore, B5.5 and D6.1, neither of which influenced adhesion of BSp73AS-β4 and BSp73AS-D6.1A cells to plastic or Ln5, significantly inhibited adhesion of BSp73AS-db cells to Ln5. The same pattern of inhibition of adhesion, though less pronounced, was seen with BSp73ASML and Progressor cells (Fig. 8A). No uniform pattern of adhesion/inhibition of adhesion was seen in PMA-treated cells, which might be due to alteration in expression/configuration of additional adhesion molecules by PKC activation and also to the transient internalization of D6.1A and α6β4 (data not shown).⇓
B5.5 had only a minor influence on migration of BSp73AS-β4 cells (Fig. 3C and Table 1) and D6.1 hardly influenced migration of BSp73AS-D6.1A cells on Ln5-coated plates. Instead, migration of BSp73AS-db and Progressor cells was clearly improved in the presence of B5.5 or, more pronounced, D6.1. Notably, the effect was not seen during the first 24 hours (Tables 2 and 4). Increased migration was not due to an increase in proliferative activity, i.e. [3H]thymidine incorporation in BSp73AS-β4, BSp73AS-D6.1A, BSp73AS-db, BSp73ASML and Progressor cells did not differ significantly when cells were cultured for 8 hours in the presence or absence of B5.5 or D6.1 (data not shown). PMA treatment, which also had no major impact on tumor cell proliferation (data not shown), strongly supported migration of BSp73AS-db and Progressor cells, while exerting a weaker, though significant effect on BSp73AS, BSp73AS-β4 and BSp73AS-D6.1A cells. It should be mentioned that BSp73AS-db and Progressor readily spread all over the area initially protected by the cover slide (6 mm diameter), whereas BSp73AS, BSp73AS-β4 and BSp73AS-D6.1A cells moved more slowly and stepwise (Table 3, Fig. 8B).
The PMA-induced increased motility of BSp73AS-db cells and metastasizing tumors was in line with the microscopically observed changes in cell shape and the redistribution of α6β4 and D6.1A. The observation that co-expression of α6β4 and D6.1A strongly supported migration on Ln5 supported our concept of the linked activity of the two molecules. Thus, it appeared worthwhile to ask whether co-expression of the two molecules might support metastasis formation.
Co-expression of α6β4 and D6.1A supports liver metastasis formation
The α6β4 integrin may be important for the early metastatic spread of pancreatic adenocarcinoma (Hermanek, 1998; Vogelmann et al., 1999) and metastastic spread of pancreatic tumors is most pronounced when cells settle in the peritoneal cavity (Z'graggen et al., 2001). One possible explanation could be that in the stimulatory surrounding of the peritoneal cavity the tetraspanin-induced redistribution of α6β4 accounts for increased metastatic capacity. This hypothesis was tested using BSp73AS tumor clones that differ only in the expression of α6β4 and D6.1A.
BSp73AS, BSp73AS-β4, BSp73AS-D6.1A and BSp73AS-db were injected intraperitoneally (Table 4). BSp73AS or BSp73AS-β4 grew only locally and developed small amounts of non-hemorrhagic ascitic fluid. BSp73AS-D6.1A cells also formed solid nodules in the peritoneal cavity, but developed hemorrhagic ascitic fluid, though only a small amount. Importantly, eight out of 10 rats showed lung metastases and four out of 10 rats liver metastases, 1 rat in each group developed metastases in the kidney and the spleen. The intraperitoneal growth of BSp73AS-db cells differed from the other lines in as much as the tumor formed miliar nodules and massively infiltrated the pancreatic gland. Animals developed high amounts (up to 30 ml) of hemorrhagic ascitic fluid and a massive infiltration of the diaphragm such that tumor cell-containing pleural effusions were generated in 10 out of 12 rats. Lung metastases were only seen in 1 rat, but 11 out of 12 rats developed multiple liver metastases. Thus, α6β4 by itself does not contribute to tumor progression, whereas the D6.1A tetraspanin supports metastasis formation. However, association of the two molecules increases the metastatic potential (miliary metastases) and is accompanied by a preferential settlement in pancreatic gland and liver.
The α6β4 integrin has repeatedly been reported to be involved in tumor progression. Depending on the tumor type, down- or upregulation has been described (reviewed by Mercurio and Rabinovitz, 2001). It is particularly the latter feature that appears surprising because matrix adhesion, e.g. of keratinocytes (Borradori and Sonnenberg, 1999), is one of the main functions of α6β4. We had noted high expression of a molecule recognized by the mAb B5.5 exclusively on metastasizing, as opposed to non-metastasizing, tumors of the rat (Claas et al., 1996; Matzku et al., 1989). The molecule was cloned and identified as α6β4. Transfection of a non-metastasizing line, which is devoid of β4, with the β4 cDNA led to expression of the α6β4 heterodimer, but this did not transfer the metastatic phenotype. However, liver metastases were abundantly observed when the non-metastasizing line co-expressed – after cDNA transfection – α6β4 and the tetraspanin D6.1A. Hence, we assume that it is an association between α6β4 and tetraspanins rather than α6β4 by itself that supports metastasis formation.
The α6β4 integrin was purified by affinity chromatography. Two bands of 130 and 200 kDa emerged and were subjected to trypsin digestion. The 200 kDa band represented the β4 integrin (Feltri et al., 1997). While human pancreatic and colon carcinoma lines mostly express the β4C variant isoform (Aplin et al., 1996; Fornaro and Languino, 1997), the three metastatic rat lines expressed the β4A chain. Functional differences between these splice variants are not yet known.
The 130 kDA band showed a high degree of homology to the murine and human α6 integrin. Because the rat α6 cDNA was unknown at the time, the molecule was cloned and sequenced. When a non-metastasizing line, which expresses α6β1, was transfected with the β4 cDNA, the cell line readily expressed α6β4, i.e. cells were stained by the B5.5 mAb, which confirmed the identity of the molecule as α6β4. A preferential association of α6 with β4 as compared to β1 has already been described (Shaw et al., 1996). Our data confirm this association.
Expression of α6β4 in BSp73AS cells was accompanied by a change in cell shape. The epitheloid cells became spindle shaped and formed extremely long and very thin filipodia when seeded on substrates. Changes in cell shape were accompanied by the formation of strong actin bundles. Double staining for α6β4 and the actin cytoskeleton revealed co-localization of these 2 molecules mainly in spikes and filipodia (data not shown). Although in epithelial cells α6β4 was described not to associate with the actin cytoskeleton (Fontao et al., 1997; Grassi et al., 1999), such an association has been seen in, for example, transfected fibroblasts (Rabinovitz et al., 1997).
Although expression of α6β4 in BSp73AS cells had a bearing on cell shape and actin bundle formation, it hardly influenced adhesion to/migration on Ln5 and did not support metastasis formation. Therefore, we speculated that additional molecules expressed by metastasizing lines may associate with α6β4 and modulate its activity. Antibody-cross-linking experiments with four molecules, which are highly expressed on metastasizing, but not on non-metastasizing lines, provided evidence for an association of α6β4 with a CD44 variant isoform and with D6.1A. The α6β4 hardly co-localized with EpCAM (D5.7A) and not with the uPAR-related C4.4A molecule. Because of the close relationship between tetraspanins D6.1A and CD9 (Claas et al., 1998) and because CD9 was described as associating with α6β4 (Baudoux et al., 2000), we also evaluated whether in the metastasizing rat tumor lines α6β4 may co-localize with CD9. This has not been the case. Although co-localization does not necessarily imply complex formation, co-immunoprecipitation after lysis in mild detergents confirmed the findings in as much as part of D6.1A was precipitated with anti-β4 and B5.5 as well as with A2.6. D6.1A was also precipitated with anti-CD9 and vice versa. But, neither anti-β4-nor B5.5-precipitates contained CD9. The apparent selectivity of the α6β4-D6.1A co-localization and the fact that tetraspanins are known to modulate the functional activity of integrins (Shaw et al., 1997) (reviewed by Mercurio and Rabinovitz, 2001), tempted us to focus on potential consequences of the co-localisation of α6β4 and D6.1A.
As already mentioned, α6β4 can associate with the tetraspanin CD151, and CD151 associates with α6β4 within hemidesmosomes (Sterk et al., 2000). This has not been the case for the co-localization of the rat α6β4 with D6.1A. Neither BSp73ASML nor Progressor cells form hemidesmosomes when grown on plastic, expression of α6β4 is not polarized and the relative amount of co-precipitating α6β4/D6.1A is not altered after PKC treatment (data not shown). In the bladder carcinoma line 804G, which forms hemidesmosomes and secretes Ln5 (Riddelle et al., 1991), expression of α6β4 is polarized. Hemidesmosomes did not contain D6.1A and co-localization/co-immunoprecipitation of α6β4 and D6.1A was only seen after disintegration of hemidesmosomes by PMA treatment (M. Herlevsen, PhD Thesis, University of Karlsruhe, Germany, 2001). Furthermore, even in BSp73ASML and Progressor cells only part of D6.1A and α6β4 co-immunoprecipitated and only using a mild detergent. This is in line with the observation that particularly after cross-linking of α6β4 on the metastasizing lines only part of D6.1A co-clustered with α6β4 (data not shown). Though an artefact was excluded by appropriate controls, these features strongly argue against a direct association between D6.1A and α6β4. Because the association of α6β4 with D6.1A was destroyed by strong detergents and since tetraspanins are known to form multimolecular tetraspanin webs (Charrin et al., 2001; Claas et al., 2001; Hemler, 2001; Levy et al., 1998; Maecker et al., 1997), we speculate that (outside hemidesmosomes/after disruption of hemidesmosomes) α6β4 may associate via a linker molecule with D6.1A. Work is in progress to identify such a linker protein; our studies so far only excluded CD9 as a potential candidate. As outlined above, a considerable amount of D6.1A was recovered in anti-CD9 precipitates and D6.1A also co-localized with CD9 (data not shown). However, CD9 apparently is not the linker between D6.1A and α6β4 because CD9 failed to co-precipitate with β4/α6β4. Because, in contrast, D6.1A co-precipitates with CD9 as well as with α6β4/β4, it is tempting to speculate that D6.1A may be found in distinct, non-overlapping membrane protein complexes. Irrespective of this open question, it is important to note that it is the association between α6β4 and D6.1A that has a bearing on cell adhesion, migration and metastasis formation. This assumption derives from the observation that (i) BSp73AS cells also express the α3β1 integrin and the tetraspanins CD9 and CD151, but (ii) α6β4 expression on BSp73AS cells hardly influenced adhesion and migration and had no impact on metastasis formation. Hence, even if D6.1A does not directly associate with α6β4, current evidence point towards the interaction between these 2 molecules being important for induced cell motility.
A higher percentage of BSp73AS-db cells adhered to Ln5-coated than to uncoated plates. Importantly, binding to Ln5 was inhibited by B5.5 and D6.1 when cells expressed both, but not when cells expressed either α6β4 or D6.1A. It should be noted that binding to BSA and to a commercially available Ln preparation, which contains unknown amounts of Ln5, was not inhibited (Claas et al., 1998) (data not shown). Antibody-mediated inhibition of binding was probably not due to blocking of an ECM binding site, because B5.5 should have blocked such an ECM binding domain independent of whether the cells expressed α6β4 or α6β4 plus D6.1A. Furthermore, tetraspanins apparently do not possess binding sites for ECM proteins. Therefore, we interpret the findings in the sense that (i) both antibodies interfere with the organization of a tetraspanin web including α6β4 and D6.1A and (ii) binding of α6β4 to Ln5 in cells not forming hemidesmosomes may depend on the formation of such tetraspanin webs.
Whereas D6.1 and B5.5 inhibited adhesion, the antibodies supported migration of double-, but not of single-positive cells on Ln5, albeit weakly. One possible explanation could be that the two phenomena, inhibition of binding and support of migration, are linked such that weakening of adhesion results in pronounced migration. We consider an alternative explanation as more likely. Cross-linking of the molecules via the antibody could well support internalization, which for PMA treatment has been shown to induce a migratory phenotype. The fact that the most dramatic changes in the migratory capacity were seen when α6β4-positive plus D6.1A-positive cells were treated with PMA supports the interpretation. After PKC stimulation both molecules were first enriched at the trail of the cell. Thereafter, they largely disappeared from the cell surface and were later seen in lamellipodia. Importantly, expression of α6β4 in lamellipodia was only seen after PMA treatment and only in cells expressing both α6β4 and D6.1A. Thus, PKC stimulation and, probably, antibody cross-linking induces transient internalization of the α6β4-tetraspanin complex, which could well interfere with adhesion and support cell motility. Whether the internalized complexes are re-expressed or whether new complexes are formed at the leading lamella remains to be explored.
There is increasing evidence that α6β4 outside hemidesmosomes is involved in actin rearrangement and cell migration (Maniero et al., 1996; Mercurio and Rabinovitz, 2001; Odintsova et al., 2000; Rabinovitz and Mercurio, 1997; O'Connor et al., 2000) and thereby influences invasiveness (Daemi et al., 2000; Gambaletta et al., 2000; Shaw et al., 1997). Considering the BSp73AS-db line, our data would certainly support the induction of the migratory phenotype by an association of α6β4/the α6β4-D6.1A complex with the actin cytoskeleton. Although we did not observe actin bundle formation in PMA-treated Progressor cells, there has been a redistribution of actin filaments after PMA treatment, which has been similar to the redistribution of α6β4 and D6.1A. Because Progressor cells also gained in motility and internalized the α6β4-D6.1A complex after PMA treatment, it remains to be explored whether actin bundle formation is essential for a gain in cell motility. Independent of this open question, our findings support the idea that a concerted activity of the two molecules is important for tumor cell motility.
An increase in tumor cell motility by the D6.1A-α6β4 `complex' was also supported by the finding of miliar metastasis formation of BSp73AS-db cells after intraperitoneal seeding. There is evidence that the intraperitoneal seeding of pancreatic adenocarcinoma cells is decisive for early metastatic progression (Z'graggen et al., 2001). In fact, the metastasizing subline BSp73ASML, expressing both α6β4 and D6.1A, was derived from the ascitic fluid of a rat with a pancreatic adenocarcinoma, whereas the non-metastasizing subline BSp73AS was derived from the locally growing tumor (Matzku et al., 1983). Neither BSp73AS nor BSp73AS-β4 (α6β4+) cells displayed metastatic growth when implanted intraperitoneally. However, BSp73AS-D6.1A (D6.1A+) cells readily metastasized to the lung. Thus, expression of the tetraspanin triggered tumor cell spreading via the lymphatic route. Co-expression of α6β4 (BSp73AS-db) additionally supported tumor cell dissemination as was apparent by the formation of miliary nodules in the peritoneal cavity and the massive infiltration of the diaphragm. Co-expression of α6β4 also played a decisive role with respect to the organ preference as evidenced by the reduction in lung metastasis and the massive infiltration of the pancreatic gland and the liver. Because, firstly, in vitro co-localization of α6β4 and D6.1A was strongly influenced by PKC activation and, secondly, metastasis formation depended on the intraperitoneal application of the tumor cells (i.e. neither pancreatic gland nor liver metastases were seen after subcutaneous tumor cell implantation; data not shown), we hypothesize that the intraperitoneal microenvironment supports metastasis formation by initiating co-localization and transient internalization of α6β4 and D6.1A. We propose, therefore, that it is not α6β4 by itself but α6β4 in association with tetraspanins that facilitates tumor cell motility and, as a consequence, metastasis formation. Transition into a motile state/increased metastasizing capacity can be induced by PKC activation, which supports a α6β4-tetraspanin complex formation, transient internalization of the complexes and, later on, expression at the leading front of the tumor cell.
This work was supported by the Tumorzentrum Heidelberg/Mannheim and the Deutsche Forschungsgemeinschaft (Zo40-8/1) (M.Z.). We greatfully acknowledge receive of rat α6 and β4 cDNA by Dr L. Feltri, Institute San Rafaelo, Milano, Italy. We also want to thank Prof. Dr J. Jones, Northwestern University Medical School, Chicago, USA, for the kind gift of 804G cells. We are most grateful to Dr M. Schnölzer, German Cancer Research Center for mass spectrometry, Dr Spring, German Cancer Research Center, for help of with confocal microscopy and G. Devitt for correction of the English.
↵* Present address: Department of Urology, PO Box 800422, Charlottesville, VA, USA
- Accepted July 8, 2003.
- © The Company of Biologists Limited 2003