The following six pancreatic cancer cell lines were used: KP-2, SUIT-2 (Dr H. Iguchi, National Shikoku Cancer Center, Matsuyama, Japan), MIA PaCa-2 (Japanese Cancer Resource Bank, Tokyo, Japan), Capan-1, Capan-2 and Hs766T (American Type Culture Collection, Manassas, VI, USA). The human immortalized pancreatic ductal epithelial cell line (HPDE6-E6E7 clone 6) was kindly provided by Dr Ming-Sound Tsao (University of Toronto, Toronto, Canada). All cells were maintained as previously described (20).

We analyzed TM4SF1 expression in 74 tissue samples obtained from patients who underwent pancreatic resection for pancreatic cancer at our institution between January 2000 and August 2009. We also obtained normal pancreatic tissue samples from intact pancreas resected for bile duct cancer as control tissues. Survival was measured from the time of pancreatic resection, with death as the end-point. Overall survival and disease-free survival analyses were performed in February 2012. The median observation time for overall survival and disease-free survival was 19 months (range 2–137 months) and 11 months (range 1–137 months), respectively. Forty-nine patients died during the follow-up. All surviving patients were censored. Histological diagnosis of specimens was in accordance with the criteria of the updated World Health Organization classification (21). Tumor stage was assessed according to the UICC classification, 7th edition (22). The clinicopathological characteristics of all patients are summarized in Table I. The study was approved by the Ethics Committee of Kyushu University and conducted according to the Ethical Guidelines for Human Genome/Gene Research enacted by the Japanese Government and the Helsinki Declaration.

One-step real-time qRT-PCR using gene-specific primers was performed as described previously (23). Primer sequences were as follows: TM4SF1 forward 5′-ATTGGAAT TGCAGGATCTGG-3′ and reverse 5′-GCCGAGGGAATCAA GACATA-3′; 18S rRNA forward 5′-GTAACCCGTTGAACC CCATT-3′ and reverse 5′-CCATCCAATCGGTAGTAGCG-3′. BLAST searches were performed to ensure the specificity of all primers. Primers were purchased from Sigma Genosys (Tokyo, Japan).

Immunohistochemistry was performed using a Histofine SAB-PO kit (Nichirei, Tokyo, Japan) (24). Sections were incubated with anti-TM4SF1 antibody (HPA002823; Sigma-Aldrich, Tokyo, Japan) overnight at 4°C. Carcinoma cells were identified according to morphology and counted in at least 20 fields per section at ×200 magnification. The distribution of TM4SF1 staining was evaluated as the percentage of stained cells, and was scored as follows: 0, no staining or <10%; 1, 11–25%; 2, 26–50%; 3, 51–75%; or 4, 76–100%. Cells were also scored for staining intensity, which was scored as 0, no staining; 1, weak; 2, moderate; or 3, strong. The multiplication product (from 0–12) from these two scores was used to assign patients into one of two groups according to TM4SF1 expression, where a score of 0–4 represented low expression and a score of 6–12 represented high expression. All slides were evaluated independently by two investigators who were unaware of the clinical features of each case.

Gene silencing was carried out using siRNA (Qiagen, MA, USA) directed against human TM4SF1. Sequences were as follows: siRNA-1 (sense, 5′-ggaccacuaugucuugauutt-3′; antisense, 5′-aaucaagacauaguggucctt-3′); siRNA-2 (sense, 5′-cgaug acugggcaagaagatt-3′; antisense, 5′-ucuucuugcccagucaucgta-3′). Qiagen all-star siRNA was used as a negative control. Transfections were performed as described previously (25). All cells were used in subsequent experiments 48 h after transfection.

Migration and invasion of cultured cancer cells were assessed by counting the number of cells migrating or invading through uncoated or Matrigel-coated transwell chambers (BD Biosciences, Franklin Lakes, NJ, USA) as described previously (20,26). Cells were maintained in 10% FBS/DMEM during these assays. Cells were transfected with siRNAs 48 h prior to experimentation. Migration was determined after a 24-h period, and invasiveness was determined after a 48-h period.

SUIT-2 and Capan-2 cells were cultured in 6-well plates (1×10⁵ cells/well) in 1% FBS/DMEM with or without 10 ng/ml recombinant human TGF-β1 (R&D, Oxon, UK) for 24 and 48 h (27). At each time-point, total RNA was extracted to evaluate TM4SF1 mRNA expression by qRT-PCR. To evaluate TM4SF1 protein levels, cells were seeded in 90-mm dishes at a density of 2×10⁵ cells/dish and cultured in 1% FBS/DMEM with or without 10 ng/ml TGF-β1 prior to cell lysis.

Western blotting was performed as described previously (28). Antibodies used in this study were as follows: anti-TM4SF1 (sc-103267; Santa Cruz, CA, USA), anti-E-cadherin (no. 3195), anti-vimentin (no. 5741), anti-N-cadherin (no. 4061), anti-claudin-1 (no. 4933) (Cell Signaling Technology, Danvers, MA, USA) and anti-β actin (ab8227; Abcam, Cambridge, MA, USA).

A χ² test was used to analyze the correlation between TM4SF1 expression and clinicopathological characteristics. Survival analysis was performed using Kaplan-Meier analysis and curves were compared using the log-rank test. For in vitro experiments, values are expressed as means ± standard deviation. Comparison between two groups was performed using the Student’s t-test. Statistical significance was defined as P<0.05. All statistical analyses were performed using JMP 9.0.2 software (SAS Institute, NC, USA).

We first sought to determine the level of TM4SF1 expression in PDAC and normal pancreatic tissues. In normal pancreatic tissues, weak-to-no staining was detected in pancreatic ductal epithelial cells and TM4SF1 expression was not observed in acinar cells or stroma. Blood vessels consistently stained positive for TM4SF1, and were used as a positive control (Fig. 1A-a and -c). In tumor tissues, strong luminal membrane surface staining and a weak-to-moderate cytoplasmic staining was detected in cancer cells (Fig. 1A-b). In addition to tumor tissue samples where moderate staining was observed, there were also tumor tissue samples where little-to-no staining was observed (Fig. 1A-c).

Survival analysis was performed for patients with PDAC. The high TM4SF1 expression group showed better survival (P=0.0332; Fig. 1B) than the low expression group, and there was an obvious difference in the disease-free time between the groups (P=0.0409; Fig. 1B). Median survival was 30 months for the high expression group and 17 months for the low expression group (Table II). Next, we performed multivariate analysis based on the Cox proportional hazard model on all parameters that were found to be significant in the univariate analyses. Overall survival was significantly dependent on residual tumor (R1) status (relative risk = 2.732, P=0.0017), vascular invasion (relative risk = 2.461, P=0.0101) and low TM4SF1 expression (relative risk = 1.987, P=0.0196) (Table III).

The relationship between TM4SF1 expression and clinicopathological factors of PDAC is shown in Table IV. There was no statistically significant correlation between TM4SF1 expression and clinicopathological factors possibly due to the relatively small size of the patient cohort, although it appears that the patients with low expression have a higher tumor grade and a more advanced clinical stage. Because there was an obvious difference in the disease-free time, we analyzed the site of first recurrence (Fig. 1C) and found that the locoregional recurrence rate in the low TM4SF1 expression group (28.21%) was higher than in the high expression group (5.71%; P=0.0079) although the total recurrence rates for the two patients groups were similar (Table V). The data suggest that tumors with low TM4SF1 expression undergo local spread more frequently than those with high expression, and this may contribute to the observed survival differences.

We also determined the level of TM4SF1 expression in a variety of pancreatic cancer cell lines, and the level of TM4SF1 mRNA expression was found to be higher in pancreatic cancer cell lines than in normal (HPDE) epithelial cells (Fig. 2A). Because local tumor invasion is more likely in patients with low TM4SF1 tumor expression, and because TM4SF1 has been reported to be involved in the migration of cancer cells (14,15), we next performed migration and invasion assays following knockdown of TM4SF1 in pancreatic cancer cells. Decreased TM4SF1 protein expression was confirmed by western blotting following transfection of cells with TM4SF1-specific siRNA (Fig. 2B). The invasion and migration of two cell lines studied, SUIT-2 and Capan-2, were enhanced following siRNA-mediated knockdown of TM4SF1 (Fig. 2C). These suggest that TM4SF1 has an inhibitory role in the invasion and migration of pancreatic cancer cells.

Because the process of cancer invasion requires epithelial tumor cells to undergo epithelial to mesenchymal transition (EMT) either transiently or stably (29), we examined the correlation between the expression of TM4SF1 and EMT markers in vitro. We found that E-cadherin expression was reduced after knockdown of TM4SF1, but there were no changes in the expression of other makers including vimentin, N-cadherin and claudin-1 (Fig. 3A). Changes in cell morphology, however, were not observed after knockdown of TM4SF1 (data not shown).

TGF-β1 is a major inducer of EMT in a number of cellular contexts (30), and plays a pivotal role in driving EMT in the pathogenesis of pancreatic cancer (27,31). To investigate whether TGF-β1 influences TM4SF1 expression, we performed additional in vitro analyses on pancreatic cancer cells treated with TGF-β1. Western blot analysis revealed that TGF-β1 treatment significantly downregulated E-cadherin expression and upregulated vimentin expression in both these cell lines. TM4SF1 protein levels were also reduced following treatment with TGF-β1 for 48 h (Fig. 3B).