The tissue samples were obtained from 20 patients with liver metastases of colorectal cancer. The tissue was obtained from partial hepatic resections. All patients in the study signed an informed consent form. The study protocol was approved by the local research ethics committee of the Second Affiliated Hospital of Zhengzhou University. The tissue samples were mechanically crushed with sterile forceps for 60 min at 37°C in 30 ml of medium 199 (Invitrogen, Carlsbad, CA, USA), and incubated in 1 mg/ml collagenase type IV (Sigma, Steinheim, Germany). Thereafter, the digested tissue pieces were placed in 30 ml Dulbecco’s modified Eagle’s medium (DMEM) with 10% fetal calf serum (FCS, Invitrogen) and 1% antibiotic/antimycotic solution (Invitrogen), consisting of penicillin, streptomycin and amphotericin B. Subsequently, samples were transferred to 145 cm² culture plates (Nunclon™) and cultured in a 5% CO₂-containing humidified atmosphere at 37°C. The first medium change occurred once the cells adhered, after approximately 48 to 72 h.

To characterize the tumor-free liver tissue phenotype, the CAFs in human metastatic tissue and tumor-free liver fibroblasts were stained with the following primary antibodies (monoclonal mouse anti-human antibodies) in PBS: α-SMA (1/50, clone 1A4, IgG2a; Dako, Glostrup, Denmark), vimentin (1/50, clone V9, Dako), Thy-1 (1/25, clone 5E10, BD Biosciences, Bedford, MA, USA), desmin (1/50, clone D33, Dako), CD45 (1/50, clone T29/33, Dako), cytokeratin 7 (1/50 clone OV-TL 12/30, Dako), cytokeratin 19 (1/100, RCK108 clone, Dako), NCAM (1/25, clone T1, Dako), ICAM-1 (1/50, clone 6.5B5, Dako), laminin (1/25, clone 4C7, Dako) and glial fibrillary acidic protein (GFAP, 1/25, clone 6F2, Dako). The staining against CD45 was carried out according to the protocol of DakoCytomation EnVision and System-HRP (Dako Co., Carpinteria, CA, USA).

In order to extract the total RNA from tissue samples, the RNA Midi kit (Qiagen, Hilden, Germany) was used. The sample was crushed by a disintegrator (MicroDismembrator, Braun Instrumente, Melsungen, Germany), and lysed in 3.8 ml RLT buffer (Qiagen). The lysed sample was centrifuged for 5 min at 4,000 rpm. Ethanol (70%; 3.8 ml) was added into the supernatant, then the supernatant was planned and mixed for 1 min. To fix the RNA in the membranes of the kit columns, several wash steps were followed according to the manufacturer’s instructions. In order to dissolve the RNA from the membrane, the filter was transferred to a new 15 ml collection tube and washed twice with 250 ml of RNase-free water (Sigma). In order to determine the concentration, an aliquot was produced at a dilution of 1/50 in RNase-free water from the kit, and the absorbance was measured at 260 and 280 nm photometrically (Hach Lange DR 2800 VIS photometer, Germany).

The transfer of nucleic acids to a nylon membrane was initially from a gel. Based on the concentration gradient of SSC solutions, the transport of RNA was performed by the gel onto the membrane. A total of 20 h after the transfer was canceled using a UV cross linker (Hoefer™, Amersham Biosciences UK Ltd., UK), the RNA could be linked to the nylon membrane. For subsequent assignment of the bands sizes, the signals of the 18S and 28S rRNA subunits were marked under ultraviolet light. The membrane was either used immediately or sealed and stored in 20X SSC at 4°C.

Total RNA of the liver samples was transferred to the inward-facing RNA-side of the nylon membrane in hybridization tubes filled with 10 ml EasyHyb hybridization buffer (Roche, USA). The RNA with membrane-covered points was prehybridized for 1 h at 50°C in a rotary hybridization oven (Amersham-Pharmacia, UK). The nonspecific binding of DIG antibody was used to prevent the blocking of the membrane. A hybridization probe was heated for 10 min at 68°C in a water bath, then stored at −20°C. After the expiry of prehybridization, buffer was discarded, and the membrane was hybridized at 50°C under constant rotation overnight. Following hybridization, the probe was frozen at −20°C again. Thereafter, the membrane was washed twice for 15 min at 20°C in wash buffer (2X SSC + 0.1% SDS) and twice for 15 min at 68°C in wash buffer (0.5X SSC + 0.1% SDS).

For measurement of PAI-1 concentration in the culture supernatant, the human PAI ELISA kit was used. Firstly, 100 μl anti-mouse HRP-conjugated PAI-1 antibodies were placed into each well. The microplate was coated with another monoclonal anti-PAI-1 antibody. The 100 μl standard sample was introduced directly into the appropriate wells, after which the immune reaction started. The PAI-1-Ag was bound to a defined epitope on the solid phase-bound monoclonal anti-PAI-1 antibodies and bound to another HRP-conjugated anti-PAI-1 monoclonal antibody defined epitope. Following gentle mixing, the plate was incubated at room temperature for 1 h and then washed with 300 μl wash solution (at least 20-fold concentrated). After the unbound proteins were washed, 200 μl of peroxidase substrate 3,3′,5,5′-tetramethylbenzidine (TMB)/H₂O₂ was added for 5 min at room temperature, and the reaction was stopped with 50 μl 0.45 M sulfuric acid. Thus the formation of a dye was induced, and the intensity was directly proportional to the concentration of the human PAI-1 in the sample. The color stabilization lasted 10 min. The absorbance was read at 450 nm and the blank value was subtracted. A microplate reader (Dynatech MR 5000) and the software Microwin 3.0 (Microtek) were applied for measurement and evaluation.

The result of each experiment was shown as the mean ± standard deviation (SD) where applicable. Statistically significant differences in each assay were determined by SPSS version 18.0 statistical software package (SPSS Inc., Chicago, IL, USA). Differences in each group were tested for significance using the ANOVA analysis of variance. P<0.05 was considered to indicate a statistically significant difference.

Fibroblastic cell populations were classified according to their immunohistochemical marker profile. We observed the cell populations in histological sections of colorectal cancer liver metastases (at least 3 cm). The cells present within the metastatic stroma showed differentiation, as occurred in stationary fibroblasts in continuous portal vessels of the liver tissue. Therefore, fibroblasts could be identified by co-expression of vimentin, Thy-1 and α-SMA, while perisinusoidal hepatic stellate cells (HSCs) were localized in the normal liver. Due to the staining in the transition area between the liver tissue and metastases, the phenomenon of fibroblastic cells in the liver tumor could be observed, as shown in Fig. 1A-C, and α-SMA, vimentin and Thy-1 were positive in tumor cells.

Clear staining of the marker Thy-1 is shown in Fig. 2. The sections showed marked staining in the septa. Thy-1 positive fibroblasts were distributed throughout the stroma and were intensified in the border region between the tumor and liver tissue (Fig. 2B). In healthy tissue, predominantly the hepatic portal area, but also the vascular wall and bile duct structures were stained (Fig. 2A).

Using a double immunofluorescence stain against α-SMA and vimentin, the proportion of α-SMA-positive cells was quantified. While all the CAFs were positive for vimentin, the percentage of α-SMA-positive cells was between 50 and 90% (Fig. 3A-B). The majority of the CAFs therefore had a vimentin+/α-SMA+/Thy-1+ fibroblast phenotype. This expression profile matched that of the CAFs in situ. Thus it was concluded that the majority of cultured CAFs was representative of the population found in fibroblastic tumors in situ.

CAFs are the targets for both cellular growth factors and pro-angiogenic factors. Therefore, CAFs are likely to play a key role in hepatic invasion and metastasis in colorectal carcinoma, as well as endothelial cells. To determine the effect of growth factors on the properties of CAFs, the established cell lines were incubated with increasing concentrations of TGF-1, PDGF, tumor necrosis factor (TNF) and epidermal growth factor (EGF). Subsequently, mRNA was isolated from the cells and analyzed by northern blotting. As shown in Fig. 4A-B, α-SMA mRNA expression increased significantly after CAFs were treated with the growth factor TGF-1 (P<0.05), while EGF and TNF led to no increase in α-SMA mRNA expression, and PDGF caused a significant suppression of α-SMA expression (P<0.05).

Compared with CAFs without stimulation, the PAI-1 mRNA expression significantly increased in those stimulated by TGF-1 (5 ng/ml), as shown in Fig. 5A and B. The PAI-1 mRNA expression was significantly lower in unstimulated CAFs (P<0.01).

The PAI-1 expression in CAFs from colorectal liver metastases and liver fibroblasts from healthy liver tissue was also analyzed by ELISA, as shown in Fig. 6. The strongest PAI-1 expression was 93.77 ng/ml in CAFs cultured with 10% FCS (control, 53.9 ng/ml), while it was 79.26 ng/ml in liver fibroblasts cultured with 10% FCS (control, 24.49 ng/ml). The influence of different growth factors on cells was strong. The average PAI-1 expression in the fibroblasts from the healthy liver tissues increased to 43.445 ng/ml under TGF-1 culture with 39.01 ng/ml for PDGF, while the average PAI-1 expression in CAFs increased to 75.45 ng/ml under TGF-1 culture, with 59.53 ng/ml for PDGF. Overall, the PAI-1 production increase in normal tissue was less than that in the metastatic CAFs.