Publication of the International Union Against Cancer Publication de l’Union Internationale Contre le Cancer Int. J. Cancer: 80, 791–795 (1999) r 1999 Wiley-Liss, Inc. INTEGRIN ␣6␤1 ROLE IN METASTATIC BEHAVIOR OF HUMAN PANCREATIC CARCINOMA CELLS Roger VOGELMANN1, Ernst D. KREUSER2, Guido ADLER1 and Manfred P. LUTZ1* 1Department of Internal Medicine I, University of Ulm, Ulm, Germany 2Department of Hematology and Oncology, University Medical Center Benjamin Franklin, Free University, Berlin, Germany The factors that determine the metastatic behavior of pancreatic tumor cells are incompletely understood. In this study, we first demonstrate differences in adhesion properties, integrin expression and in vivo integrin function in the metastatic tumor cell line PaTu 8988s compared with the non-metastatic cell line PaTu 8988t. Both cell lines were derived from the same original tumor and exhibit identical genetic fingerprints. Using in vitro adhesion assays performed on purified extracellular matrix components, adhesion of PaTu 8988s cells was significantly increased on the basal membrane component laminin and decreased on the interstitial matrix protein fibronectin compared to PaTu 8988t cells. By immunocytochemistry and flow cytometry, and in correspondence with their adhesive properties, the metastatic PaTu 8988s cells did express a distinct pattern of integrin subunits. Laminin-binding integrins ␣6 and ␤4 were overexpressed in PaTu 8988s cells. Fibronectin-binding ␣5 integrins were present at higher levels in the non-metastatic PaTu 8988t cells, whereas the ␤1 subunit expression did not differ. Adhesion to laminin or fibronectin was specific and was mediated via integrins ␣6␤1 and ␣5␤1, respectively. In addition, metastasis formation in vivo after injection of cells into the tail vein of nude mice was inhibited by preincubation of PaTu 8988s cells with antibodies directed against the integrin ␣6 or ␤1. We conclude that ␣6␤1 integrins are overexpressed and functionally active in metastatic human pancreatic carcinoma cells, and participate in metastasis formation probably through binding to the basal membrane component laminin. Int. J. Cancer 80:791–795, 1999. r 1999 Wiley-Liss, Inc. Pancreatic cancer is a major cause of cancer-related deaths in Western countries. Despite intensive efforts to improve therapy of advanced disease, palliation remains unsatisfactory and most patients die within months due to rapid locoregional growth of the tumor or early metastasis formation. The biological characteristics that are responsible for the aggressive behavior of these tumors are not well understood. Tumor metastasis is considered to be a multistep process. During this process, tumor cells must detach from the primary tumor, migrate through adjacent tissues and enter the circulation where they face the host defenses. In addition, metastasizing cells need to arrest at a distant vascular site, cross the basal membrane, which functions as a barrier between circulation and target organ, and be able to colonize and grow under these novel conditions. Many of these steps require changes in the adhesive properties of cells to the extracellular matrix (Heino, 1996). Cell adhesion to extracellular matrix is mediated at least in part by the integrin family of transmembrane receptor proteins. Integrins are dimeric proteins composed of non-covalently associated ␣ and ␤ subunits. Theoretically, the 8 known ␤ subunits can combine with one of at least 16 ␣ subunits, resulting in integrin heterodimers with distinct adhesion properties. The integrins are divided into subgroups depending on their binding preferences to extracellular matrix proteins or to cell surface molecules. To date, 22 different ␣␤ complexes have been described, with 15 of these acting as receptors for fibrous protein components of the extracellular matrix (Heino, 1996). These matrix proteins are classified into structural proteins (e.g., collagen or elastin) or proteins with mainly adhesive properties (e.g., fibronectin or laminin). The adhesion of cells to interstitial matrix is mediated by fibronectin together with collagen type I as structural component, whereas laminin is the predominant adhesive protein in the basal membrane together with collagen type IV as a backbone. Decreased expression of fibronectinbinding integrins or increased expression of laminin-binding proteins correlates with aggressive growth and with the ability to metastasize in several tumor tissues (Shaw et al., 1996; Schreiner et al., 1991; Giancotti and Ruoslahti, 1990; Varner et al., 1995). In the pancreas, the expression of several integrin subunits has been described. These include the fibronectin-binding integrin ␣5; the laminin-binding integrins ␣2, ␣3 and ␣6; and the vitronectinbinding ␣v together with the ␤1, ␤4 and ␤5 subunits. In general, expression of the fibronectin-binding ␣5 subunit appears to be decreased in tumor tissues (Hall et al., 1991). Using functional assays, pancreatic tumor cells can attach to laminin, fibronectin or collagens (Weinel et al., 1995; Löhr et al., 1996), and adhesion to laminin appears to be mediated by the laminin-specific integrin ␣6␤1 (Weinel et al., 1995; Rosewicz et al., 1997). Integrin expression does not correlate with tumor differentiation, and the number of tumors or cell lines examined has been too small to define the changes relevant to the growth characteristics or the metastatic behavior of these cells (Hall et al., 1991; Rosendahl et al., 1993; Löhr et al., 1996). Therefore, we used 2 closely related cell lines, both derived from the same original tumor, but differing in metastatic ability (Elsässer et al., 1992) to examine adhesion properties, integrin expression and integrin function. Using expression studies and adhesion assays on purified matrix components, we could define a metastasisrelated expression pattern of functionally intact integrins with binding specificity for laminin. In addition, inhibition experiments with integrin-specific antibodies demonstrated a role for the laminin-binding integrins ␣6␤1 in experimental in vivo metastasis. MATERIAL AND METHODS Cell lines The human pancreatic carcinoma cell lines PaTu 8988s and PaTu 8988t (Elsässer et al., 1992) were obtained from the Deutsche Sammlung für Mikroorganismen (Braunschweig, Germany). Both cell lines express a unique fingerprint profile when using the (gtg)5 and pYNH24 probes. Cells were maintained under standard culture conditions in Dulbecco’s modified Eagle’s medium (DMEM) with high glucose, 10% fetal calf serum (FCS) and 1% antibiotics/ antimycotics (GIBCO, Eggenstein, Germany). Antibodies, proteins and peptides Flow cytometry and immunocytochemistry were performed using monoclonal antibodies (MAbs) specific for the integrins ␣1 (clone TS2/7; Serotec, Oxford, UK), ␣2 (Gi9), ␣3 (PiB5), ␣4, (HP2/1), ␣5 (SAM-1), ␣6 (GoH3), ␣v (AMF7), ␤1 (K20), ␤2 (BL5), ␤3 (SZ.21), ␤4 (3E1) and ␣v␤5 (P1F6; all from Dianova, Grant sponsor: Deutsche Forschungsgemeinschaft; Grant number: Lu 441/2-1. *Correspondence to: Department of Internal Medicine I, University of Ulm, 89081 Ulm, Germany. Fax: (49) 731-502-4323. E-mail: [email protected] Received 29 September 1998 792 VOGELMANN ET AL. Hamburg, Germany), and directed against IgG1 (MOPC-21; Sigma, Deisenhofen, Germany). Collagen type I, collagen type IV and fibronectin were from Sigma, and laminin was obtained from Biomol (Hamburg, Germany). The specificity of cell adhesion was tested using the MAbs described above, except for ␤1 integrin (DE9; Biomol). The peptides RGD, GRGDS, GRADSPK and YIGSR were purchased from Bachem (Heidelberg, Germany). Purified rat IgG and purified mouse IgG for metastasis assays were from Sigma and Dianova, respectively. All other reagents were of analytical grade. Flow cytometry As described in detail by Herzberg et al. (1996), trypsinized cells were washed once and resuspended at a concentration of 1 ⫻ 107 cells/ml in calcium and magnesium-free phosphate-buffered saline (PBS) containing 5% FCS. After incubation for 30 min with saturating concentrations of antibodies at 4°C, cells were washed once in PBS/5% FCS. Labeling was carried out with fluorescein isothiocyanate (FITC)-conjugated F(ab8)2 fragments for 30 min at 4°C. After an additional washing step, cells were analyzed using a FACScan (Becton-Dickinson, Heidelberg, Germany) with LYSIS II software. Antibodies against IgG1 were used as a basal control. A live gate was set by staining with 2 µg/ml propidium iodide for 5 min. The threshold level for significance of integrin expression was arbitrarily set at channel 50. Adhesion assay Semiconfluent cells in culture were washed with PBS containing 1 mM EDTA and were detached by short treatment with PBS/0.2% EDTA/0.5% trypsin. After 2 washing steps in Tris-buffered saline containing 0.01% (w/v) soybean trypsin inhibitor (Boehringer Mannheim, Germany), cells were resuspended to a density of 50,000 cells/100 µl Tris-buffered saline containing 0.1% heatdenatured bovine serum albumin (BSA), 2 mM glucose and 2 mM MgCl2 (TBS). Extracellular matrix protein-covered cell culture plates were prepared by incubation of sterile, non-tissue culture 96-well flat-bottom plates (Greiner, Frickenhausen, Germany) with the indicated concentrations of matrix proteins overnight at room temperature. Unspecific adhesion was blocked by incubation with 1% BSA (low endotoxin; Sigma) in PBS for 60 min at room temperature. For adhesion assays, 100 µl of cell suspension was added to each well and adhesion was allowed to proceed at 37°C/5% CO2 for the indicated time periods. Non-adherent cells were removed by 2 careful washing steps in PBS. The fraction of adherent cells was determined by staining with sulforhodamine B (Skehan et al., 1990). Briefly, cells were fixed by adding 125 µl of ice-cold trichloroacetic acid (TCA) solution (10% in DMEM) for 60 min. After 5 washing steps with H2O, the cells were stained with 0.4% sulforhodamine B in 1% acetic acid for 30 min. Following 4 additional washing steps in 1% acetic acid, the absorbed dyes were redissolved in 10 mM Tris base, pH 10.5, and the extinction was measured at 577 nm. Extinction did correlate with the cell number in a linear fashion under the assay conditions used. To calculate the fraction of adhering cells, the extinctions obtained in the absence of cells were subtracted, and values were expressed as percentage of the extinctions obtained without washing steps. For inhibition experiments, detached cells were preincubated in TBS with or without the indicated concentrations of MAbs, peptides or laminin for 60 min at 37°C under constant rotation, and adhesion assays were then performed as described above. In preliminary experiments, increasing concentrations were tested to determine the concentrations necessary for maximum inhibition. Adhesion was expressed as percent of extinction measured after preincubation with TBS alone. Metastasis assay To determine their metastatic potential, cells were prepared as described above and were suspended in cell culture medium; 100 µl of the cell suspension containing 2 ⫻ 106 cells was injected into the tail vein of NMRI nu-nu mice. After 4 weeks, the animals were sacrificed and both lungs were excised. Lung tissue was fixed in 4% formaldehyde. The total number of colonies were blindly counted in serial sections after staining with hematoxylin-eosin (H&E). The animal care protocol and the experimental design were approved by a governmental animal care review committee. Statistics The mean values and standard errors were calculated out of at least 3 experiments, each performed at least in quadruplicate. For statistical analysis, the Wilcoxon non-parametric test was used, and p ⬍ 0.05 was considered significant. RESULTS Adhesion assays In vitro adhesion was examined on purified extracellular matrix proteins. The results of time-dependent adhesion on the major interstitial matrix components collagen type I and fibronectin and the basal membrane components collagen type IV and laminin are shown in Figure 1. Assay conditions had been validated before by using increasing amounts of extracellular matrix proteins and by varying the divalent cation concentration. Adhesion did increase with substrate concentration and reached a plateau above 3.125 µg/ml for collagen I and collagen IV, and above 12.5 µg/ml for laminin in both cell lines. Maximal adhesion on fibronectin was observed above 6.25 µg/ml for PaTu 8988s cells and above 12.5 µg/ml for PaTu 8988t cells. Adhesion increased with the concentration of Mg2⫹, reaching its maximum level at 4 mM Mg2⫹ on all 4 substrates. Changing the Ca2⫹ concentration did not influence maximal adhesion. Therefore, 12.5 µg/ml substrate protein was used for surface coating in subsequent experiments and a concentration of 2 mM Mg2⫹ was used in the adhesion assays to approximate physiological conditions. Maximal adhesion was similar for both cell lines on the structural matrix proteins collagen types I and IV. In contrast, binding to the adhesive proteins laminin and fibronectin differed significantly between cell lines. Maximal adhesion of the metastatic cell line PaTu 8988s on laminin was 2.8 times that of the non-metastatic cell line PaTu 8988t ( p ⫽ 0.004). In contrast, adhesion on fibronectin was 8.1 times ( p ⫽ 0.03) higher for PaTu 8988t cells than for PaTu 8988s cells (Fig. 1). Expression of integrin subunits Integrin subunit expression in the metastatic cell line PaTu 8988s and in the non-metastastic cell line PaTu 8988t was quantified by flow cytometry after immunocytochemical staining (Fig. 2). FIGURE 1 – Adhesion of PaTu 8988s (䊊) and PaTu 8988t (䊏) cells on purified extracellular matrix proteins. Shown are the fractions of adhering cells after various time periods on the interstitial matrix proteins fibronectin (c) and collagen type I (a), and on the basal membrane proteins laminin (d) and collagen type IV (b). Values are expressed as means ⫾ SEM of at least 3 independent experiments. INTEGRIN ␣6␤1 AND PANCREATIC CARCINOMA 793 FIGURE 2 – Integrin subunit expression in pancreatic carcinoma cell lines PaTu 8988s (white bars) and PaTu 8988t (black bars). Cells were labeled with integrin-specific antibodies and analyzed by flow cytometry. Values are expressed as means ⫾ SEM of 2 independent determinations. Expression of the ␤1 subunit, and low expression levels of the ␤4 and ␤5 subunits were demonstrated in both cell lines. The ␤1 subunit that is the most ubiquitous integrin ␤ subunit in epithelial cells and is able to associate with most ␣ subunits was present in similar amount in both cell lines. Expression of the ␤4 subunit that is known to bind to laminin in combination with the ␣6 subunit, as well as the ␤5 subunit that mediates binding to vitronectin if dimerized to ␣v (Heino, 1996), was more prominent in the metastatic cell line PaTu 8988s. Expression of ␤2 and ␤3 subunits was below the threshold level. Several ␣ subunits were expressed in both cell lines. In PaTu 8988s cells, expression of ␣1 (preferentially binding collagen type I and laminin), ␣2 (preferentially binding collagen type VI and laminin), ␣6 (binding to laminin) and ␣v (binding to fibronectin with low specificity) was more marked. In contrast, expression levels of ␣5, which is the main fibronectin-binding subunit, were elevated in PaTu 8988t cells. Expression of the ␣4 subunit was below the threshold level in both cell lines. Specificity of adhesion To test the specificity of adhesion for the substrates, cells were preincubated either with laminin or with fibronectin-derived peptides. As illustrated in Figure 3, binding to laminin was inhibited after preincubation with the soluble laminin molecule (3.125 µg/ml) in PaTu 8988s cells by 47.0 ⫾ 10.2% ( p ⫽ 0.02) and in PaTu 8988t by 39.5 ⫾ 5.0% ( p ⫽ 0.02). Under identical conditions, the YIGSR peptide (0.5 mM), which represents the laminin region binding to the 67 kDa laminin receptor (Graf et al., 1987), had no effect on the adhesion of both cell lines on laminin. Next, binding to fibronectin was inhibited using known integrin-binding sequences of the fibronectin molecule (Hardan et al., 1993). Inhibition by the RGD peptide (0.5 mM) was less effective than preincubation with the GRGDS peptide (0.5 mM), which reduced the adhesion of PaTu 8988t on fibronectin to 42.8 ⫾ 8.4% ( p ⫽ 0.02) of controls. In control experiments, the same peptides did not inhibit adhesion of either cell lines to laminin, and laminin at a concentration of 3.125 mg/ml, which had been shown to maximally inhibit binding to laminin, did not decrease adhesion to fibronectin. Participation of specific integrin chains in cell adhesion was examined by preincubation of cells with integrin-specific MAbs. As illustrated, adhesion to laminin in PaTu 8988s cells was inhibited by 40.2 ⫾ 6.6% ( p ⬍ 0.005) with antibodies against ␤1 (1.25 µg/ml) and by 39.0 ⫾ 11.5% when using anti-␣6 antibodies (1.25 µg/ml, p ⬍ 0.005), whereas anti-␤4 antibodies (60 µg/ml) did not have any effect (Fig. 3). Similar results were observed using PaTu 8988t cells, where anti-␤1 decreased binding to laminin by 48.9 ⫾ 9.3% ( p ⬍ 0.005) and anti-␣6 by 50.0 ⫾ 4.7% ( p ⫽ 0.02). As expected, antibodies against the fibronectin-specific ␣5 subunit FIGURE 3 – Inhibition of maximal adhesion of PaTu 8988s cells (white bars) or PaTu 8988t cells (black bars) on laminin (a) or fibronectin (b). Cells were preincubated with the indicated concentrations of proteins, peptides or antibodies for 60 min and were allowed to adhere for 60 min. Shown are the fractions of adhering cells in the presence of inhibitors relative to controls. Values are expressed as means ⫾ SEM of at least 3 independent experiments. (1.25 µg/ml) as well as purified polyclonal mouse IgG (10 mM) did not decrease attachment to laminin. The same anti-␣5 antibody decreased binding of PaTu 8988s and PaTu 8988t cells to fibronectin by 52.0 ⫾ 5.4% ( p ⬍ 0.005) and by 45.0 ⫾ 16.6% (not significant), respectively. Fibronectin binding was inhibited with antibodies directed against the ␤1 subunit (PaTu 8988s: 32.2 ⫾ 6.4%, p ⬍ 0.005; PaTu 8988t: 48.6 ⫾ 4.9%, p ⬍ 0.005), whereas antibodies against ␤4 and ␣6 as well as IgG did not change adhesion significantly. Of note, this occurred even though ␤4 integrin subunits were present in our cell lines on the plasma membrane (data not shown) and ␤4 may associate with ␣6 in pancreatic tumor cells (Weinel et al., 1995). Metastasis formation In vivo integrin function was examined using a mouse model of metastasis formation in lung tissue. Injection of PaTu 8988s cells into the tail vein of NMRI nu-nu mice led to growth of 7.6 ⫾ 2.7 colonies/animal (n ⫽ 20, range 0–44) metastases in both lungs. In contrast, and as reported originally (Elsässer et al., 1992), PaTu 8988t cells did not colonize lung tissue (n ⫽ 13). As shown above, PaTu 8988s cells when compared with PaTu 8988t cells adhered more strongly to laminin and expressed more laminin-specific integrin subunits; in addition, binding to laminin was decreased by antibodies against the integrin subunits ␣6 and ␤1. To test the potential role of these laminin-binding integrins for metastasis formation, 8988s cells were preincubated with integrin-directed antibodies before injection. As shown in Figure 4, preincubation with the antibody directed against the laminin-binding ␣6 integrin significantly decreased lung colony formation to 0.6 ⫾ 0.2 colonies/animal (n ⫽ 13, p ⫽ 0.02, compared to control), whereas antibodies directed against the ubiquitous ␤1 integrin subunit led to a reduction to 1.1 ⫾ 0.4 lung colonies/animal (n ⫽ 8, p ⫽ 0.2). In control experiments, immunoglobulin controls for the anti-␣6 (rat VOGELMANN ET AL. 794 FIGURE 4 – Metastasis formation in lung tissue. Cells were injected into the tail vein of NMRI nu-nu mice after preincubation for 60 min in buffer alone or buffer with antibodies against integrin ␣6, integrin ␤1 or immunoglobulin controls. The number of metastases was counted after 4 weeks by histological examination. Values are expressed as means ⫾ SEM. IgG, n ⫽ 9) and anti-␤1 (mouse IgG, n ⫽ 5) antibodies had no effect on metastasis formation. DISCUSSION In this study, we have demonstrated that PaTu 8988s, a metastasizing human pancreatic carcinoma cell line, preferentially adheres to laminin and expresses a different pattern of integrin subunits than its twin cell line PaTu 8988t, which was originally derived from the same tumor tissue but lacks metastatic potential (Elsässer et al., 1992). Binding to laminin is thought to represent an essential step in metastasis (Iwamoto et al., 1987) and might explain the differences in metastatic capacity observed between the 2 cell lines. We, therefore, first confirmed that adhesion to laminin was indeed specific and could be inhibited by laminin but not by fibronectin-derived peptides, and then examined whether integrin expression could explain the observed differences in adhesion. Significant differences were observed for several ␣ integrin subunits relevant to binding of cells to laminin or to the interstitial matrix component fibronectin. The ␣6 subunit, which binds with high preference to laminin, was over-expressed in the metastatic cell line 8988s, whereas expression of the ␣5 subunit, which confers binding to fibronectin, was elevated in the non-metastasizing cell line 8988t. As expected, expression of the ␤1 subunit was similar in both cell lines, corresponding to its ability to associate with most ␣ subunits with similar affinity (Heino, 1996). In other tumor tissues, e.g., prostate cancer (Rabinovitz et al., 1995) or malignant melanoma (Danen et al., 1993), expression of the ␣6 integrin correlates with the ability of tumor cells to metastasize. The function of this integrin chain appears to center around attachment to or invasion through the basal membrane (Heino, 1996), which is supported by in vitro experiments where overexpression of ␣6␤1 integrins in chemically transformed human osteosarcoma cells correlated with their ability to migrate through the reconstituted basal membrane Matrigel (Dedhar and Saulnier, 1990), and where invasion into Matrigel was inhibited by antibodies against the ␣6 subunit in fibrosarcoma cells (Ramos et al., 1991), prostate carcinoma (Rabinovitz et al., 1995) and pancreatic carcinoma cell lines (Weinel et al., 1995). In addition, inhibition of ␣6 integrin function may be one of the mechanisms by which transfection of breast carcinoma cells with a dominant negative integrin ␤4 chain inhibits their ability to adhere and migrate on laminin in vitro (Shaw et al., 1996). In our model, the relative overexpression of ␣6 integrins in the metastatic PaTu 8988s cell line is consistent with the hypothesis that integrin-mediated laminin binding may be important for metastasis of pancreatic tumor cells. Cells can bind to laminin through a variety of laminin receptors, which include the integrins ␣6␤1 and ␣6␤4, but also the 67 kDa laminin receptor. In our experiments, adhesion was inhibited by antibodies against the ␣6 and the ␤1 integrin chains, but not by antibodies directed against the ␤4 integrin subunit or by the peptide YIGSR, which represents the laminin-binding sequence of the 67 kDa laminin receptor and which competitively inhibits binding through this mechanism (Graf et al., 1987). Therefore, similar to 5 other pancreatic carcinoma cell lines (Weinel et al., 1995; Rosewicz et al., 1997), binding of PaTu 8988s cells to laminin was mediated by the integrin ␣6␤1, and not via ␣6␤4 integrin complexes or the 67 kDa laminin receptor. To confirm the actual role of the integrin ␣6␤1 for in vivo metastasis, we did inhibit their function by incubation of PaTu 8988s cells with specific antibodies and could significantly decrease lung colonization in a nude mouse metastasis model. A similar mechanism has been described for mouse melanoma cells, where integrin ␣6 antibodies inhibit allogeneic metastasis formation when injected before or simultaneously with the tumor cells, or when cells had been precoated with antibodies (Ruiz et al., 1993). However, adhesion of melanoma cells to laminin as well as metastasis formation can be inhibited by the YIGSR sequence, which binds to the 67 kDa laminin receptor (Iwamoto et al., 1987). The same peptide does not inhibit adhesion of PaTu 8988s cells or other pancreatic tumor cells to laminin (Weinel et al., 1995) to further strengthen the central position of the ␣6␤1 integrin for pancreatic cancer cell metastasis. We conclude that overexpression of ␣6␤1 integrins is one of the determinants of metastasis formation in pancreatic carcinoma. Therefore, the diagnostic evaluation of ␣6␤1 integrin expression might provide valuable prognostic information to establish riskadapted treatment strategies. In addition, inhibition of integrin function might open ways to novel therapeutic approaches. ACKNOWLEDGEMENTS We thank Mrs. F. Genze and Mrs. E. Wolf-Hieber for expert technical assistance. REFERENCES DANEN, E.H., VAN MUIJEN, G.N., VAN DE WIEL VAN KEMENADE, E., JANSEN, K.F., RUITER, D.J. and FIGDOR, C.G., Regulation of integrin-mediated adhesion to laminin and collagen in human melanocytes and in nonmetastatic and highly metastatic human melanoma cells. Int. J. Cancer, 54, 315–321 (1993). DEDHAR, S. and SAULNIER, R., Alterations in integrin receptor expression on chemically transformed human cells: specific enhancement of laminin and collagen receptor complexes. J. Cell Biol., 110, 481–489 (1990). ELSÄSSER, H.P., LEHR, U., AGRICOLA, B. and KERN, H.F., Establishment and characterisation of two cell lines with different grade of differentiation derived from one primary human pancreatic adenocarcinoma. Virchows Arch. B Cell Pathol. Incl. mol. Pathol., 61, 295–306 (1992). GIANCOTTI, F.G. and RUOSLAHTI, E., Elevated levels of the alpha 5 beta 1 fibronectin receptor suppress the transformed phenotype of Chinese hamster ovary cells. Cell, 60, 849–859 (1990). GRAF, J., OGLE, R.C., ROBEY, F.A., SASAKI, M., MARTIN, G.R., YAMADA, Y. and KLEINMAN, H.K., A pentapeptide from the laminin B1 chain mediates cell adhesion and binds the 67,000 laminin receptor. Biochemistry, 26, 6896–6900 (1987). HALL, P.A., COATES, P., LEMOINE, N.R. and HORTON, M.A., Characterization of integrin chains in normal and neoplastic human pancreas. J. Pathol., 165, 33–41 (1991). HARDAN, I., WEISS, L., HERSHKOVIZ, R., GREENSPOON, N., ALON, R., CAHALON, L., REICH, S., SLAVIN, S. and LIDER, O., Inhibition of metastatic INTEGRIN ␣6␤1 AND PANCREATIC CARCINOMA cell colonization in murine lungs and tumor-induced morbidity by nonpeptidic Arg-Gly-Asp mimetics. Int. J. Cancer, 55, 1023–1028 (1993). HEINO, J., Biology of tumor cell invasion: interplay of cell adhesion and matrix degradation. Int. J. Cancer, 65, 717–722 (1996). HERZBERG, F., SCHONING, M., SCHIRNER, M., TOPP, M., THIEL, E. and KREUSER, E.D., IL-4 and TNF-alpha induce changes in integrin expression and adhesive properties and decrease the lung-colonizing potential of HT-29 colon carcinoma cells. Clin. exp. Metastasis, 14, 165–175 (1996). IWAMOTO, Y., ROBEY, F.A., GRAF, J., SASAKI, M., KLEINMAN, H.K., YAMADA, Y. and MARTIN, G.R., YIGSR, a synthetic laminin pentapeptide, inhibits experimental metastasis formation [published erratum appears in Science 239(4837), 245 (1988)]. Science, 238, 1132–1134 (1987). LÖHR, M., TRAUTMANN, B., GOTTLER, M., PETERS, S., ZAUNER, I., MAIER, A., KLÖPPEL, G., LIEBE, S. and KREUSER, E.D., Expression and function of receptors for extracellular matrix proteins in human ductal adenocarcinomas of the pancreas. Pancreas, 12, 248–259 (1996). RABINOVITZ, I., NAGLE, R.B. and CRESS, A.E., Integrin alpha 6 expression in human prostate carcinoma cells is associated with a migratory and invasive phenotype in vitro and in vivo. Clin. exp. Metastasis, 13, 481–491 (1995). RAMOS, D.M., CHENG, Y.F. and KRAMER, R.H., Role of laminin-binding integrin in the invasion of basement membrane matrices by fibrosarcoma cells. Invasion Metastasis, 11, 125–138 (1991). ROSENDAHL, A., NEUMANN, K., CHALOUPKA, B., ROTHMUND, M. and WEINEL, R.J., Expression and distribution of VLA receptors in the pancreas: an immunohistochemical study. Pancreas, 8, 711–718 (1993). 795 ROSEWICZ, S., WOLLBERGS, K., VON LAMPE, B., MATTHES, H., KAISER, A. and RIECKEN, E.O., Retinoids inhibit adhesion to laminin in human pancreatic carcinoma cells via the alpha 6 beta 1-integrin receptor. Gastroenterology, 112, 532–542 (1997). RUIZ, P., DUNON, D., SONNENBERG, A. and IMHOF, B.A., Suppression of mouse melanoma metastasis by EA-1, a monoclonal antibody specific for alpha 6 integrins. Cell Adhesion Comm., 1, 67–81 (1993). SCHREINER, C., FISHER, M., HUSSEIN, S. and JULIANO, R.L., Increased tumorigenicity of fibronectin receptor deficient Chinese hamster ovary cell variants. Cancer Res., 51, 1738–1740 (1991). SHAW, L.M., CHAO, C., WEWER, U.M. and MERCURIO, A.M., Function of the integrin alpha 6 beta 1 in metastatic breast carcinoma cells assessed by expression of a dominant-negative receptor. Cancer Res., 56, 959–963 (1996). SKEHAN, P., STORENG, R., SCUDIERO, D., MONKS, A., MCMAHON, J., VISTICA, D., WARREN, J.T., BOKESCH, H., KENNEY, S. and BOYD, M.R., New colorimetric cytotoxicity assay for anticancer-drug screening. J. nat. Cancer Inst., 82, 1107–1112 (1990). VARNER, J.A., EMERSON, D.A. and JULIANO, R.L., Integrin alpha 5 beta 1 expression negatively regulates cell growth: reversal by attachment to fibronectin. Mol. Biol. Cell, 6, 725–740 (1995). WEINEL, R.J., ROSENDAHL, A., PINSCHMIDT, E., KISKER, O., SIMON, B. and SANTOSO, S., The alpha 6-integrin receptor in pancreatic carcinoma. Gastroenterology, 108, 523–532 (1995).