Transforming Growth Factor-β1 Induces Desmoplasia in an Experimental Model of Human Pancreatic Carcinoma¹

Desmoplasia is a characteristic feature of the growth of some carcinomas (1). To date, it is not clear whether this process is a mechanism to protect the tumor from the host or represents a defense mechanism by the host (2), although there are hints that this stroma is beneficial for the tumor (3). To tackle desmoplasia therapeutically by either supporting or suppressing this development, it becomes necessary to study the etiology and to attribute this feature to either the tumor cells themselves or the host. Desmoplasia is of particular predominance in ductal adenocarcinomas of the pancreas exhibiting a strong stromal reaction(4). Therefore, pancreatic carcinoma has become a model system to study the interrelation of epithelial tumor cells, matrices,fibroblasts, and growth factors (5, 6, 7, 8).

Desmoplastic tissue consists of fibroblasts, as the main cellular component, and extracellular matrix proteins (9). The pancreatic tumor cells themselves are able to produce ECM³proteins (10, 11, 12, 13) and interact with ECM by expressing functionally active integrins (6, 14, 15).

To test the hypothesis of desmoplasia induction by a tumor-derived growth factor, we conducted a deductive analysis correlating the ability to induce desmoplasia with the expression of certain growth factors. Furthermore, we reasoned that the overexpression of such a growth factor, e.g., TGF-β1 in a pancreatic tumor cell line known neither to induce desmoplasia nor to express substantial amounts of TGF-β1 and FGF-2, should result in the gain of the ability to induce fibroblast growth and in an induction of desmoplasia in a xenografted nude mouse model by virtue of direct and indirect effects of TGF-β1.

AsPC-1, BxPC-3, Capan-1, and PANC-1 cells, all from American Type Culture Collection, were cultivated in DMEM with GlutaMAX I (Life Technologies, Inc.) supplemented with 10% heat-inactivated FCS and antibiotics (100 units/ml penicillin, 100 μg/ml streptomycinsulfate,and 250 ng/ml amphotericin B; Life Technologies, Inc.; Ref.10). Mature human recombinant TGF-β1 was purchased from R&D Systems. Full-length cDNA of TGF-β1(16) was cut out of pRK5β1E (BamHI) and cloned into the pcDNA3 vector(Invitrogen) under the control of a cytomegalovirus promoter. PANC-1 cells were transfected with this construct or with the empty pcDNA3 plasmid (mock) by calcium phosphate coprecipitation in N,N-bis(2-hydroxyethyl)-2-aminoethanesulfonic acid buffered saline using standard protocols as described(17). This plasmid also codes for the neoresistance gene, enabling selection of transfectants with the antibiotic G418 (Sigma; 400 μg/ml). Resistant clones were expanded,and expression of the transfected cDNA was confirmed by Northern blot,Western blot, and ELISA (R&D).

Subconfluent layers of PANC-1/TGF-β1 cells, mock transfected,untransfected PANC-1 cells, and AsPC-1 and BxPC-3 cells were lysed in ice-cold guanidine thiocyanate. RNA preparation was performed as described (18). Ten μg of total RNA were subjected to standard formamide gel electrophoresis as described. Gels were blotted to nylon membranes (Qiagen) and hybridized with cDNA probes for TGF-β1 (EcoRI/HindIII digest of pcDNA3/TGF-β1), type I collagen (pHCAL1U; Refs. 10 and19), PDGF (Amersham), FGF-2,(20), and CTGF(21) using the nonradioactive Dig labeling kit (Boehringer Mannheim, Mannheim, Germany). In addition, RT-PCR was performed using published primers for TGF-β1, PDGF-A, type I collagen, and GAPDH. The primers were as follows: TGF-β1 (22), sense 5′-CAG AAA TAC AGC AAC AAT TCC TGG-3′ and antisense 5′-TTG CAG TGT GTT ATC CCT GCT GTC-3′ (190-bp product); PDGF-A (23), sense 5′CAG TCA GAT CCA CAG CAT CC-3′ and antisense 5′-AAT GAC CGT CCT GGT CTT GC-3′(200-bp product); collagen type I (23), sense 5′-ACG TGA TCT GTG ACG AGA CC-3′ and antisense 5′-AGC AAA GTT TCC TCC GAG GC-3′(250-bp product); and GAPDH (24), sense 5′-ACC ACA GTC CAT GCC ATC AC-3′ and antisense 5′-TCC ACC ACC CTG TTG CTG TA-3′ (450-bp product). PCR conditions were the following: denaturing for 30 s at 94°C; annealing for 60 s at 60°C (TGF-β1) or at 64°C(collagen, GAPDH, and PDGF); and extension for 60 s at 72°C. Amplified DNA was sampled after 21, 24, 27, and 30 cycles, and the resulting PCR products for TGF-β1, collagen, and PDGF-A were loaded in the same gel pockets as the GAPDH amplificate.

Expression of genes of several growth factors, receptors, and genes of ECM proteins was investigated by reverse slot blot. For this purpose,plasmid DNA corresponding to 1 μg of cDNA insert was blotted onto a nylon membrane (Qiagen) by use of a slot blot apparatus (Schleicher &Schuell). Hybridization was performed according to standard procedures with a probe obtained by Dig labeling (Boehringer Mannheim) of 7.5 μg of total RNA in a reverse transcription reaction (25, 26). Hybrids were detected using the chemiluminescent Dig detection system(Boehringer Mannheim) according to the manufacturer’s instructions.

PANC-1/TGFβ1 cells (5 × 10⁴)were seeded onto Transwell inserts (Costar) and were cocultivated with fibroblasts (5 × 10⁴ cells/well)seeded in six-well tissue culture plates. After 7 days of incubation in DMEM/1% FCS, the cells were trypsinized and counted (trypan blue exclusion test). Controls were cocultivated of fibroblasts with fibroblasts, PANC-1, PANC-1 mock transfected, and BxPC-1 cells(American Type Culture Collection). From respective parallel experiments, conditioned media (see below) and RNA (see above) were prepared for subsequent analysis.

PANC-1/TGFβ1 cells were seeded in DMEM/10% FCS. After 2 days of incubation, cells were washed three times with PBS (pH 7.4) to remove any FCS traces and refed with fresh medium containing no FCS. After incubation for another 2 days, the supernatants were collected and filter sterilized. Skin fibroblasts were incubated with serial dilutions of this concentrated medium for 2 days, and induction of proliferation was investigated by the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide test(Boehringer Mannheim). Controls included conditioned media of PANC-1 and mock-transfected PANC-1 cells.

1.3 × 10⁴ cells of TGF-β1-transfected and mock-transfected PANC-1 were plated in six-well plates with DMEM and 10% FCS. After 2 days, cells were grown with DMEM without FCS (transfected cells all of the time with 400μg/ml G418) for 1, 2, or 3 days, after that the supernatant was collected. TGF-β1 and PDGF were quantified using the Quantikine TGF-β1 and PDGF immunoassays (R&D) according to the instructions of the manufacturer.

Proteins were separated by SDS-PAGE, transferred to a polyvinylidene difluoride membrane (Roche), and blocked for 1 h in Tris-buffered saline (TBS; 10 mm Tris, 10 mm NaCl) containing 1% skim milk and 0.01% Tween 20. After incubation with the primary antibody for 1 h, blots were developed using alkaline phosphatase-labeled secondary antibodies and chemiluminescence(CDP-star; Roche). The following primary antibodies were used in a dilution of 1:1000: PCNA (Santa Cruz; sc-56), TGF-β-1 (sc-146),p21^wafI (sc-6246), p-Tyr (sc-7020), Erk 1(sc-94-G), Erk 2 (sc-1647), and Erk 3 (sc-6268). As detection antibodies, mouse-antigoat immunoglobulin (Dako; 1:5000),rabbit-antimouse immunoglobulin (Dako; 1:5000), and swine-antirabbit immunoglobulin-AP (Dako, 1:5000) were used (27).

Cells were scraped, washed with Tris-buffered saline, resuspended in hypotonic buffer (10 mm HEPES, 10 mm KCl, 1.5 mm MgCl₂, and 0.5 mmEDTA), and allowed to swell on ice for 20 min. The nuclei were collected by centrifugation at 12,000 × gfor 5 min in a microcentrifuge and analyzed by Western blotting(28).

A suspension of 1 × 10⁶PANC-1/TGFβ1 cells or mock-transfected PANC-1 cells were injected orthotopically into athymic nude mice (29, 30). Nude mice were killed after solid tumors were palpable. Tumors were removed,fixed in 4% formaldehyde, and examined after H&E or Masson-Goldner trichrome staining. Immunocytochemistry was performed as described before with antibodies against type I collagen (1:100; Calbiochem) and fibronectin (1:400; Sigma; Ref. 10). Detection was performed using horseradish peroxidase-conjugated rabbit-antimouse and swine-antirabbit IgGs (Dako Diagnostika) as secondary and third antibodies and 3-amino-9-ethylcarbazole as substrate.

In a deductive analysis on human pancreatic carcinoma cells in vitro and in vivo (6, 31), using all of the published information on the expression of various growth factors described in pancreatic carcinoma (22, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42), the stromal reaction (Fig. 1) was found to be positively correlated with the expression of TGF-β1 and/or FGF-2 (Ref. 31; Table 1). We therefore chose PANC-1 cells that did not express significant amounts of TGF-β1 as a model for the subsequent experiments investigating the role of TGF-β1 in desmoplasia.

Expression of the transfected TGF-β1 cDNA in PANC-1/TGF-β1 cells was verified by Northern and Western blots (Figs. 2,A and 3 A). The TGF-β1 protein was released into the culture medium as demonstrated by ELISA of serum-free supernatants; it was native, i.e., inactive, and had to be activated by acidification before quantification.

In the TGF-β1-transfected PANC-1 cells, the expression of collagen type I was increased (Fig. 2,B). Also, PDGF-A was increased(Fig. 2,C), whereas the expression of FGF-2 (data not shown),epidermal growth factor, and the α₅ integrin subunit (Fig. 2 C) was similar in TGF-β1-transfected and mock-transfected PANC-1 cells by Northern blot or reverse slot blot.

TGF-β1 inhibits growth by acting on the cell cycle by modulating, for example, p21^wafI and PCNA. TGF-β1-transfected PANC-1 cells exhibited a substantial increase in p21^wafI expression on the protein level on Western blot of nuclear extracts (Fig. 3,B); on ELISA,p21^wafI was 5.4 units/mg protein in untransfected and 16.3 units/mg protein in transfected cells. Conclusively, the transfected cells demonstrated decreased nuclear levels of PCNA on the protein level (Fig. 3 B).

Cocultivation of fibroblasts with PANC-1/TGF-β1 cells in the Transwell system led to an increase in proliferation of both the fibroblasts and the tumor cells (Fig. 4,A), whereas cocultivation with mock-transfected PANC-1 cells did not exhibit this effect. On the RNA level, induction of collagen type I in the fibroblasts could be demonstrated after incubation with conditioned media from TGF-β1-transfected PANC-1 cells (Fig. 4,B); moreover, an up-regulation of TGF-β1 expression could be demonstrated by RT-PCR under this conditions (Fig. 4,B). Although CTGF expression remained unchanged in the PANC-1 cells after TGF-β1 transfection (data not shown), a significant increase in CTGF mRNA was detectable in fibroblasts after cocultivation with TGF-β1-transfected PANC-1 cells (Fig. 4 B).

Incubation of fibroblasts in conditioned media of TGF-β1-transfected PANC-1 cells resulted in more pronounced tyrosine phosphorylation of proteins in fibroblasts than incubation with supernatants from mock-transfected and untransfected PANC-1 cells (Fig. 5, top). Furthermore, mitogen-activated protein kinases were activated as indicated by a mobility shift of Erk 1/2 (Fig. 5, bottom). As mentioned with the tyrosine phosphorylation, the most pronounced phosphorylation of Erk 1/2 and 3 could be demonstrated after incubation with supernatants of PANC-1/TGF-β1 (Fig. 5, bottom). Here, an increase in the activated, i.e., phosphorylated, kinases was evident.

PANC-1/TGF-β1 transfected cells and mock-transfected cells were injected orthotopically into the nude mouse pancreas. Tumors were harvested after 2 months. The tumors grown from TGF-β1-transfected cells demonstrated an increased desmoplasia surrounding the tumor cells as compared with the mock-transfected cells (Fig. 6). This was evident both on the tumor margin toward the normal mouse pancreas as well as within the tumor. In addition, collagen type I and fibronectin could be detected in increased amounts surrounding the tumor cells (Fig. 6).

The desmoplastic reaction is one of the morphological hallmarks of several human tumors (1) originating from solid epithelial glands, such as pancreatic adenocarcinoma, that sets it apart from other epithelial tumors. Beside the description and static expression analysis of potential factors, no detailed analysis has been performed to dissect this phenomenon. The pancreatic tumor cells themselves produce matrix proteins (10) and express a variety of integrins (6, 15). Furthermore, the expression of growth factors and their receptors has been demonstrated conclusively,however, mostly related to a demonstration of the autocrine growth-promoting effect (35, 36, 43).

To test our hypothesis of a positive correlation of stroma induction and TGF-β1 expression, we successfully transfected the tumor cell line PANC-1 with a TGF-β1 expression vector.

For TGF-β1, it has been suggested that the major regulatory step controlling TGF-β1 activity takes place extracellularly. The same was true for the transfected PANC-1 cells; TGF-β1 was released into the culture medium in a latent, i.e., not activated, state. Recently, it was demonstrated that latent TGF-β1 can bind to and be activated by the α_vβ₆integrin (44). These integrin subunits are also expressed by pancreatic carcinomas (15), i.e., the PANC-1 cells (6). Thus, activation of the released TGF-β1 may be accomplished in this way. Expression of TGF-β1 resulted in an up-regulation of the matrix proteins collagen type I and fibronectin in the tumor cells themselves. Furthermore, PDGF expression was increased in the transfected cells. This altered gene expression resulted in several paracrine effects on fibroblasts in cocultivation experiments. We could demonstrate an increase in collagen type I synthesis in the fibroblasts after stimulation with supernatants from TGF-β1-transfected PANC-1 cells. Similarly, the activation of collagen type VII regulatory elements by TGF has been described recently (45). The fibroblasts themselves produced more TGF-β1 upon stimulation (cocultivation or conditioned media) by the TGF-β1-transfected PANC-1 cells. This is supported by the observation that in pancreatic carcinoma tissue, TGF-β1 is most predominant in the stroma (46). Furthermore, collagen type I, the most predominant basal membrane matrix protein in pancreatic carcinoma(10), is also up-regulated, both in the tumor cells themselves and in the fibroblasts upon cocultivation. This up-regulation, however, may only be in part attributed to TGF-β1 itself; it could also be the result of the up-regulation of PDGF-A that has been shown to be a cofactor in TGF-β1-induced collagen type I stimulation (23). In the fibroblasts, after cocultivation with TGF-β1-transfected PANC-1 cells, CTGF, one of the index TGF-β1 response genes (47), was increased. Inhibition of CTGF abrogated the TGF-β1-induced collagen gene up-regulation, confirming the pivotal role of this growth factor(48). As a result of these alterations in gene expression mentioned above, the transfection of TGF-β1 in the pancreatic tumor cell line PANC-1 led to a gain of stromal tissue after orthotopic transplantation in the nude mouse when compared with mock-transfected PANC-1 cells.

The influence of the matrix on signal transduction has long been under debate (49, 50). We have shown that a single growth factor, TGF-β1, is capable of conferring the desmoplastic potential to tumor cells not capable of these features. Some of the effects may be attributed to a direct effect of TGF-β1, whereas others, e.g., the up-regulation of collagen type I(51), may be the result of indirect effects of TGF-β1 intimately associated with the signal transduction pathway involved in TGF-β1 activities.

We thank Roland M. Schmid for assistance in subcloning the TGF-β1 plasmid and Thomas Gress (both of University of Ulm, Ulm,Germany) for supplying us with the CTGF plasmid.