For gene expression analysis, 111 gastric cancer and 20 normal gastric tissues were collected from Ruijin Hospital. Total RNA was extracted using TRIzol reagent (Invitrogen, Carlsbad, CA, USA), according to the manufacturer’s instructions. Gene expression profiling was conducted using U133 plus 2.1 array (Affymetrix, Santa Clara, CA, USA).

Gastric cancer tissues, confirmed by pathological diagnosis, were obtained from 89 patients who underwent radical resection for gastric cancer between 2006 and 2008 at the Department of Surgery, Ruijin Hospital, Shanghai, China. The corresponding non-tumor gastric tissues were obtained at least 6 cm from the tumors. All tissue samples were formalin-fixed and paraffin-embedded. TNM staging was classified based on the criteria of the American Joint Committee on Cancer (AJCC, 7th edition) for gastric cancer. The present study was approved by the Shanghai Jiao Tong University Medical School Institutional Review Board.

Immunohistochemistry staining was performed by using a highly sensitive streptavidin-biotin-peroxidase detection system with gastric cancer tissue microarrays. Rabbit monoclonal anti-FRZB (working dilution 1:100) was purchased from LifeSpan Biosciences (Seattle, WA, USA) and Rabbit anti-β-catenin (working dilution 1:100) was purchased from Cell Signaling (Danvers, MA, USA). Immunolabeling was conducted using Envision+ Rabbit Polymer (Dako Carpinteria, CA, USA). The slides were counterstained with hematoxylin and coverslipped.

The histology of the samples was examined by two histopathologists independently without knowing the clinicopathological information. We scored the slides according to a previous publication (15). The percentage of positive tumor cells was assigned to 5 categories: ≤5% (0), 5–25% (1), 25–50% (2), 50–75% (3), and ≥75% (4). ≤5% positive cells were used as the cutoff to define negative tumors. The intensity of immunostaining was scored as: weak (1), moderate (2), and strong (3). The percentage of positivity of tumor cells and staining intensity were multiplied to produce a weighted score for each tumor specimen. The intensity scores were grouped as low (which included scores 0 to +4) and high (which included scores +6 and +12).

Human gastric cancer cell line MKN45 was obtained from Shanghai Institute of Cell Biology, Chinese Academy of Sciences. The cells were grown in RPMI-1640 medium containing 10% fetal bovine serum (FBS), penicillin and streptomycin (Gibco-BRL, Gaithersburg, MD, USA).

Cell proliferation was assessed using the Cell Counting Kit-8 (CCK-8; Dojindo Laboratories, Kumamoto, Japan) assay. Soft agar colony formation assay was performed by using 0.3% agar in complete medium (RPMI-1640 medium containing 10% FBS) with cells as the feeder layer and 0.6% agar in complete medium as the bottom layer.

The sequence used for construct of FRZB-shRNA was 3′-GGAGATTCTAAAGTCCTCTTTCAAGAGAAGAGGACTTTAGAATCTCC-5′. The shRNAs were cloned into pGPU6/Neo (Shanghai GenePharma, Shanghai, China). Plasmids were transfected into gastric cancer cells using Lipofectamine 2000 (Invitrogen). G418 was used for stable clone selection.

RT-PCR was carried out and the set of primers for FRZB was: F, 5′-GAGGAGCTGCCAGTGTACGAC-3′ and R, 5′-GAAAATCAGCTCCGTCCGC-3′; GAPDH: F, 5′-GGACCTGACCTGCCGTCTAG-3′ and R, 5′-GTAGCCCAGGATGCCCTTGA-3′ respectively (11). Primers for Ecad were: F, 5′-TTCCCTCGACACCCGATTCA-3′ and R, 5′-CCAGAAACGGAGGCCTGATG-3′; Ncad: F, 5′-CCCGGTTCATTTGAGGGCA-3′ and R, 5′-GGCATTGGGATCGTCGCAT-3′; MMP7: F, 5′-GTCTCTGGACGGCAGCTATG-3′ and R, 5′-TAGTCCTGAGCCTGTTCCCA-3′; cyclin D1: F, 5′-GATGCCAACCTCCTCAACGA-3′ and R, 5′-GGAAGCGGTCCAGGTAGTTC-3′. Relative density of target gene levels were measured by the following equation: Relative density = density of band of target genes/density of GAPDH.

Whole cell lysates were harvested using RIPA cell lysis buffer supplemented with a protease inhibitor cocktail (Sigma). A total of 50 μg protein were separated by SDS-polyacrylamide gel electrophoresis and blotted onto 0.22 μm polyvinylidene difluoride membranes (Millipore, Billerica, MA, USA). For extracting nuclear protein, nuclear extract kit (Active Motif, Carlsbad, CA, USA) was used following the manufacturer’s instructions. Antibodies against FRZB (LifeSpan) were used at 1:1,000 dilutions. Antibodies against GAPDH (Sigma) were used at a 1:5,000 dilution. Antibodies against Lamin B1 and β-catenin (Cell Signaling, Boston, MA, USA) were used at 1:1,000 dilutions. The signals were visualized using Odyssey-Sa model 9260 (Li-COR, Lincoln, NE, USA) and images were captured and managed using Odyssey-Sa Infrared Image System (Li-COR).

Cells were trypsinized in the presence of EDTA, washed twice in PBS and suspended at a concentration of 2.5×10⁵ cells/ml in DMEM with 10% FBS. Cell culture dishes (diameter, 10 cm) were pre-coated with poly-2-hydroxyethyl-methacrylate (poly-HEMA; Sigma, St. Louis, MO, USA) to block cell attachment. Same amount of MKN45 (parental), vector-only (MKN45/vector) and FRZB-shRNA (MKN45/FRZB-KD) cells were seeded and cultured for 24 h (16). Cells were observed using IX71 microscope (Olympus, Tokyo, Japan). Images were captured using Digital Sight DS-U1 (Nikon, Tokyo, Japan) and NIS elements F3.0 software was used (Nikon).

Cell migration was analyzed by a Transwell chamber assay. Cell invasion assays were performed using BioCoat Matrigel Invasion Chambers (BD Biosciences, Franklin Lakes, NJ, USA) and 10% FBS was used as the chemoattractant. Cells on the lower surface of the insert were fixed and stained followed by counting under a light microscope. Cells were visualized using BX50 microscope (Olympus). Images were captured using Digital Sight DS-U2 (Nikon) and NIS elements F3.0 software was used (Nikon).

For IHC staining, the differences in clinicopathological features between the different groups were determined using Pearson’s χ² test. P<0.05 was considered to indicate a statistically significant difference. The significance of differences between experimental groups was analyzed using the Student’s t-test and two-tailed distribution. Statistical Package for the Social Sciences version 13.0 (SPSS, Inc., Chicago, IL, USA) was used for all statistical analyses.

As demonstrated by previous studies (3,5), FRZB is associated with the activity of β-catenin. To investigate the correlation between FRZB and β-catenin, we first analyzed mRNA levels of FRZB and β-catenin in non-tumor and tumor gastric tissues. cDNA microarray dataset using a cohort of 111 gastric cancer tissues and 20 normal gastric mucosal tissues was analyzed. This analysis demonstrated that FRZB mRNA levels were higher in gastric cancer tissues than in non-tumor tissues (Fig. 1A). On the contrary, β-catenin mRNA levels were lower in gastric cancer tissues than in non-tumor tissues (Fig. 1B). Pearson correlation analysis revealed that the negative correlation between FRZB and β-catenin levels was statistically significant in gastric cancer tissues (Corr=−0.263, P=0.005) but not in normal gastric tissues (Corr=−0.063, P=0.402) (Table I). These data showed a negative correlation between FRZB and β-catenin in gastric cancer tissues.

To investigate the possible role of FRZB in regulating the function of β-catenin, we performed IHC staining using the same cohort of specimens as we used for FRZB staining. We discovered that FRZB was correlated with the sub-cellular localization of β-catenin, which might suggest the activation of β-catenin related signaling pathway. Among these 90 cases, 14 cases displayed membrane staining of β-catenin, while 18 cases showed nuclear β-catenin. As shown in Table II, membrane β-catenin tended to exist in the high FRZB expression group (11/14 cases), while nuclear β-catenin tended to exist in the low FRZB expression group (13/18 cases). The FRZB protein levels were associated with β-catenin membrane location with statistical significance (P=0.005). Although not statistically significant (P=0.112), gastric cancers in the FRZB-low group had more nuclear β-catenin. As shown in Fig. 2A, strong FRZB expression was shown in an intestinal-type gastric cancer specimen (upper left) and membrane β-catenin was also observed in the same case (upper right, black arrows). Weak FRZB expression was shown in a diffuse-type gastric cancer specimen (lower left) and nuclear β-catenin translocation was observed in the same case (lower right, red arrows) (Fig. 2B). These data suggested FRZB expression levels were negatively correlated with β-catenin levels and strongly associated with sub-cellular location of β-catenin.

Since we observed strong associations between FRZB and β-catenin in gastric cancer tissue, we next investigated whether the alteration of FRZB levels changed the activity of β-catenin. Thus, we established an FRZB-knockdown (FRZB-KD) model by stable transfection of FRZB-specific shRNA into gastric cancer cells. FRZB mRNA and protein levels were significantly inhibited by shRNA (Fig. 3A). We first examined the effect of FRZB knockdown on cell growth using CCK-8 analysis. FRZB knockdown significantly increased cell growth rate from day 2 after plating in monolayer culture (Fig. 3B). We then examined whether FRZB-KD had an effect on clonogenicity, using an in vitro assay to investigate tumorigenicity. We performed soft agar culture using parental cells and cells stably transfected with either control-vector or FRZB-shRNA-vector. After 14 days of culture, we observed more clones growing in soft agar in FRZB-KD cells than vector control and parental cells (Fig. 3C, upper). These data suggested that FRZB knockdown increased gastric cancer cell growth.

Metastasis is a multi-step process which includes the capacity of detaching from original site, escaping from anoikis and forming secondary tumors (17). We next employed suspension culture assay to examine the anti-anoikis activity, which reflected anchorage-independent growth ability and correlated with metastatic ability of tumor cells. After 24 h of suspension culture, FRZB-KD cells showed bigger cell aggregation clusters than the other two control groups (Fig. 4A). Apoptosis analysis also showed a lower apoptotic rate in FRZB-KD cells which was consistent with the results from suspension culture (Fig. 4B). These data suggested that anoikis resistance was enhanced by FRZB knockdown.

In addition to anoikis, we observed epithelial-to-mesenchymal (EMT) morphological changes in FRZB-KD cells. FRZB-KD cells exhibited scattered clone, while the vector-only and parental cells grew intense clone which showed strong cell interactions (Fig. 5A). We further performed in vitro migration and Matrigel invasion assay to investigate the effect of FRZB-KD on cell motility. As expected, FRZB-KD cells migrated and invaded faster than the other two control cells (Fig. 5B). These data suggested that FRZB knockdown promoted metastatic ability in gastric cancer cells.

As we discovered earlier that FRZB and β-catenin showed negative correlation in gastric cancer tissues, we used an FRZB-KD gastric cancer cell model to further validate this correlation. We first found nuclear β-catenin level was increased in FRZB-KD cells and total β-catenin level was also slightly increased (Fig. 6A). We further examined the expression levels of EMT markers and downstream targets which are known to be altered by β-catenin signaling, such as E-cadherin, N-cadherin, MMP7 and cyclin D1. As expected, levels of the epithelial marker E-cadherin were downregulated while the levels of the mesenchymal marker N-cadherin were upregulated by FRZB knockdown (Fig. 6B). MMP7 and cyclin D1, two direct downstream targets of β-catenin, were upregulated by FRZB knockdown. Our data indicated activation of β-catenin signaling in FRZB knockdown gastric cancer cells.