The effect of the characteristic desmoplastic reaction of pancreatic cancer on tumor progression is largely unknown. We investigated whether pancreatic stellate cells, which are responsible for the desmoplastic reaction, support tumor progression. Immunohistology revealed that matrix metalloproteinase-2 (MMP-2), which is suggested to promote pancreatic cancer progression, is present in stellate cells adjacent to cancer cells. In vitro, stellate cells exhibited a much higher basal expression of MMP-2 compared with cancer cells. Panc1-, MiaPaCa2- and SW850-conditioned media stimulated MMP-2 release of stellate cells as detected by zymography. Cancer cells expressed and released basigin [BSG, extracellular matrix metalloproteinase inducer (EMMPRIN), CD147], a glycoprotein that is known to stimulate MMP-2 in mesenchymal cells, as detected by immunostaining, western blot and reverse transcription-polymerase chain reaction. Tumor cell-conditioned medium and BSG purified by affinity chromatography from supernatants of cancer cells, but not supernatants depleted from BSG, stimulated expression of MMP-1 and MMP-2 of stellate cells as demonstrated by western blot and zymography. Moreover, the interaction of stellate cells and cancer cells promoted the invasiveness of Panc-1 cells in the chorioallantoic membrane assay and increased the weight of tumors induced by all carcinoma cell lines in nude mice by 2.1-3.7-fold. Our findings support the assumption that the interaction of stellate cells and cancer cells promotes progression of pancreatic cancer.

Adenocarcinomas of the pancreas are a highly aggressive malignancy with a 5-year survival rate of only 1-4%. They remain one of the leading causes of cancer-related death (Jemal et al., 2006), and efficient therapies are currently unavailable (Landis et al., 1999; Smeenk et al., 2005). Further insights into the mechanisms causing the highly aggressive nature of this disease might reveal new prospects for therapy and diagnosis of pancreatic adenocarcinomas.

Compared with other epithelial neoplasms, pancreatic adenocarcinomas have an extremely rich and dense fibrotic stroma surrounding the malignant cells (Bachem et al., 2005; Seymour et al., 1994). We recently demonstrated that the desmoplastic reaction in pancreatic cancer is the result of the cross-talk between carcinoma cells and the pancreatic stellate cells (PSC) (Bachem et al., 2005). PSC are also involved in the development of pancreatic fibrosis following pancreatitis (Haber et al., 1999) and share a lot of similarities with hepatic stellate cells (Buchholz et al., 2005), which synthesize `fibrotic' matrix in diseased human liver (Bachem et al., 1998; Friedman, 1993). Both are activated from a quiescent state upon stimulation with TGF-β1 (Bachem et al., 1998). This process leads to the acquisition of a myofibroblast-like phenotype by the PSC, which has the following changes compared with its quiescent counterpart: (1) loss of retinyl droplets, (2) change in cellular size and morphology, (3) expression of iso-α smooth muscle actin, and (4) the ability to proliferate and (5) synthesize significant amounts of extracellular matrix proteins (Apte et al., 1998; Bachem et al., 1998).

Cancer cell-derived matrix metalloproteinase-2 (MMP-2) was recently shown to be associated with the development of the desmoplastic reaction and seems to promote tumor cell invasion in human pancreatic cancer (Ellenrieder et al., 2000). However, in most epithelial cancers the majority of MMP-2 is produced by stromal cells (Egeblad and Werb, 2002) and PSC also express the gelatinases MMP-2 and MMP-9 (Phillips et al., 2003). Mesenchymal cells are stimulated to produce MMP-1, MMP-2 and MMP-3 by basigin (BSG), a transmembrane glycoprotein enriched on several tumor cell types, which is identical to extracellular matrix metalloproteinase inducer (EMMPRIN) (CD147) (Biswas et al., 1995). Although BSG is a transmembrane protein, cancer cells do not stimulate MMPs through direct cell-cell contact with mesenchymal cells. Instead, the malignant cells stimulate MMPs in a paracrine fashion after BSG is shed from their cell surface via microvesicles (Sidhu et al., 2004) or by matrix-metalloproteinases (Haug et al., 2004).

It has been suggested that BSG plays an important role in tumor-mesenchymal cell interaction and in promoting tumor progression (Biswas et al., 1995; Caudroy et al., 2002; Kanekura et al., 2002; Sameshima et al., 2000; Taylor et al., 2002; Zucker et al., 2001). Experimental overexpression of BSG in MDA-MB-436 human mammary tumor cells, which express lower levels of BSG than most malignant cell lines, promotes invasive capacity (Caudroy et al., 2002) and anchorage-independent growth in vitro (Marieb et al., 2004). In vivo, one of the most marked responses to overexpression of BSG was the formation of fast-growing tumors in nude mice (Zucker et al., 2001).

Tumor-host communication is crucial for the progression of many human tumors and PSC are abundant in the desmoplasia, producing matrix-associated growth factors, as well as MMP-2, both of which have been associated with accelerated pancreatic cancer progression. The effect of PSC on pancreatic cancer progression, however, remains elusive.

To elucidate the role of cancer cell-PSC interaction with respect to progression of pancreatic cancer, we first investigated the source and the regulation of MMP-2 in the desmoplastic reaction. Moreover, we analyzed the influence of PSC on cancer cell invasion in the chorioallantoic membrane (CAM) assay and on tumorigenity in a xenograft model in nude mice.

We demonstrate that PSC are an important source of MMP-2 in pancreatic cancer and that the interaction of PSC and tumor cells stimulates MMP-2 in PSC and promotes invasiveness (CAM assay) as well as tumorigenity (nude mice xenograft model) of cancer cells. These findings suggest that the desmoplasia of pancreatic cancer promotes tumor progression.

PSC are the major source of MMP-2 in pancreatic adenocarcinomas

MMPs play a pivotal role in tumor progression. Since MMP-2 is associated with the development of the desmoplastic reaction and enhanced tumor progression in pancreatic cancer (Ellenrieder et al., 2000), we elucidated the role of PSC as a source of MMPs in the desmoplasia of human adenocarcinomas of the pancreas.

Expression of MMPs by PSC was compared with MMP expression by tumor cells. Gelatinase activity was hardly detectable by zymography in supernatants of tumor cells but was very strong in PSC supernatants (Fig. 1A). A prominent band of approximately 65 kDa corresponding to proMMP-2 was detected in PSC supernatants and Panc1 supernatant (Fig. 1A). If supernatant of SW850 and MiaPaCa2 was concentrated 10-fold, minimal amounts of MMP-2 could be detected here as well (data not shown).

Since pancreatic cancer cells express much lower amounts of MMP-2 compared with PSC, we speculated that tumor cells might stimulate PSC to serve as a source for MMP-2. To test this hypothesis, we treated cultivated PSC with tumor cell-conditioned medium. Zymography revealed that culturing PSC in the presence of conditioned media from each tested carcinoma cell line resulted in an increase in the concentration of MMP-2 in the PSC supernatant (Fig. 1B). Densitometric scan of these zymographies shows that cancer cell-conditioned medium increased MMP-2 concentration by up to 1.32-fold (MiaPaCa2), 1.23-fold (Panc1) or 1.38-fold (MiaPaCa2), respectively (Fig. 1C). We went on to identify the factors involved. Several lines of evidence suggest that BSG is a paracrine factor that stimulates mesenchymal cells to increase MMP expression. We thus investigated the role of BSG produced by cancer cell lines in paracrine stimulation of MMP synthesis by PSC.

Fig. 1.

MMP expression in pancreatic cancer cell lines and mesenchymal cells. (A) Gelatin zymography (Zymo) of tumor cell and PSC supernatant reveals that MMP-2 is predominantly expressed in PSC. Supernatant was loaded adapted to the DNA level of the wells corresponding to the samples. (B) Gelatin zymography: the indicated volume of supernatant conditioned by MiaPaCa2, Panc1 and SW850 (TC SN) was incubated with PSC for 24 hours. Serum-free culture media was added to yield a total volume of 2000 μl. Supernatant of cancer cells stimulated the expression of MMP-2 in PSC. Supernatant was loaded adjusted to the DNA level of the wells corresponding to the samples. Results are representative of at least three independent experiments. (C) Densitometry of the zymography: each condition was determined in quandruplicate and values are expressed as means. Boxes represent the s.e.m. and error bars the s.d. The asterisk denotes a statistically significant difference (P<0.01) for MMP-2 in the supernatant of the group of PSC stimulated with 1000 μl cancer cell supernatant compared with control.

Fig. 1.

MMP expression in pancreatic cancer cell lines and mesenchymal cells. (A) Gelatin zymography (Zymo) of tumor cell and PSC supernatant reveals that MMP-2 is predominantly expressed in PSC. Supernatant was loaded adapted to the DNA level of the wells corresponding to the samples. (B) Gelatin zymography: the indicated volume of supernatant conditioned by MiaPaCa2, Panc1 and SW850 (TC SN) was incubated with PSC for 24 hours. Serum-free culture media was added to yield a total volume of 2000 μl. Supernatant of cancer cells stimulated the expression of MMP-2 in PSC. Supernatant was loaded adjusted to the DNA level of the wells corresponding to the samples. Results are representative of at least three independent experiments. (C) Densitometry of the zymography: each condition was determined in quandruplicate and values are expressed as means. Boxes represent the s.e.m. and error bars the s.d. The asterisk denotes a statistically significant difference (P<0.01) for MMP-2 in the supernatant of the group of PSC stimulated with 1000 μl cancer cell supernatant compared with control.

Fig. 2.

BSG expression in cultured tumor cell lines MiaPaCa2, SW850 and Panc1. (A) Immunofluorescence staining of BSG: the human pancreatic adenocarcinoma cell lines were immunostained for BSG. Immunoreactivity was detected by fluorescein isothiocyanate (FITC) (green). Control was performed by omission of the primary antibody. Initial magnification was 200×. (B) Western-blot analysis of RIPA lysates and supernatant of MiaPaCa2, SW850, Panc1 cells as well as PSC. A band of 40-50 kDa according to the molecular mass of HG BSG was detected in supernatants as well as lysates. The LG form of BSG appeared as a band of 30-35 kDa in tumor cell lysates. Highly glycosylated proteins do not focus if separated by SDS-PAGE and glycosylation differs among the different cell types. Supernatants were concentrated 10-fold before western blot analysis. Lysates and supernatants were loaded adjusted to their protein concentration or adjusted to the surface area of the confluent well. Control was performed by omission of the primary antibody. (C) BSG immunofluorescence staining (FITC, green) of cancer cells co-cultured with PSC. Compared with cancer cells, the intensity of the BSG fluorescence was much lower on PSC as the result of a lower density of BSG in the cell membrane of PSC. The large PSC with their irregular polygonal fibroblast-like shape can be distinguished from the small round cancer cells in the phase-contrast image. Nuclei were counterstained with propidium iodide (red). (D) Agarose gel electrophoresis demonstrating the qRT-PCR products of BSG with a predicted length of 120 bp that were reverse transcribed and amplified from RNA of cultured MiaPaCa2, SW850 and Panc1 carcinoma cell lines as well as PSC (left panel). Controls comprised substitution of water for either cDNA (no template) or reverse transcriptase (no RT). In addition, the results of qRT-PCR were expressed normalized using XS13 as a housekeeping gene (right panel). The diagram is representative for three independent experiments.

Fig. 2.

BSG expression in cultured tumor cell lines MiaPaCa2, SW850 and Panc1. (A) Immunofluorescence staining of BSG: the human pancreatic adenocarcinoma cell lines were immunostained for BSG. Immunoreactivity was detected by fluorescein isothiocyanate (FITC) (green). Control was performed by omission of the primary antibody. Initial magnification was 200×. (B) Western-blot analysis of RIPA lysates and supernatant of MiaPaCa2, SW850, Panc1 cells as well as PSC. A band of 40-50 kDa according to the molecular mass of HG BSG was detected in supernatants as well as lysates. The LG form of BSG appeared as a band of 30-35 kDa in tumor cell lysates. Highly glycosylated proteins do not focus if separated by SDS-PAGE and glycosylation differs among the different cell types. Supernatants were concentrated 10-fold before western blot analysis. Lysates and supernatants were loaded adjusted to their protein concentration or adjusted to the surface area of the confluent well. Control was performed by omission of the primary antibody. (C) BSG immunofluorescence staining (FITC, green) of cancer cells co-cultured with PSC. Compared with cancer cells, the intensity of the BSG fluorescence was much lower on PSC as the result of a lower density of BSG in the cell membrane of PSC. The large PSC with their irregular polygonal fibroblast-like shape can be distinguished from the small round cancer cells in the phase-contrast image. Nuclei were counterstained with propidium iodide (red). (D) Agarose gel electrophoresis demonstrating the qRT-PCR products of BSG with a predicted length of 120 bp that were reverse transcribed and amplified from RNA of cultured MiaPaCa2, SW850 and Panc1 carcinoma cell lines as well as PSC (left panel). Controls comprised substitution of water for either cDNA (no template) or reverse transcriptase (no RT). In addition, the results of qRT-PCR were expressed normalized using XS13 as a housekeeping gene (right panel). The diagram is representative for three independent experiments.

BSG released by tumor cells stimulates MMP synthesis in PSC

We first examined BSG expression by pancreatic cancer cell lines and PSC (Fig. 2). Immunocytology demonstrated BSG immunoreactivity in cultured MiaPaCa2, SW850 and Panc1 cells (Fig. 2A). A band of approximately 40-50 kDa was detected by western blotting in lysates (Fig. 2B, top panel) and supernatants (Fig. 2B, bottom panel) of MiaPaCa2, Panc1, SW850 and PSC. The low glycosylated precursor (LG) of highly glycosylated (HG) BSG appeared as a band with a molecular mass of approximately 30-35 kDa (Fig. 2B). The HG form shows the typical appearance of an HG protein, which tend not to focus when separated by SDS-PAGE. The observed band pattern is in accordance with previously published data (Tang et al., 2004). When adjusted to the total protein concentration, PSC exhibited an expression of cell-associated BSG that was comparable to cancer cells, yet the concentration of soluble BSG shed by PSC into the supernatant was considerably lower (Fig. 2B). When samples were loaded adjusted to the surface area of the cells, the expression level of cell-associated BSG in PSC was much lower than in cancer cells, indicating a lower density of BSG in the membrane of PSC (Fig. 2B). The lower concentration of BSG in the membrane of PSC was also demonstrated by immunofluorescence microscopy of stellate cells co-cultured with cancer cells (Fig. 2C). The PSC can be distinguished from the small round cancer cells in the phase-contrast image by their larger size and their irregular polygonal shape (Fig. 2C). BSG expression by tumor cells was also demonstrated to be higher than that of PSC on the RNA level by quantitative reverse transcription-polymerase chain reaction (qRT-PCR). The PCR product with a predicted length of 120 bases was amplified with BSG-specific primers on a light-cycler platform and XS13 was used as a housekeeping gene (Fig. 2D, right panel). The negative controls comprised a no-template control and omission of the reverse transcriptase (Fig. 2D, left panel).

To evaluate the relative importance of BSG in paracrine stimulation of MMP synthesis by PSC compared with the other components of the cancer cell-conditioned medium, we compared the ability of purified BSG, complete tumor cell supernatant and BSG-immunodepleted tumor cell supernatant to stimulate MMP expression when added to PSC. Purified BSG exhibited the same effect on MMP-2 concentration in PSC supernatant as cancer cell-conditioned medium that had the same final BSG concentration (Fig. 3A). In addition, the effect of cancer cell-conditioned medium was completely abolished by immunodepletion of BSG from the cancer cell-conditioned medium (Fig. 3A). The gelatin zymography data were verified by western blot analysis (Fig. 3B). In addition, western-blot analysis revealed a strong increase in MMP-1 concentration in PSC supernatant, occurring in parallel to the effect of BSG on MMP-2 (Fig. 3B).

The next question we addressed was whether the in vitro observation that tumor cells secrete BSG to stimulate MMP-2 expression by PSC is also likely to occur in pancreatic cancer in vivo.

In vivo BSG is expressed in human pancreatic cancer mainly by tumor cells, whereas MMP-2 is enriched in peritumoral stroma

We examined the localization of BSG and MMPs in relation to cancer cells and PSC respectively by immunofluorescence microscopy of serial sections of 10 human pancreatic adenocarcinomas (Fig. 4). The cancer cells, identified by pancytokeratin staining, are surrounded by an extensive peritumoral fibrosis, which was identified by its strong reactivity with antibodies to the matrix proteins fibronectin (Fig. 4A) and collagen type I (data not shown). To demonstrate that PSC are located in the peritumoral stroma, immunostaining with the PSC marker iso-α smooth muscle actin was performed. Iso-α smooth muscle actin (Fig. 4A) was colocalized with the proteins of the extracellular matrix of the peritumoral fibrosis. BSG immunostaining revealed an intense BSG expression predominantly co-localized with carcinoma cells (Fig. 4A-C) but not with normal monolayer epithelial cells of pancreatic duct structures (Fig. 4B). In accordance with the in vitro results, immunoreactivity to MMP-2 was mainly detected in the stroma of pancreatic cancer, and particularly at the tumor-stroma interface (Fig. 4C).

PSC promote growth and invasion of pancreatic tumors in vivo

Since the effects of the interaction of tumor cells with PSC suggested an influence of PSC on cancer progression, we employed the CAM assay and a xenograft model in nude mice to test whether PSC influence tumor progression in vivo. These models were intended to reflect the effect of the sum of the interactions between PSC and cancer cells with respect to tumor progression, not just that of MMP-2 stimulation. In the CAM assay we inoculated Panc1 cells either alone or in combination with PSC on chick CAMs (Fig. 5A). Microscopic examination revealed no tumor formation after 4 days of incubation if Panc1 cells were cultured alone (Fig. 5A). In the presence of PSC, Panc1 cells invaded the CAM and formed submembraneous tumors in all cases (n=6) (Fig. 5A). MiaPaCa2 and SW850 failed to form tumors on the CAM whether PSC were present or not and therefore the effect of PSC on these cells could not be evaluated (data not shown). Therefore, we used a xenograft to test the effect of PSC on cancer progression.

Fig. 3.

Effect of tumor cell supernatant-derived BSG on MMP expression in PSC. (A,B) As demonstrated by zymography (A) and western blot (B), BSG isolated from MiaPaCa2, Panc1 and SW850 stimulates MMP-2 and MMP-1-expression in PSC to the same extent as complete supernatant. Final concentration of purified BSG and BSG in supernatant were adjusted to the same concentration. Supernatant depleted of BSG by immunoprecipitation (ID supernatant) did not increase MMP levels compared with control.

Fig. 3.

Effect of tumor cell supernatant-derived BSG on MMP expression in PSC. (A,B) As demonstrated by zymography (A) and western blot (B), BSG isolated from MiaPaCa2, Panc1 and SW850 stimulates MMP-2 and MMP-1-expression in PSC to the same extent as complete supernatant. Final concentration of purified BSG and BSG in supernatant were adjusted to the same concentration. Supernatant depleted of BSG by immunoprecipitation (ID supernatant) did not increase MMP levels compared with control.

To assess the influence of PSC on cancer cell tumorigenicity, xenograft tumors were induced by subcutaneous injection of Panc1, MiaPaCa2 or SW850 cells in the presence of PSC on the left-side of six nude mice. As a control, tumor cells were injected on the right-side in the absence of PSC. After 11 days of incubation, palpable tumors formed in all mice at the site of injection. PSC alone were not tumorigenic in nude mice (data not shown).

Fig. 4.

Immunohistologic examination of BSG expression in human ductal adenocarcinomas of the pancreas. (A) Immunofluorescence micrographs demonstrating the distribution of BSG, cytokeratin, fibronectin and iso-α smooth muscle actin (FITC, green) expression in serial sections of a human carcinoma. The PSC marker iso-α smooth muscle actin is colocalized with the excessive deposition of fibronectin, a major constituent of the extracellular matrix of the tumor desmoplasia. BSG was only detected in cancer cells, which are identified by pancytokeratin staining, whereas the stromal cells of the desmoplasic reaction exhibit no immunoreactivity to BSG. Nuclei were counterstained with propidium iodide. Initial magnification was 100×. (B) Immunofluorescence micrograph of BSG (FITC, green) expression in an adenocarcinoma of the pancreas. The duct-like structures (#) with multilayered malignant cells exhibit a strong BSG immunoreactivity, whereas an adjacent normal duct structure (*) with regular epithelial cells shows no BSG expression. Nuclei were counterstained with propidium iodide. Initial magnification was 200×. (#, duct-like structure of an adenocarcinoma; *, regular pancreatic duct.) (C) Immunostaining of BSG and MMP-2 in sections of human pancreatic cancer using APAAP detection with Fast Red as a chromogen. BSG is expressed by malignant epithelial cells. MMP-2 could be detected in peritumoral stroma at the tumor-stroma interface adjacent to malignant epithelial cells with high BSG immunoractivity. Nuclei were counterstained with Mayer's hemalaun. Initial magnification was 200×.

Fig. 4.

Immunohistologic examination of BSG expression in human ductal adenocarcinomas of the pancreas. (A) Immunofluorescence micrographs demonstrating the distribution of BSG, cytokeratin, fibronectin and iso-α smooth muscle actin (FITC, green) expression in serial sections of a human carcinoma. The PSC marker iso-α smooth muscle actin is colocalized with the excessive deposition of fibronectin, a major constituent of the extracellular matrix of the tumor desmoplasia. BSG was only detected in cancer cells, which are identified by pancytokeratin staining, whereas the stromal cells of the desmoplasic reaction exhibit no immunoreactivity to BSG. Nuclei were counterstained with propidium iodide. Initial magnification was 100×. (B) Immunofluorescence micrograph of BSG (FITC, green) expression in an adenocarcinoma of the pancreas. The duct-like structures (#) with multilayered malignant cells exhibit a strong BSG immunoreactivity, whereas an adjacent normal duct structure (*) with regular epithelial cells shows no BSG expression. Nuclei were counterstained with propidium iodide. Initial magnification was 200×. (#, duct-like structure of an adenocarcinoma; *, regular pancreatic duct.) (C) Immunostaining of BSG and MMP-2 in sections of human pancreatic cancer using APAAP detection with Fast Red as a chromogen. BSG is expressed by malignant epithelial cells. MMP-2 could be detected in peritumoral stroma at the tumor-stroma interface adjacent to malignant epithelial cells with high BSG immunoractivity. Nuclei were counterstained with Mayer's hemalaun. Initial magnification was 200×.

Tumors grew progressively up to a final weight of 40±27 mg (mean ± s.d.) if induced with Panc1, up to 33±17 mg with MiaPaCa2, and up to 60±12 mg with SW850, respectively (Fig. 5B). PSC significantly increased the weight of Panc1 tumors by 3.3-fold to 131±66 mg, the weight of MiaPaCa2 tumors by 3.7-fold to 123±35 mg, and the weight of SW850 tumors by 2.1-fold to 126±46 mg (Fig. 5B).

Fibroblasts were long presumed to be passive structural elements, but now their importance in malignant tumors is increasingly recognized, and many reports indicate that continuous interactions between carcinoma and stromal cells are a prerequisite for tumor development and progression (Liotta and Kohn, 2001). The phenomenon of tumor-associated desmoplasia is most frequently described in squamous cell carcinomas and bilo-pancreatic carcinomas (Kunz-Schughart et al., 2002). We recently demonstrated that the characteristic dense fibrous peritumoral stroma of pancreatic adenocarcinoma forms as a result of the interaction between tumor cells and PSC (Bachem et al., 2005). However, it is not known whether the desmoplastic reaction promotes, inhibits or has no effect on tumor progression.

It has recently been demonstrated that MMP-2 is associated with development of the desmoplasia of pancreatic carcinoma (Ellenrieder et al., 2000). MMPs have long been associated with metastasis, poor prognosis and the malignant phenotype of tumors (Egeblad and Werb, 2002) and they are now reported to be involved in the invasion of pancreatic cancer as well (Ellenrieder et al., 2000). In recent years it has become clear that transformed cells secrete MMPs, and yet a major fraction of MMPs is produced by fibroblasts in the tumor stroma (Egeblad and Werb, 2002; Noel et al., 1994). Our current results show that PSC constitutively express high levels of MMP-2 compared with the pancreatic tumor cell lines MiaPaCa2, SW850 and Panc1. Immunohistology confirmed that MMP-2 immunoreactivity is found on mesenchymal cells adjacent to cancer cells colocalized with iso-α-smooth muscle actin. Our results demonstrate for the first time that the stromal PSC also significantly contribute to MMP-2 secretion in pancreatic cancer in vivo and in vitro. Moreover, our results suggest that secretion of MMP-2 by PSC can exceed that of cancer cells by far. Although the exact sources for MMPs in pancreatic cancer need to be identified in further detailed studies, these result indicate that PSC are a major source and further emphasize the role of the host reaction in pancreatic cancer progression.

Fig. 5.

Pancreatic stellate cells promote invasion and growth of pancreatic tumors in the CAM model and a xenograft model in nude mice. (A) If CAMs of chick embryos were inoculated with Panc1 cells alone, after 4 days no invasion of the membrane was observed by microscopic examination of hematoxylin-eosin-stained paraffin sections. In combination with human PSC, Panc1 cells formed significant tumors in all cases and markedly invaded the CAM. Initial magnification was 200×. (B) After subcutaneous injection of tumor cells into nude mice either alone on the right-side or in combination with PSC on the left-side, palpable tumors formed at the site of injection in all cases after 11 days. Tumor weight of explanted pancreatic tumors induced by subcutaneous injection of Panc1, MiaPaCa2 or SW850 either in the presence (+PSC) or absence (-PSC) of PSC in nude mice was determined at the end of the experiment. The squares represent the mean, the circles the single measured values, the boxes the standard error and the error bars the standard deviation observed for six animals. All groups with PSC were significantly larger than the respective group without PSC (*P<0.05).

Fig. 5.

Pancreatic stellate cells promote invasion and growth of pancreatic tumors in the CAM model and a xenograft model in nude mice. (A) If CAMs of chick embryos were inoculated with Panc1 cells alone, after 4 days no invasion of the membrane was observed by microscopic examination of hematoxylin-eosin-stained paraffin sections. In combination with human PSC, Panc1 cells formed significant tumors in all cases and markedly invaded the CAM. Initial magnification was 200×. (B) After subcutaneous injection of tumor cells into nude mice either alone on the right-side or in combination with PSC on the left-side, palpable tumors formed at the site of injection in all cases after 11 days. Tumor weight of explanted pancreatic tumors induced by subcutaneous injection of Panc1, MiaPaCa2 or SW850 either in the presence (+PSC) or absence (-PSC) of PSC in nude mice was determined at the end of the experiment. The squares represent the mean, the circles the single measured values, the boxes the standard error and the error bars the standard deviation observed for six animals. All groups with PSC were significantly larger than the respective group without PSC (*P<0.05).

We observed that the MMP-2 synthesis of PSC was even further increased in the presence of tumor cell-conditioned medium, suggesting an interaction of tumor cells and PSC by soluble factors. Here we demonstrate by RT-PCR, western blot and immunofluorescence staining that BSG, a type-2 membrane glycoprotein that induces MMPs in stromal cells (Biswas et al., 1995), is expressed and released by MiaPaCa2, SW850 and Panc1. We and others have demonstrated that the soluble form of BSG is a result of shedding of the membrane-associated form (Haug et al., 2004; Sidhu et al., 2004). BSG expression by pancreatic adenocarcinomas was also shown by immunohistology of surgically resected human tumors in vivo. BSG has been demonstrated to be enriched in many human malignancies including lymphomas, gliomas, breast cancer, ovarian cancer and certain forms of skin cancer (Davidson et al., 2003; Sameshima et al., 2000; Taylor et al., 2002; Thorns et al., 2002) and the expression of BSG correleates with malignant potential and poor survival (Dalberg et al., 2000; Davidson et al., 2003; Marionnet et al., 2003).

The molecular mechanisms by which BSG promotes the malignant potential of tumors are poorly understood but its property to stimulate MMPs in mesenchymal cells has gained considerable attention (Kanekura et al., 2002; Li et al., 2001; Taylor et al., 2002). In our study, BSG purified from tumor cell supernatant stimulated MMP-1 and MMP-2 synthesis in PSC to the same extent as tumor cell supernatant. These findings demonstrate that BSG is a soluble factor released by cancer cells that plays a significant role in the interaction of cancer cells and PSC resulting in the stimulation of MMP-2. BSG seems to be a major factor for MMP stimulation compared with other cancer cell-derived soluble factors. This assumption is supported by our observation that the effect of conditioned medium was almost completely abolished after BSG immunodepletion.

Although we are only beginning to understand how mesenchymal cells and MMPs support tumor growth (Liotta and Kohn, 2001; Zucker et al., 2000), it has been reported that fibroblasts from MMP knockout mice have lost their ability to support tumor growth (Itoh et al., 1999). This emphasizes the pivotal role of mesenchymal cell-derived MMP secretion for tumor progression.

One of the most remarkable properties of PSC was their ability to support tumor progression, as demonstrated by the CAM- and xenograft assay. PSC substantially promote the growth and invasion of Panc1 cells through the CAM of chick embryos and support tumorigenity of tumor xenografts in nude mice. These results demonstrate that the interaction of cancer cells with PSC propagates the invasion of pancreatic cancer in vivo. In breast cancer and prostatic cancer, the role of the stromal microevironment, which consists primarily of fibroblasts, is also being increasingly appreciated (De Wever and Mareel, 2002; Liotta and Kohn, 2001).

Here we demonstrate that PSC represent a significant source of MMP-2 in pancreatic cancer, which is known to promote tumor progression. Moreover, cancer cells release BSG to stimulate MMP-2 in PSC, indicating that cross-talk of cancer cells and mesenchymal cells might further increase MMP-2 production in the desmoplasia. Finally, we show by in vivo models that the interaction of human cancer cells with PSC promotes invasiveness and tumorigenity of malignant epithelial cells of the pancreas. In conclusion, our data suggest that the tumor-stroma interaction in the desmoplastic reaction of pancreatic cancer promotes tumor progression.

Cell culture

Human PSC were isolated by outgrowth, using surgically resected pancreas as described elsewhere (Bachem et al., 1998). Cells were maintained in Dulbecco's modified Eagle's medium (DMEM)/Ham's F12 (1:1 vol./vol.) containing 10% fetal calf serum (FCS) and used from passage 3-8. To study stimulated MMP synthesis, BSG or tumor cell-conditioned media were added to 3-5×104 PSC/cm2 grown in six-well plates in the absence of FCS. Cultures were stopped 24 hours later.

To obtain carcinoma cell-conditioned media, 10 ml DMEM was conditioned by incubation with confluent tumor cell lines (MiaPaCa2, Panc1 and SW850) in 75 cm2 culture flasks for 48 hours.

Purification of BSG

Panc1-, SW850- or MiaPaCa-conditioned media were dialyzed against 20 mM Tris, pH 8.0 at 4°C overnight and applied to an equilibrated anion-exchange column (BioCAD Workstation, Applied Biosystems, Foster City, CA). Bound proteins were eluted with a linear salt gradient. The BSG-containing fraction was added to an immunoaffinity matrix for 12 hours at 4°C, which was prepared by coupling a monoclonal anti-human BSG antibody (Serotec, Oxford, UK) to protein-G-agarose using ImmunoPure plus an Immunoglobulin G (IgG) Orientation Kit (Pierce, Rockford, IL). The immunoaffinity matrix was washed with 20 mM Tris, pH 8.0, several times before BSG was eluted with elution buffer (Pierce, Rockford, IL). The eluted protein was neutralized with 1 M Tris, pH 9.5, and analyzed for purity by SDS-PAGE and silver staining.

BSG immunoassay

To quantify BSG, Nunc-maxisorb 96-well plates (Nunc, Roskilde, Denmark) were coated with 50 μl/well polyclonal rabbit anti-mouse antibody (DAKO, Glostrup, Denmark) diluted 1:220 in coating buffer (50 mM NaHCO3, pH 9.1) at 4°C overnight. Subsequently, 100 μl/well monoclonal anti-CD147 antibody (R&D Systems, Minneapolis, MN) diluted 1:2000 in coating buffer was allowed to bind for 4 hours at room temperature. Unspecific binding sites were blocked by incubation with assay buffer [50 mM Tris, 150 mM NaCl, 0.5% (w/v) albumin, 0.5% (w/v) NaN3, pH 7.7] for 2 hours at room temperature before incubation with 100 μl/well of sample or standard at 4°C overnight. Following incubation of 100 μl/well biotinylated polyclonal anti-CD147 antibody (R&D Systems) diluted 1:1000 in assay buffer for 4 hours, streptavidin europium (PerkinElmer/Wallac, Turku, Finland) was incubated diluted 1:500 in assay buffer for 1 hour. Finally, time-resolved fluorescence was measured on a Victor multilabelcounter (PerkinElmer/Wallac) after enhancement solution was added. Each incubation step was followed by a washing step with 150 mM NaCl, 50 mM Tris, 0.5% (v/v) Tween, pH 7.4. A dilution series of conditioned tumor cell supernatant was used as a standard and results were expressed as arbitrary units. All samples were determined in duplicate.

Immunodepletion of conditioned media

Monoclonal anti-BSG IgG (150 μl) coupled to protein-G-agarose was incubated with 2 ml conditioned media for 2 hours at 4°C. After centrifugation, immunodepletion of the supernatant was confirmed by western blotting.

SDS-PAGE, silver staining and western blot

Basigin western blot was performed as previously described (Haug et al., 2004). Briefly, cells were lysed in a modified radioimmunoprecipitation assay (RIPA) buffer containing proteinase inhibitors and resolved by SDS-PAGE using an 8% gel. Supernatants were concentrated ten times using centrifugal filter devices (Millipore, Billerica, MA) with a 10 kDa cut-off before they were loaded adjusted to the DNA content of the corresponding wells. DNA was determined according to Labarca and Paigen (Labarca and Paigen, 1980) by fluorophotometry using Hoechst 33258 dye.

Gels were either stained using GelCode SilverSNAP kit (Pierce) or immunoblotted with antibodies against BSG (Serotec), MMP-1 (Chemicon, Temecula, CA) and MMP-2 (R&D Systems) and visualized by chemiluminescence.

Zymography

Supernatant adjusted to the DNA content of the corresponding well mixed with sample buffer (0.98 M Tris, pH 6.8, 1% SDS, 4% glycerol, 0.006% bromphenol blue) was separated on a 7% polyacrylamide gel containing 1 mg/ml gelatin. Thereafter, gels were soaked in 2.5% Triton X-100 for 1 hour, followed by incubation in developing buffer (50 mM Tris, pH 7.5, 200 mM NaCl, 5 mM CaCl2, 0.2% Brij-35) at 37°C for 16 hours.

Immunohistology and cytology

For detection of BSG, iso-α smooth muscle actin, fibronectin, collagen type I and pancytokeratin cells seeded on glass coverslips and cryosections of human pancreatic adenocarcinomas or tumors induced in nude-mice were fixed in ice-cold acetone. After fixation, unspecific binding sites were blocked by incubation with TNB (0.1 M Tris, 0.15 M NaCl, 0.5% bovine serum albumin, pH 7.5). Subsequently, the specimens were incubated with anti-MMP-2 (R&D Systems), anti-BSG (clone MEM-M6/1; Serotec), anti-iso-α smooth muscle actin (DAKO), anti-pancytokeratin (clone MNF-116; DAKO), anti-collagen type I (Chemicon) or anti-fibronectin (Behring, Marburg, Germany) antibodies diluted in TNB. Immunoreactivity was visualized using tyramide signal amplification (NEN Life Science Products, Boston, MA) and a streptavidin-conjugated fluorochrome or alkaline phosphatase-anti-alkaline phosphatase (APAAP) staining using FastRed (DAKO) as substrate. Nuclei were counterstained with propidium iodide, Hoechst 33258 or Mayer's hemalaun.

RNA-isolation and PCR

RNA was extracted from MiaPaCa, Panc1 and SW850 by phenol/chloroform following the method described by Chomczynski and Sacchi (Chomczynski and Sacchi, 1987). mRNA (1 μg) was used to synthesize complementary DNA using 200 U SuperScript reverse-transcriptase (Invitrogen, Carlsbad, CA, USA) and 0.5 μg oligo(dT) primer. PCR amplification was performed on a light-cycler platform using the Roche Light-Cycler Fast-Start SYBR-Green DNA Kit (Roche, Mannheim, Germany) according to the instructions of the manufacturer. The sequence of the sense primer was 5′-CCGAGGACGTCCTGGAT-3′ and 5′-CGGGCCACCTGCCTCA-3′ for the antisense primer. Negative controls comprised replacement of either cDNA or reverse transcriptase with water. XS13 was used as a housekeeping gene. The PCR product was analyzed by etidium-bromide staining on a 1.2% agarose gel. The predicted size of the PCR product was 120 bp.

CAM assay

The CAM assay was performed as previously described (Kunzi-Rapp et al., 2001). Briefly, 0.5×106 Panc1 cells were seeded either alone or in combination with 0.5×106 PSC on the CAM of fertilized chicken eggs. Following four days of incubation at 37°C and 60% relative humidity, the tumor cells with the surrounding CAM were sampled, fixed in 4% paraformaldehyde, paraffin-embedded and stained with hematoxylin-eosin for histologic examination.

Animals and tumors

Xenografts of all carcinoma cell lines were initiated by subcutaneous injection of 2×106 tumor cells into six nude mice on the right-side alone or in combination with 1.8×106 PSC on the left-side. After 11 days the mice were sacrificed and the tumors were dissected, weighed and prepared for further histological examination. All experiments were performed in accordance with institutional guidelines for the care and use of experimental animals.

Statistical analysis

Multivariate analysis of variance followed by Fischer LSD post-hoc analysis was performed using Statistica® 6.1 for Windows® to test for significant statistical differences between groups.

This work was supported in part by the DFG grant SFB 518 (A7).

Apte, M. V., Haber, P. S., Applegate, T. L., Norton, I. D., McCaughan, G. W., Korsten, M. A., Pirola, R. C. and Wilson, J. S. (
1998
). Periacinar stellate shaped cells in rat pancreas: identification, isolation, and culture.
Gut
43
,
128
-133.
Bachem, M. G., Schneider, E., Gross, H., Weidenbach, H., Schmid, R. M., Menke, A., Siech, M., Beger, H., Grunert, A. and Adler, G. (
1998
). Identification, culture, and characterization of pancreatic stellate cells in rats and humans.
Gastroenterology
115
,
421
-432.
Bachem, M. G., Schunemann, M., Ramadani, M., Siech, M., Beger, H., Buck, A., Zhou, S., Schmid-Kotsas, A. and Adler, G. (
2005
). Pancreatic carcinoma cells induce fibrosis by stimulating proliferation and matrix synthesis of stellate cells.
Gastroenterology
128
,
907
-921.
Biswas, C., Zhang, Y., DeCastro, R., Guo, H., Nakamura, T., Kataoka, H. and Nabeshima, K. (
1995
). The human tumor cell-derived collagenase stimulatory factor (renamed EMMPRIN) is a member of the immunoglobulin superfamily.
Cancer Res.
55
,
434
-439.
Buchholz, M., Kestler, H. A., Holzmann, K., Ellenrieder, V., Schneiderhan, W., Siech, M., Adler, G., Bachem, M. G. and Gress, T. M. (
2005
). Transcriptome analysis of human hepatic and pancreatic stellate cells: organ-specific variations of a common transcriptional phenotype.
J. Mol. Med.
83
,
795
-805.
Caudroy, S., Polette, M., Nawrocki-Raby, B., Cao, J., Toole, B. P., Zucker, S. and Birembaut, P. (
2002
). EMMPRIN-mediated MMP regulation in tumor and endothelial cells.
Clin. Exp. Metastasis
19
,
697
-702.
Chomczynski, P. and Sacchi, N. (
1987
). Single-step method of RNA isolation by acid guanidinium thiocyanate-phenol-chloroform extraction.
Anal. Biochem.
162
,
156
-159.
Dalberg, K., Eriksson, E., Enberg, U., Kjellman, M. and Backdahl, M. (
2000
). Gelatinase A, membrane type 1 matrix metalloproteinase, and extracellular matrix metalloproteinase inducer mRNA expression: correlation with invasive growth of breast cancer.
World J. Surg.
24
,
334
-340.
Davidson, B., Goldberg, I., Berner, A., Kristensen, G. B. and Reich, R. (
2003
). EMMPRIN (extracellular matrix metalloproteinase inducer) is a novel marker of poor outcome in serous ovarian carcinoma.
Clin. Exp. Metastasis
20
,
161
-169.
De Wever, O. and Mareel, M. (
2002
). Role of myofibroblasts at the invasion front.
Biol. Chem.
383
,
55
-67.
Egeblad, M. and Werb, Z. (
2002
). New functions for the matrix metalloproteinases in cancer progression.
Nat. Rev. Cancer
2
,
161
-174.
Ellenrieder, V., Alber, B., Lacher, U., Hendler, S. F., Menke, A., Boeck, W., Wagner, M., Wilda, M., Friess, H., Buchler, M. et al. (
2000
). Role of MT-MMPs and MMP-2 in pancreatic cancer progression.
Int. J. Cancer
85
,
14
-20.
Friedman, S. L. (
1993
). Seminars in medicine of the Beth Israel Hospital, Boston. The cellular basis of hepatic fibrosis. Mechanisms and treatment strategies.
New Engl. J. Med.
328
,
1828
-1835.
Haber, P. S., Keogh, G. W., Apte, M. V., Moran, C. S., Stewart, N. L., Crawford, D. H., Pirola, R. C., McCaughan, G. W., Ramm, G. A. and Wilson, J. S. (
1999
). Activation of pancreatic stellate cells in human and experimental pancreatic fibrosis.
Am. J. Pathol.
155
,
1087
-1095.
Haug, C., Lenz, C., Diaz, F. and Bachem, M. G. (
2004
). Oxidized low-density lipoproteins stimulate extracellular matrix metalloproteinase Inducer (EMMPRIN) release by coronary smooth muscle cells.
Arterioscler. Thromb. Vasc. Biol.
24
,
1823
-1829.
Itoh, T., Tanioka, M., Matsuda, H., Nishimoto, H., Yoshioka, T., Suzuki, R. and Uehira, M. (
1999
). Experimental metastasis is suppressed in MMP-9-deficient mice.
Clin. Exp. Metastasis
17
,
177
-181.
Jemal, A., Siegel, R., Ward, E., Murray, T., Xu, J., Smigal, C. and Thun, M. J. (
2006
). Cancer statistics, 2006.
CA Cancer J. Clin.
56
,
106
-130.
Kanekura, T., Chen, X. and Kanzaki, T. (
2002
). Basigin (CD147) is expressed on melanoma cells and induces tumor cell invasion by stimulating production of matrix metalloproteinases by fibroblasts.
Int. J. Cancer
99
,
520
-528.
Kunz-Schughart, L. A. and Knuechel, R. (
2002
). Tumor-associated fibroblasts (part I): Active stromal participants in tumor development and progression?
Histol. Histopathol.
17
,
599
-621.
Kunzi-Rapp, K., Genze, F., Kufer, R., Reich, E., Hautmann, R. E. and Gschwend, J. E. (
2001
). Chorioallantoic membrane assay: vascularized 3-dimensional cell culture system for human prostate cancer cells as an animal substitute model.
J. Urol.
166
,
1502
-1507.
Labarca, C. and Paigen, K. (
1980
). A simple, rapid, and sensitive DNA assay procedure.
Anal. Biochem.
102
,
344
-352.
Landis, S. H., Murray, T., Bolden, S. and Wingo, P. A. (
1999
). Cancer statistics, 1999.
CA Cancer J. Clin.
49
,
8
-31.
Li, R., Huang, L., Guo, H. and Toole, B. P. (
2001
). Basigin (murine EMMPRIN) stimulates matrix metalloproteinase production by fibroblasts.
J. Cell. Physiol.
186
,
371
-379.
Liotta, L. A. and Kohn, E. C. (
2001
). The microenvironment of the tumour-host interface.
Nature
411
,
375
-379.
Marieb, E. A., Zoltan-Jones, A., Li, R., Misra, S., Ghatak, S., Cao, J., Zucker, S. and Toole, B. P. (
2004
). Emmprin promotes anchorage-independent growth in human mammary carcinoma cells by stimulating hyaluronan production.
Cancer Res.
64
,
1229
-1232.
Marionnet, C., Lalou, C., Mollier, K., Chazal, M., Delestaing, G., Compan, D., Verola, O., Vilmer, C., Cuminet, J., Dubertret, L. et al. (
2003
). Differential molecular profiling between skin carcinomas reveals four newly reported genes potentially implicated in squamous cell carcinoma development.
Oncogene
22
,
3500
-3505.
Noel, A. C., Polette, M., Lewalle, J. M., Munaut, C., Emonard, H. P., Birembaut, P. and Foidart, J. M. (
1994
). Coordinate enhancement of gelatinase A mRNA and activity levels in human fibroblasts in response to breast-adenocarcinoma cells.
Int. J. Cancer
56
,
331
-336.
Phillips, P. A., McCarroll, J. A., Park, S., Wu, M. J., Pirola, R., Korsten, M., Wilson, J. S. and Apte, M. V. (
2003
). Rat pancreatic stellate cells secrete matrix metalloproteinases: implications for extracellular matrix turnover.
Gut
52
,
275
-282.
Sameshima, T., Nabeshima, K., Toole, B. P., Yokogami, K., Okada, Y., Goya, T., Koono, M. and Wakisaka, S. (
2000
). Glioma cell extracellular matrix metalloproteinase inducer (EMMPRIN) (CD147) stimulates production of membrane-type matrix metalloproteinases and activated gelatinase A in co-cultures with brain-derived fibroblasts.
Cancer Lett.
157
,
177
-184.
Seymour, A. B., Hruban, R. H., Redston, M., Caldas, C., Powell, S. M., Kinzler, K. W., Yeo, C. J. and Kern, S. E. (
1994
). Allelotype of pancreatic adenocarcinoma.
Cancer Res.
54
,
2761
-2764.
Sidhu, S. S., Mengistab, A. T., Tauscher, A. N., LaVail, J. and Basbaum, C. (
2004
). The microvesicle as a vehicle for EMMPRIN in tumor-stromal interactions.
Oncogene
23
,
956
-963.
Smeenk, H. G., Tran, T. C., Erdmann, J., van Eijck, C. H. and Jeekel, J. (
2005
). Survival after surgical management of pancreatic adenocarcinoma: does curative and radical surgery truly exist?
Langenbecks Arch. Surg.
390
,
94
-103.
Tang, W., Chang, S. B. and Hemler, M. E. (
2004
). Links between CD147 function, glycosylation, and caveolin-1.
Mol. Biol. Cell
15
,
4043
-4050.
Taylor, P. M., Woodfield, R. J., Hodgkin, M. N., Pettitt, T. R., Martin, A., Kerr, D. J. and Wakelam, M. J. (
2002
). Breast cancer cell-derived EMMPRIN stimulates fibroblast MMP2 release through a phospholipase A(2) and 5-lipoxygenase catalyzed pathway.
Oncogene
21
,
5765
-5772.
Thorns, C., Feller, A. C. and Merz, H. (
2002
). EMMPRIN (CD 174) is expressed in Hodgkin's lymphoma and anaplastic large cell lymphoma. An immunohistochemical study of 60 cases.
Anticancer Res.
22
,
1983
-1986.
Zucker, S., Cao, J. and Chen, W. T. (
2000
). Critical appraisal of the use of matrix metalloproteinase inhibitors in cancer treatment.
Oncogene
19
,
6642
-6650.
Zucker, S., Hymowitz, M., Rollo, E. E., Mann, R., Conner, C. E., Cao, J., Foda, H. D., Tompkins, D. C. and Toole, B. P. (
2001
). Tumorigenic potential of extracellular matrix metalloproteinase inducer.
Am. J. Pathol.
158
,
1921
-1928.